Cave Prop PDF

A MODEL FOR CAVE PROPAGATION AND SUBSIDENCE
ASSESSMENT IN JOINTED ROCK MASSES
Bre-Anne Sainsbury
B.E. (Geological Engineering) Royal Melbourne Institute of Technology
M.E. (Mining) The University of New South Wales
A Thesis submitted to The University of New South Wales in fulfilment of the

requirements for the degree Doctor of Philosophy
August 2012
ORIGINALITY STATEMENT
I hereby declare that this thesis is my own work and that, to the best of my
knowledge, it contains no materials previously published or written by another
person, or substantial proportions of material which have been accepted for the
award of any other degree or diploma at UNSW or any other educational
institution, except where due acknowledgment is made within the thesis. Any
contribution made to the research by others, with whom I have worked at UNSW
or elsewhere, is explicitly acknowledged. I also declare that the intellectual content
of the thesis is the product of my own work unless otherwise acknowledged.
COPYRIGHT STATEMENT
I hereby grant The University of New South Wales or its agents the right to archive
and to make available my thesis or dissertation in whole or part in the University
libraries in all forms of media, now or here after known, subject to the provisions
of the Copyright Act 1968. I retain all proprietary rights, such as patent rights. I
also retain the right to use in future works (such as articles or books) all or part of
this thesis or dissertation. I also authorise University Microfilms to use the 350
word abstract of my thesis in Dissertation Abstract International. I have either
used no substantial portions of copyright material in my thesis or I have obtained
permission to use copyright material.
AUTHENTICITY STATEMENT
I certify that the Library deposit digital copy is a direct equivalent of the final
officially approved version of my thesis. No emendation of content has occurred
and if there are any minor variations in formatting, they are the result of the
conversion to digital format.
ACKNOWLEDGEMENTS
The Author wishes to thank the following persons/organisations for directly and
indirectly providing assistance, support and guidance during the research.
Dr. David Sainsbury, my long suffering husband, for encouraging me to complete

this body of work and providing support along the way. In addition to being a
great husband and father you are also a remarkable engineer.
The research presented herein builds upon the initial work completed by Dr. Loren
Lorig, Dr. Mark Board, Dr. Peter Cundall and Dr. Matthew Pierce, from Itasca
Consulting Group, completed during the International Caving Study (ICS I and ICS
II) and Mass Mining Technology (MMT I) Project. My thanks for their initial work
and their ongoing support, input and interest in caving mechanics.
Thanks to the Mass Mining Technology (MMT II) Sponsors for providing financial
support to complete this body of work and for their provision of data for some of
the case-study applications. In particular, thanks to Dr. Andrew Haile and Dr.
Jonny Sjoberg who were the caving mechanics area monitors and devoted many
hours to reviewing this research.
I would also like to thank Professor Bruce Hebblewhite and Dr. Rudrajit Mitra from
the Mining Engineering Department of the University of New South Wales who
supervised the research. My thanks for your continued support and guidance.
ABSTRACT
Cave mining methods allow for the bulk extraction of large, low grade orebodies in
a cost effective manner. The fundamental mechanics of caving involves the self-
propagating yield (failure) of an in situ rock mass in response to production draw
from a mining horizon located at depth. Since the inception of large-scale
mechanised cave mining methods in the iron ore mines of northern Michigan, USA,
during the early part of the 20th century, researchers have sought to understand
and predict the nature of cave propagation through simple one-dimensional
volume based relationships and empirical methods. Although historically these
methods have successfully been applied to many cave operations, numerical
modelling is considered to be able to provide a more fundamental, rigorous and
robust assessment of cave propagation behaviour now and in the future.
A numerical model for cave propagation and subsidence assessment has been
developed based on fundamental rock mass behaviour and the development of
numerical modelling techniques. Unlike most existing techniques, the cave
volume is not introduced manually into the model; rather it is allowed to develop
based on the specified mass-based production schedule, evolving stress conditions
and the simulated constitutive behaviour of the rock mass. In doing so, hang-ups,
over-breaks and rapid advance rates can all be predicted.
The resulting numerical model is able to accurately capture rock mass strength
and deformation modulus anisotropy and scale effects as well as the effect of large-
scale discontinuities on cave propagation behaviour. In addition, the strain-
softening and bulking behaviour during the complex process of caving induced
yield and mobilisation is also considered. A production draw algorithm has been
developed that accurately reflects the mass withdrawn and drawpoint production
variability for all cave mining methods; block, panel and sub-level caving. This
algorithm is complemented by an algorithm that updates the evolving ground
surface profile to reflect the development of a crater. The methodology has been
applied to four large-scale case study back-analyses that provide validation of the
numerical techniques and assessment criteria.
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses i
TABLE OF CONTENTS
1 INTRODUCTION ............................................................................................................................................ 1
1.1 Cave Mining Method .......................................................................................................................... 1
1.2 Development of the Cave Mining Method ................................................................................ 3
1.3 Caving Mechanics ............................................................................................................................... 5
1.4 Terminology....................................................................................................................................... 10
1.5 Thesis Outline.................................................................................................................................... 14
2 DEVELOPMENT OF CAVE PROPAGATION AND ROCK MASS MODELLING
TECHNIQUES .............................................................................................................................................. 16
2.1 Cave Propagation Assessment ................................................................................................... 16
2.1.1 Analytical ................................................................................................................................... 16
2.1.2 Empirical .................................................................................................................................... 18
2.1.3 Numerical .................................................................................................................................. 21
2.1.3.1 Two-Dimensional Elastic Models .............................................................................. 22
2.1.3.2 Two-Dimensional Plasticity Models ......................................................................... 24
2.1.3.3 Axis-Symmetric Strain-softening Models ............................................................... 29
2.1.3.4 Two-Dimensional Distinct Element Models.......................................................... 32
2.1.3.5 Three-Dimensional Distinct Element Models....................................................... 33
2.1.3.6 Three-Dimensional Continuum Methods ............................................................... 36
2.1.3.7 Hybrid Techniques .......................................................................................................... 41
2.1.4 Summary .................................................................................................................................... 42
2.2 Rock Mass Modelling Techniques ............................................................................................. 44
2.2.1 Introduction.............................................................................................................................. 44
2.2.2 Hoek-Brown Strength Criterion ....................................................................................... 46
2.2.2.1 Cohesion and Tension Softening................................................................................ 47
2.2.2.2 Estimation of Critical Plastic Strain ( ps
crit )........................................................... 51
2.2.2.3 Mesh Dependency ............................................................................................................ 52

2.2.2.3.1 Cosserat Theory ......................................................................................................... 53
2.2.2.3.2 Standard Regularisation Method ........................................................................ 53
2.2.2.3.3 Scale-Dependent Brittleness ................................................................................ 54
2.2.2.3.4 Localisation and Bifurcation ................................................................................. 57
2.2.2.4 Independent Tension Softening ................................................................................. 58
2.2.2.5 Friction Hardening .......................................................................................................... 60
2.2.2.6 Summary .............................................................................................................................. 63
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses ii
2.2.3 The Synthetic Rock Mass (SRM) Modelling Approach ............................................ 66

2.2.3.1 Application of Synthetic Rock Mass Modelling for Cave Propagation
Assessment ......................................................................................................................... 71
2.2.3.2 Summary .............................................................................................................................. 76
3 RESEARCH OUTLINE ............................................................................................................................... 78
3.1 Objectives ............................................................................................................................................ 78
3.2 Methodology ...................................................................................................................................... 80
3.2.1 Simulation of Rock Mass Response to Cave Propagation ...................................... 81
3.2.1.1 Rock Mass Cohesion/Tension Softening and Post-Peak Brittleness .......... 81
3.2.1.2 Rock Mass Dilation .......................................................................................................... 81
3.2.1.3 Rock Mass Deformation Modulus Softening ........................................................ 81
3.2.1.4 Simulation of Large-Scale Discontinuities ............................................................. 82
3.2.2 Production Draw Simulation ............................................................................................. 82
3.2.2.1 Mass-Based Production Draw Algorithm ............................................................... 82
3.2.2.2 Development of an Algorithm to Update Ground Surface Profile ................ 83
3.2.2.3 Sub-Level Caving Algorithm ........................................................................................ 83
3.2.3 Validation................................................................................................................................... 83
4 DEVELOPMENT OF A CAVE PROPAGATION DEMONSTRATION MODEL ......................... 85
4.1 Geomechanical Conditions........................................................................................................... 85
4.2 Production Draw Simulation ...................................................................................................... 86
4.3 Cave Propagation Sensitivity Studies ...................................................................................... 87
4.3.1 Effect of Rock Mass Peak Strength on Cave Propagation ....................................... 87
4.3.2 Effect of Post-Peak Softening Rate on Cave Propagation....................................... 91
4.3.3 Effect of Estimation of mi value on Cave Propagation ............................................. 93
4.3.4 Effect of Stress/Depth on Cave Propagation ............................................................... 95
4.4 Summary ............................................................................................................................................. 97
5 DEVELOPMENT OF THE UBIQUITOUS JOINT ROCK MASS MODEL (UJRM) .................... 98
5.1 Establishment of a Standard Laboratory Environment ................................................. 101
5.1.1 Sample Geometry and Generation.................................................................................101
5.1.2 Sample Zone Resolution ....................................................................................................103
5.1.3 Sample Loading Conditions ..............................................................................................104
5.1.4 Large Strain/Small-strain Calculation Mode.............................................................106
5.2 Calibration of UJRM Response.................................................................................................. 108
5.2.1 Summary of SRM Responses............................................................................................108
5.2.1.1 Intact Calibration............................................................................................................108
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses iii
5.2.1.2 Discrete Fracture Network.........................................................................................109

5.2.1.3 Estimated Joint Strength .............................................................................................111
5.2.1.4 SRM Simulation Results ...............................................................................................112
5.2.2 Calibration of Deformation Modulus and Poisson’s ratio....................................112
5.2.3 Calibration of Matrix Friction, Cohesion and Tension ..........................................112
5.2.4 Calibration of Ubiquitous Joint Properties ................................................................113
5.2.5 Calibration of Critical Plastic Strain ( ps
crit ) ..............................................................113
5.2.6 Calibrated Laboratory Stress-Strain Curves .............................................................114

5.3 Application and Validation of the UJRM Methodology ................................................... 117
5.3.1 Calibrated SRM-UJRM in Laboratory Environment ...............................................117
5.3.1.1 Calibration of Intact Response..................................................................................117
5.3.1.2 Selection of Joint Properties ......................................................................................120
5.3.1.3 Development and Validation of a Discrete Fracture Network ....................121
5.3.1.4 Calibrated Continuum Responses ...........................................................................123
5.3.2 UJRM Large-Scale Response.............................................................................................126
5.4 Summary ........................................................................................................................................... 128
6 CONSIDERATION OF THE VOLUMETRIC CHANGES THAT ACCOMPANY CAVE
PROPAGATION ......................................................................................................................................... 130
6.1 Rock Mass Density......................................................................................................................... 130
6.2 Rock Mass Dilation ........................................................................................................................ 132
6.2.1 Implementation of Non-Constant Dilation in the Cave Demonstration
Model .........................................................................................................................................136
6.3 Deformation Modulus .................................................................................................................. 138
6.3.1 Implementation of Non-Linear Deformation Modulus Softening in the
Cave Demonstration Model ..............................................................................................143
7 IMPACT OF LARGE-SCALE DISCONTINTIES ON CAVE PROPGATION AND
SUBSIDENCE BEHAVIOUR .................................................................................................................. 145
7.1 Subsidence Behaviour ................................................................................................................. 145
7.2 General Characteristics of Caving Induced Subsidence ................................................. 149
7.3 Conceptual Models of Caving Induced Subsidence .......................................................... 150
7.3.1 Block Caving ...........................................................................................................................150
7.3.2 Chimney Caving.....................................................................................................................151
7.3.3 Plug Caving ..............................................................................................................................152
7.4 Subsidence Features Related to Cave Mines ...................................................................... 154
7.4.1 Description of Active Subsidence Features ...............................................................154
7.4.1.1 Caved Rock Zone.............................................................................................................154
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses iv
7.4.1.2 Zone of Large-Scale Fracturing.................................................................................155

7.4.1.3 Small-Scale Displacement Zone (Continuous Zone of Subsidence)...........156
7.4.1.4 Stable (Elastic) Zone .....................................................................................................157
7.4.2 Long-Term Time-Dependent Subsidence ..................................................................158
7.4.2.1 Residual Subsidence .....................................................................................................158
7.4.2.2 Sub-Surface Erosion ......................................................................................................159
7.5 Effect of Large-Scale Discontinuities on Subsidence Limits ........................................ 161
7.5.1 Fault Impacted Caving ........................................................................................................164
7.5.1.1 San Manuel Mine ............................................................................................................164
7.5.1.2 Ridgeway Deeps Sub-Level Cave .............................................................................166
7.5.1.3 Questa Mine ......................................................................................................................167
7.5.1.4 Henderson Mine .............................................................................................................168
7.5.1.5 Kimberly Mine .................................................................................................................169
7.5.1.6 Summary ............................................................................................................................169
7.6 Fault Properties .............................................................................................................................. 171
7.7 Numerical Simulation of Large-Scale Discontinuities in the Cave
Demonstration Model ................................................................................................................ 172
7.7.1 Implicit Fault Representation .........................................................................................173
7.7.2 Explicit Fault Representation ..........................................................................................177
7.8 Summary ........................................................................................................................................... 180
8 DEVELOPMENT OF A PRODUCTION DRAW ALGORITHM..................................................... 181
8.1 Influence of Production Schedule on Cave Propagation Behaviour ......................... 181
8.1.1 Impact of Production Draw Strategy in the Demonstration Model .................182
8.2 Influence of Rock Mass Bulking Behaviour on Cave Propagation Behaviour ....... 184
8.2.1 Impact of Bulking Factor in the Cave Demonstration Model .............................184
8.3 Limitations to Height of Draw Scheduling .......................................................................... 186
8.4 Development of a Mass-Based Production Schedule ...................................................... 188
8.4.1 Simulation of Undercutting ..............................................................................................188
8.4.2 Simulation of Production Draw ......................................................................................189
8.4.3 Selection of Maximum Draw Velocity (Vmax) .............................................................190
8.4.4 Development of a Tonnes Based Production Cut-Off Algorithm ......................192
8.5 Summary ........................................................................................................................................... 192
9 DEVELOPMENT OF AN ALGORITHM TO CONSIDER EVOLVING GROUND SURFACE
PROFILE ...................................................................................................................................................... 195
9.1 Impact of Topography on Subsidence Limits..................................................................... 195
9.2 Toppling Failure Mechanism .................................................................................................... 197
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses v
9.3 Limitations of the Small-Strain Numerical Approach..................................................... 197

9.4 Development of an Algorithm to Simulate Crater Development ............................... 199
10 DEVELOPMENT OF A SUB-LEVEL CAVING ALGORITHM ....................................................... 205
10.1 Sub-Level Caving Mining Method ........................................................................................... 205
10.2 Simulation of Blast Damage ....................................................................................................... 206
10.3 Mobilisation of a Previous Sub-Level .................................................................................... 208
10.4 Incremental Mass-Based Calculation .................................................................................... 209
11 CASE STUDY VALIDATION: CAVING INDUCED FAILURE OF THE PALABORA
OPEN PIT .................................................................................................................................................... 210
11.1 Background ...................................................................................................................................... 210
11.2 Geomechanical Conditions......................................................................................................... 211
11.3 In situ Stress .................................................................................................................................... 213
11.4 Production History ........................................................................................................................ 214
11.5 Simulation Results......................................................................................................................... 215
11.5.1 Cave Initiation ........................................................................................................................215
11.5.2 Yielding of the Crown Pillar – Q4 2002 .......................................................................217
11.5.3 Cave Break-Through – Q1 2004 .....................................................................................218
11.5.4 North Wall Failure – Q4 2004 .........................................................................................220
11.5.4.1 North Wall Failure Mechanism ..........................................................................221
11.6 Summary ........................................................................................................................................... 223
12 CASE STUDY VALIDATION : STRUCTURALLY CONTROLLED CAVING AT THE
HENDERSON MINE................................................................................................................................. 225
12.1 History of the Henderson Mine ................................................................................................ 225
12.2 Model Geometry and Production Schedule ........................................................................ 226
12.3 Material Properties and Pre-Mining Stresses .................................................................... 228
12.4 Simulation Results......................................................................................................................... 229
12.5 Summary ........................................................................................................................................... 232
13 CASE STUDY VALLIDATION : CAVING INDUCED SUBSIDENCE AT THE
ABANDONED GRACE MINE PANEL CAVE..................................................................................... 233
13.1 Background ...................................................................................................................................... 233
13.2 Geomechanical Properties ......................................................................................................... 235
13.3 Local Geology................................................................................................................................... 235
13.4 Pre-Mining Stress State ............................................................................................................... 237
13.5 Caving Induced Subsidence ....................................................................................................... 237
13.5.1 Evolution of Subsidence Crater and Trough .............................................................237
13.5.2 Visual Observation of the Limit of Large-Scale Surface Cracking ....................239
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses vi
13.5.3 Conceptual Model of Subsidence Formation at the Grace Mine........................240

13.5.4 Modelling Methodology .....................................................................................................241
13.5.5 Predicted Evolution of Cave Mobilised and Yield Zones ......................................242
13.5.6 Model Validation ...................................................................................................................244
13.6 Summary ........................................................................................................................................... 245
14 CASE STUDY VALIDATION : CAVING INDUCED SUBSIDENCE AT THE
KIIRUNAVAARA LAKE OREBODY SLC............................................................................................ 247
14.1 Introduction ..................................................................................................................................... 247
14.2 Historical Mining Record ............................................................................................................ 247
14.3 Evolution of Surface Subsidence ............................................................................................. 248
14.3.1 Limits of Large-Scale Fracturing / Yield Zone ..........................................................249
14.3.2 Limits of Continuous Deformation ................................................................................250
14.4 Numerical Simulation of Caving Induced Subsidence .................................................... 250
14.4.1 In situ Stress............................................................................................................................251
14.4.2 Rock Mass Properties .........................................................................................................251
14.4.3 Production Schedule ...........................................................................................................251
14.4.4 Simulation Results ...............................................................................................................252
14.5 Summary ........................................................................................................................................... 256
15 CONCLUSIONS AND RECCOMENDATIONS................................................................................... 257
15.1 Summary of Original Contributions ....................................................................................... 258
15.1.1 Rock Mass Behaviour..........................................................................................................258
15.1.1.1 Development of the Ubiquitous Joint Rock Mass (UJRM) Model..........258
15.1.1.2 Consideration of the Volumetric Changes that Accompany Cave
Propagation ......................................................................................................................258
15.1.1.3 Impact of Large-Scale Discontinuities on Cave Propagation and
Subsidence Behaviour ..................................................................................................259
15.1.2 Production Simulation .......................................................................................................259
15.1.2.1 Development of a Mass-Based Production Schedule ................................259
15.1.2.2 Development of a Sub-Level Caving Algorithm ...........................................259
15.1.2.3 Development of an Algorithm to Consider Evolving Surface Profile..260
15.2 Validation .......................................................................................................................................... 260
15.3 Recommendations for Further Work .................................................................................... 260
15.3.1 Rock Mass Behaviour..........................................................................................................260
15.3.1.1 Ubiquitous Joint Rock Mass .................................................................................260
15.3.1.2 Time-Dependent Processes .................................................................................261
15.3.2 Numerical Techniques........................................................................................................261
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses vii
15.3.2.1 Small-Strain Calculation Model ..........................................................................261

15.3.2.2 Interactive Draw.......................................................................................................261
15.3.2.3 Hardware.....................................................................................................................261
15.3.3 Validation.................................................................................................................................261
16 REFERENCES ............................................................................................................................................ 262
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses viii
TABLE OF FIGURES
Figure 1. Location of some historical and currently operating caving mines

around the world. ...................................................................................................................... 1
Figure 2. Schematic representation of a typical block cave mine (modified after
Atlas Copco, 2011). ................................................................................................................... 2
Figure 3. (a) Laubscher’s stability chart showing E26 predicted caveability (b)
section showing cave stall at Northparkes E26 prior to plug failure
(modified after Ross and van As, 2005). .......................................................................... 4
Figure 4. Typical caving stress-paths representing stress and gravity caving
mechanisms. ................................................................................................................................ 6
Figure 5. Conceptual stress-state development around a propagating cave
(modified after Pierce, 2010). .............................................................................................. 7
Figure 6. Conceptual diagram of showing the main behavioural regions of a
propagating cave based on underground observations and
instrumentation. ..................................................................................................................... 10
Figure 7. Conceptual schematic diagram showing the main behavioural regions
of a cave that has propagated through to the ground surface. ............................ 12
Figure 8. Analytical solution: relationship between thickness of caved ore
withdrawn , thickness of in-place ore extracted and rise of the caving
brow (after Panek, 1984). .................................................................................................. 16
Figure 9. Empirical method for predicting caveability: Laubscher’s stability
chart (after Laubscher, 1994). .......................................................................................... 19
Figure 10. Empirical method for predicting caveability: Extended Mathews
Stability Chart (after Trueman and Mawdesley, 2003). ......................................... 20
Figure 11. Impact of principal stress orientation in relation to an undercut as
defined by two-dimensional numerical modelling (after Palma and
Agarwal 1973). ........................................................................................................................ 22
Figure 12. Conceptual diagram of effect of principal stress direction and undercut
footprint dimensions on caveability. .............................................................................. 23
Figure 13. Development of two-dimensional numerical modelling approaches for
cave propagation analysis. (a) Model mesh (b) section through the
mining geometry (c) simulated undercutting process (d) contours of
resultant mobilised strength (after Barla et al., 1980). .......................................... 25
Figure 14. Methodology for the application of a continuum based numerical
model for the prediction of onset of caving (after Rech and Lorig,
1992). .......................................................................................................................................... 27
Figure 15. Determination of the Cave Propagation Factor at Northparkes E26 Lift
1 (after Karzulovic and Flores 2003). .......................................................................... 28
Figure 16. Development of axis-symmetric numerical methods for cave
propagation (a) axis-symmetric concept (b) evolution of the undercut
pressure and height (c) stepwise reduction of undercut pressure (d)
details of the pressure evolution with a simulated reduction step (after
Brown, 2003). .......................................................................................................................... 30
Figure 17. Development of cave geometry resulting from a two dimensional
strain-softening numerical simulation (after Lorig, 2000). .................................. 31
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses ix
Figure 18. Development of a discrete element model to study cave propagation

(a) particle clusters early in the caving process with superimposed
contact force chains (after Lorig et al., 1995). (b) particle clusters after
significant cave propagation showing internal fractures of blocks in the
caving zone chains (after Lorig et al., 1995). (c) forces arching around
the unstable rock mass (after Brown, 2003). ............................................................. 32
Figure 19. Large-scale (mine-wide) discrete element modelling of caving and
subsidence phenomena in three-dimensions. Cross section of
subsidence mass movement from block caving and simulated synthetic
rock mass triaxial test of PFC material (after Gilbride et al., 2005). ................. 34
Figure 20. Large-scale discrete element modelling of caving and subsidence
phenomena in three-dimensions. (after Sharrock et al., 2011). ......................... 35
Figure 21. Three-dimensional strain-softening, continuum models for cave
propagation (a) logic sequence to simulate caving (b) typical
simulation results (after Pierce and Lorig, 1998). .................................................... 37
Figure 22. Simulation of production draw from large-scale, three-dimensional
strain-softening continuum models based on velocities. ...................................... 38
Figure 23. Large-scale back-analysis of cave propagation behaviour at the
Northparkes E26 Lift 2 Mine. (after Pierce et al., 2006). ...................................... 39
Figure 24. Example of a mine-wide, three-dimensional, multi-scale simulation
(after Beck et al., 2011). ....................................................................................................... 40
Figure 25. Simulation of cave development using a hybrid, two-dimensional
approach (after Rogers et al., 2010). .............................................................................. 41
Figure 26. Measured rock strength-scale effect including large size specimens of
in situ test (after Pratt et al., 1972). ................................................................................ 44
Figure 27. Applicability of the Hoek-Brown empirical rock mass strength
criterion at different scales (after Li et al., 2008). .................................................... 45
Figure 28. Development of equivalent Mohr-Coulomb property estimates from a
fit to the Hoek-Brown curve. ............................................................................................. 47
Figure 29. Idealised stress-strain curves representing different material
behaviour used in numerical modelling. ...................................................................... 48
Figure 30. Stages of damage within a three-dimensional, strain-softening
specimen. ................................................................................................................................... 50
Figure 31. Summary of FLAC3D critical strain relation and data points used for
fitting. .......................................................................................................................................... 51
Figure 32. Post-peak response as a function of zone resolution controlled by
sample width. ........................................................................................................................... 55
Figure 33. Schematic diagram of a tensile failure mechanism that does not affect
cohesive strength. .................................................................................................................. 59
Figure 34. Development of equivalent linear Mohr-Coulomb strength parameters
based on a fit to the Hoek-Brown strength envelope. ............................................. 60
Figure 35. Schematic diagram of the mobilisation of the strength components
cohesion and friction (a) in the laboratory (b) around an underground
opening (after Hajiabdolmajid, Kaiser and Martin, 2002). ................................... 61
Figure 36. Implementation of the CWFS model in a two-dimensional numerical
model of a tunnel failure (after Barton and Pandey, 2011). ................................. 62
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses x
Figure 37. Example of simulated bi-linear, strain-softening response (after

Sainsbury et al., 2010). ......................................................................................................... 65
Figure 38. The Smooth Joint Contact Model. (after Mas Ivars et al., 2011). ........................ 68
Figure 39. Components of a Synthetic Rock Mass sample. (after Mas Ivars et al.,
2011). .......................................................................................................................................... 69
Figure 40. Three-dimensional response of a synthetic rock mass sample tested in
three-opposing directions under unconfined compression; (after
Sainsbury et al., 2009). ......................................................................................................... 70
Figure 41. Stress-path dependent Synthetic Rock Mass approach (a) stress path,
fitted peak-strength envelope (b) estimates of brittleness obtained
from SRM testing (after Pierce et al., 2006). ............................................................. 72
Figure 42. Validation of Synthetic Rock Mass response based on observed and
measured fracture modes and fragmentation (after Pierce et al., 2006)........ 73
Figure 43. Development of a large-scale caving model using stress-path
dependent Synthetic Rock Mass strengths (after Mas Ivars et al., 2011). ...... 75
Figure 44. Research methodology plan. ........................................................................................... 84
Figure 45. Development of a numerical demonstration model: geomechanical
conditions. ................................................................................................................................. 86
Figure 46. Hoek-Brown failure envelopes and simulated rock mass stress-strain
curves for the rock mass domains in the numerical demonstration
model. .......................................................................................................................................... 87
Figure 47. Empirical estimates of rock mass caveability for four rock mass
domains simulated in the numerical demonstration model. ............................... 88
Figure 48. Predicted cave propagation behaviour for variable peak strength rock
masses in the numerical demonstration model. ....................................................... 89
Figure 49. Simulated variable post-peak softening responses for the same peak
strength rock mass................................................................................................................. 91
Figure 50. Variation in cave propagation behaviour based on variable post-peak
softening rates simulated in the numerical cave propagation model. ............. 92
Figure 51. Hoek-Brown curves and equivalent bi-linear Mohr-Coulomb property
estimates for varying mi values........................................................................................ 93
Figure 52. Effect of estimates of mi on predicted cave propagation behaviour in
the numerical demonstration model.............................................................................. 94
Figure 53. Cave propagation results for increasing stress /depth in the numerical
demonstration model. ........................................................................................................ 96
Figure 54. Subiquitous constitutive model in FLA3D; assignment of matrix and
joint properties........................................................................................................................ 99
Figure 55. Stages of damage within a simulated UCS test on a subiquitous sample. .....100
Figure 56. Development of a UJRM sample (a) variation in sample size with equal
zone sizes; (b) joint assignment as a function of sample size. ...........................102
Figure 57. UJRM sample testing geometry (a) sample loading conditions (b)
orientation for anisotropy tests completed for each sample loading
condition. .................................................................................................................................103
Figure 58. Examples of (a) poor (b) low and (c) good mesh resolution required
for large-scale analysis of cave propagation. ............................................................104
Figure 59. Discontinuity shear strength in a triaxial cell (after Brady and Brown,
2006). ........................................................................................................................................105
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xi
Figure 60. Simulation of different boundary loading conditions on the response of

UJRM material in the numerical UCS test environment. ......................................106
Figure 61. Investigation of UJRM response as a result of small-strain/large-strain
calculation modes.................................................................................................................107
Figure 62. Joint orientations considered in the development of the DFN for (a)
carbonatite (b) micaceous pyroxenite (c) dolerite (d) foskorite (after
Sainsbury et al., 2008). .......................................................................................................110
Figure 63. Representation of DFN in a UJRM sample (a) actual DFN (b) DFN
represented in numerical model (after Sainsbury et al., 2008). .......................111
Figure 64. UJRM sample stress-strain responses (a) calibrated 40x80m
carbonatite UCS UJRM rock mass samples showing strength anisotropy
(b) calibrated 40x80m carbonatite triaxial UJRM rock mass samples
showing strength anisotropy. .........................................................................................114
Figure 65. Calibrated UJRM: SRM results at 5 MPa confinement for each lithology
at Palabora in three testing directions. .......................................................................115
Figure 66. UJRM UCS results for the carbonatite domain at Palabora compared to
SRM results at three different sample sizes in three loading directions. ......116
Figure 67. Calibrated stress-strain curves within PFC for three rock mass
domains. ...................................................................................................................................119
Figure 68. Domain 1 fracture network views: 18m REV edge length (after
Sainsbury, Mas Ivars and Darcel, 2008)......................................................................121
Figure 69. Domain 2 fracture network views: 40m edge length (after Sainsbury,
Mas Ivars and Darcel, 2008). ...........................................................................................122
Figure 70. Domain 3 Fracture Network views : 18m edge length (after Sainsbury,
Mas Ivars and Darcel, 2008). ...........................................................................................122
Figure 71. Quantification of GSI chart (after Cai et al., 2007). .................................................123
Figure 72. Domain 1 SRM test results and UJRM response represented in FLAC3D :
1 MPa triaxial tests (after Sainsbury, Mas Ivars and Darcel, 2008). ...............124
Figure 73. Domain 2 SRM test results and UJRM response represented in FLAC3D :
1 MPa triaxial tests (after Sainsbury, Mas Ivars and Darcel, 2008). ...............125
Figure 74. Domain 3 SRM test results and UJRM response represented in FLAC3D: 1
MPa triaxial tests (after Sainsbury, Mas Ivars and Darcel, 2008). ...................125
Figure 75. Cave propagation behaviour for varying joint orientations simulated in
the numerical demonstration model............................................................................127
Figure 76. Simulated porosity profile during propagation of a block cave. .......................130
Figure 77. Conceptual diagram of dilation associated with sliding along micro-
cracks and particles (after Zhao and Cai, 2010). .....................................................132
Figure 78. Typical stress-strain curve for uniaxial compression of brittle,
crystalline rock (after Rudnicki and Rice, 1975).....................................................132
Figure 79. Evolution of peak dilation estimate on a rock mass during cave
propagation using the Alejano and Alonso relation...............................................136
Figure 80. Implementation of a non-constant dilation relation and its impact on
cave propagation behaviour in the numerical demonstration model
compared to the simulation of a constant dilation angle.....................................137
Figure 81. Schematic linear relationship for rock mass deformation modulus
reduction based on Pierce et al. (2006) relation. ....................................................139
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xii
Figure 82. In situ rock mass deformation modulus versus GSI for Disturbance
Factors of 0, 0.5 and 1.0 (after Hoek and Diederichs, 2006). .............................139
Figure 83. Softened deformation modulus versus porosity for particulate matter
determined by laboratory testing..................................................................................140
Figure 84. Best-fit deformation modulus softening equation to compiled
laboratory test data. ............................................................................................................141
Figure 85. Typical deformation modulus softening curves of caving rock masses
using the non-linear softening relation.......................................................................142
Figure 86. Impact on cave propagation behaviour by implementing the non-linear
modulus softening relation in the cave demonstration model..........................143
Figure 87. Simulated evolution of the bulk modulus in the back of demonstration
model undercut; the linear and non-linear relations compared in the
cave demonstration model. ..............................................................................................144
Figure 88. Conceptual models of subsidence a) continuous subsidence (after
Kratzsch, 1983) b) discontinuous subsidence (after Whittaker and
Reddish, 1989). .....................................................................................................................146
Figure 89. Terminology used to describe subsidence features for block- and
panel-cave mines (modified after van As et al., 2003). .........................................148
Figure 90. Conceptual model of the development of block caving subsidence
(after Sainsbury and Lorig, 2005). ................................................................................151
Figure 91. Conceptual model of chimney cave development (Betourney et al.,
1994), b) surface expression of a chimney pipe in a kimberlite caving
operation (after van As et al., 2003). ............................................................................152
Figure 92. Plug subsidence mechanism at the Athens Mine in Michigan USA (after
Obert and Duvall, 1967). ...................................................................................................153
Figure 93. Geometry of Lift 1 cave a) before and b) after plug caving (after Pierce,
1999). ........................................................................................................................................153
Figure 94. Photo showing crater and caved rock zone at Henderson Mine (after
Lupo, 1998). ............................................................................................................................155
Figure 95. Photo showing large-scale surface cracking at Northparkes E26 Lift 1
Mine (after van As et al, 2003). .......................................................................................156
Figure 96. Photo showing tension crack within small-scale displacements at the
Kiirunavaara Mine (after Villegas, 2008). ..................................................................157
Figure 97. Simplified subsurface erosion mechanism (after Van der Merwe 1999). ....159
Figure 98. Photos of subsurface erosion pot holes (after Van der Merwe, 1999)...........160
Figure 99. Photo of sinkhole located outside the limit of large-scale cracking at
the abandoned Grace Mine (after Sainsbury and Lorig, 2005). ........................160
Figure 100. Schematic diagram of how crater shape can be modified by major
geological structure (after Stacey and Swart, 2001). ............................................162
Figure 101. Conceptual development of surface subsidence at the San Manuel Mine
(after Hatheway, 1966)......................................................................................................164
Figure 102. Plan view, section view of subsidence crater at the San Manuel Mine
(after Hatheway 1966).......................................................................................................165
Figure 103. Photos showing cave propagation controlled by weak vertical fault at
the Ridgeway Mine (Brunton, 2009). ...........................................................................166
Figure 104. Photo of Goathill Crater at the Questa Mine (after Gilbride et al., 2005). .....167
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xiii
Figure 105. Irregular cave growth along a weak intrusive contact at the Henderson
Mine 7210 Level (after Sainbury et al., 2011) ..........................................................168
Figure 106. Section through Kimberly Mine showing over-hang (after Laubscher,
2000). ........................................................................................................................................169
Figure 107. Simulation of subsidence crater formation for different two-
dimensional fault orientations (modified after Vyazmensky et al.,
2010). ........................................................................................................................................170
Figure 108. Estimated shear strength of filled discontinuities (after Wyllie and
Mah, 2007). .............................................................................................................................171
Figure 109. Conceptual mesh of implicit versus explicit technique for fault
representation. ......................................................................................................................172
Figure 110. Conceptual geological structures simulated in numerical
demonstration model. ........................................................................................................173
Figure 111. Simulated direct shear test; normal stress 10 MPa using ubiquitous
joints in FLAC3D. ......................................................................................................................174
Figure 112. Ubiquitous joint faults used to simulate faults within a cave-scale
model. ........................................................................................................................................174
Figure 113. Cross-section of mobilised zone (2m displacement) – implicit,
ubiquitous joint approach used to simulate conceptual discontinuity
surfaces.....................................................................................................................................176
Figure 114. Schematic diagram showing interface logic and how it can be used to
represent a discontinuity in a numerical model of caving. .................................177
Figure 115. Cross-section of mobilised zone (2m displacement) – explicit, interface
approach used to simulate conceptual discontinuity surfaces..........................179
Figure 116. Plan view of subsidence limits at the Grace Mine determined by
observations. ..........................................................................................................................182
Figure 117. Effect of draw strategy on the caveability of a rock mass in the
numerical demonstration model. ..................................................................................183
Figure 118. Cave simulation results for variable maximum bulking rates in the
numerical demonstration model. ..................................................................................185
Figure 119. Schematic representation of a HOD based schedule interpreted for
numerical mesh. ....................................................................................................................186
Figure 120. Representation of (a) typical production schedule (b) mining
increment schedule (c) improved drawpoint scheduling method. .................187
Figure 121. Identification of perimeter gridpoints for production draw simulation
in a numerical mesh. ...........................................................................................................189
Figure 122. Simulated large-scale laboratory tests at different applied loading
velocities and the impact on the sample strength response. .............................190
Figure 123. Impact of selection of draw velocity on cave propagation behaviour in
the numerical demonstration model............................................................................191
Figure 124. Schematic diagram of the mass-based production draw algorithm
developed. ...............................................................................................................................193
Figure 125. Example of evolving mobilised zone based on drawpoint tonnes
algorithm..................................................................................................................................194
Figure 126. Photo showing the effect of topography on subsidence crate at the
Questa Mine (after Blodgett, 2002). .............................................................................196
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xiv
Figure 127. Survey displacement map above Questa Mine D Orebody (after
Gilbride et al., 2005). ...........................................................................................................196
Figure 128. Schematic diagram showing the three primary modes of toppling
(after Goodman and Bray, 1976). ..................................................................................197
Figure 129. Schematic diagram showing the simulation of evolving surface crater
in small-strain calculation mode. ...................................................................................198
Figure 130. Schematic diagram of subsidence algorithm logic. ................................................200
Figure 131. Geometry and undercut footprint of test model used to validate crater
development algorithm. ....................................................................................................201
Figure 132. Subsidence limits predicted with/without surface update algorithm. ........202
Figure 133. Updated surface elevation in the model after the simulation of mining
with the surface update algorithm. ...............................................................................203
Figure 134. Vertical displacement simulated in the test model and the surfaces
zones that have been nulled to represent the development of the
surface crater. ........................................................................................................................204
Figure 135. Schematic diagram of sub-level caving algorithm logic. ......................................207
Figure 136. Conceptual model of the volumetric changes in the sub-level caving
algorithm logic. ......................................................................................................................209
Figure 137. Photo of the failure in north wall at the Palabora open pit.................................210
Figure 138. The spatial location of each of the rock mass domains and faults
throughout the Palabora model mesh. ........................................................................211
Figure 139. Location of large-scale structure simulated in the Palabora numerical
mesh. ..........................................................................................................................................212
Figure 140. Estimated in situ stress orientation and magnitude at Palabora based
on back-analysis of pit slope failure and stress measurement testing. .........213
Figure 141. Historical mining record at the Palabora block cave mine. ................................214
Figure 142. Observed seismicity at the Palabora Mine during cave initiation and
propagation.............................................................................................................................215
Figure 143. Numerical prediction of seismogenic zones during early production
simulation at the Palabora mine. ...................................................................................216
Figure 144. Numerical simulation - yielding of the crown pillar during Q4 2002.............217
Figure 145. Numerical simulation – cave breakthrough during Q1 2004.............................218
Figure 146. (a) Cave profiles at the Palabora Mine; April 2002 to December 2003
(after Glazer, 2006) compared to the simulated cave profile (b).....................219
Figure 147. Numerical simulation – north wall failure during Q4 2004. ..............................220
Figure 148. North wall failure: observed versus simulated limits. ..........................................221
Figure 149. Development of the pit slope failure mechanism at the Palabora Mine
at various stages of production. .....................................................................................222
Figure 150. Development of the Palabora block cave between 2003 and 2004 in
relation to fault structure..................................................................................................223
Figure 151. Cross section of the Henderson Mine (after Rech, 2001). ...................................225
Figure 152. Geological domains at the Henderson Mine a) plan view of weak
contact; b) 7210 Level yield zone during December 2007. ................................226
Figure 153. Development of the numerical model of the Henderson Mine a)
regional extents of model; b) existing cave volumes. ............................................227
Figure 154. Interface used to simulate the weak Seriate contact at the Henderson
Mine. ..........................................................................................................................................227
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xv
Figure 155. Simulated production schedule (cumulative solid rock height of draw)
based on actual draw heights. .........................................................................................228
Figure 156. Simulated evolution of the cave yield zone at Henderson compared to
underground instrument observations.......................................................................230
Figure 157. Comparison of modelled verses actual cave shape at break-through to
the overlying lift at the Henderson Mine. ...................................................................231
Figure 158. Comparison of cave initiation at the Henderson Mine simulated with
and without a weak structure. ........................................................................................231
Figure 159. Geometry of the Grace Mine a) extent of underground production
drives (plan view) b) two-dimensional schematic of panel caving at
Grace Mine (after Stafford, 2002). .................................................................................234
Figure 160. Photo showing present day subsidence lake at the Grace Mine. ......................234
Figure 161. Isometric view of orebody and surface topography at the Grace Mine. ........236
Figure 162. Photos showing the evolution of subsidence from 1963 to 1978 at the
Grace Mine...............................................................................................................................239
Figure 163. Limit of large-scale surface cracking observed at the Grace Mine during
2005 (after Sainsbury and Lorig, 2005). ....................................................................240
Figure 164. Conceptual model of subsidence formation at the abandoned Grace
Mine (after Sainsbury and Lorig, 2005). .....................................................................241
Figure 165. Development of a numerical model of the Grace Mine a) regional
extents of model b) simulated undercut footprint. ................................................242
Figure 166. Predicted evolution of cave mobilised and yield zones (looking south)
during simulation of production from the Grace Mine. ........................................243
Figure 167. Measured verses predicted vertical displacements from numerical
model of Grace Mine............................................................................................................244
Figure 168. Comparison of cave propagation results (a) HOD schedule - uniform
bulking factor of 0.2 (b) mass balance (tonnes-based) schedule. ....................245
Figure 169. Location and extent of the Lake and Main Orebody at the Kiirunavaara
Mine. ..........................................................................................................................................247
Figure 170. Subsidence regions at Kiirunavaara. ............................................................................248
Figure 171. Photos showing the development of a crater at northern extent of Lake
Orebody during 2006 and its subsequent enlargement. .....................................249
Figure 172. Fracture mapping at Kiirunavaara (a) plan above Lake Orebody (b)
section through Main Orebody (modified after Villegas et al., 2011). ...........249
Figure 173. Evolution of measured surface total displacement profile around the
Lake Orebody. ........................................................................................................................250
Figure 174. Regional extents of Kiirunavaara numerical mesh. ...............................................250
Figure 175. Simulated evolution of crater and limits of large-scale fracturing within
the numerical model of Kiirunavaara. .........................................................................252
Figure 176. Production rate versus simulated mobilised zone within the numerical
model of Kiirunavaara. .......................................................................................................253
Figure 177. Simulated evolution of total surface displacements within the
numerical model of Kiirunavaara compared to observations onsite. ............254
Figure 178. Plan view of simulated displacement and strain-based subsidence
criteria and subsidence zone of influence at the end of 2010 as
simulated in the numerical model of Kiirunavaara................................................254
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xvi
Figure 179. Plan view of simulated subsidence limits at the end of 2010 compared
to observations at Kiirunavaara. ....................................................................................255
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xvii
TABLE OF TABLES
Table 1. Documented yield zone propagation rates from caving operations

around the world (after Sainsbury and Sainsbury, 2010). ................................... 17
Table 2. Cave height as a function of brittleness (after Lorig, 2000). ................................ 54
Table 3. Estimate of Hoek-Brown and equivalent Mohr-Coulomb rock mass
strength properties for four simulated domains in the numerical
demonstration model. .......................................................................................................... 87
Table 4. Mean target intact rock block properties for the lithology at Palabora. .......109
Table 5. Measured joint frequencies and persistence from mapping at Palabora
(after Mas Ivars et al., 2008) ............................................................................................109
Table 6. Estimated joint properties for the rock mass domains at Palabora
(after Mas Ivars et al., 2008). ...........................................................................................112
Table 7. SRM-derived strengths for the rock mass domains at Palabora - triaxial
5-MPa confinement (after Mas Ivars et al., 2008)...................................................112
Table 8. Calibrated UJRM properties for the rock mass domains at Palabora. ............114
Table 9. Summary of laboratory test results for three rock mass domains. .................118
Table 10. Calibrated PFC micro-properties for three rock mass domains (after
Table 11. Calibrated intact foliation strength properties in PFC3D (after
Table 12. Estimated open joint strength properties for simulation of joints in
SRM sample (after Sainsbury, Mas Ivars and Darcel, 2008)...............................121
Table 13. Calibrated continuum material properties for seven rock mass
domains. ...................................................................................................................................124
Table 14. Dilation angle in large-scale triaxial tests on rock fill material (after
Marachi et al., 1972)............................................................................................................134
Table 15. Summary of terminology used to define discontinuous subsidence
(after Flores and Karzolovic, 2004). .............................................................................147
Table 16. Observed residual subsidence duration over longwall mines (after
Singh, 2003). ...........................................................................................................................158
Table 17. Conceptual fault shear strength and stiffness parameters represented
in numerical demonstration model. .............................................................................173
Table 18. Example of gridpoint velocity scaling based on variable production
draw. ..........................................................................................................................................190
Table 19. Rock mass properties used for the representation of the granite
domain. .....................................................................................................................................211
Table 20. Rock mass geomechanical properties of the porphyry at the Henderson
Mine. ..........................................................................................................................................228
Table 21. Conceptual fault shear strength and stiffness parameters estimated for
the seriate contact at the Henderson Mine. ...............................................................229
Table 22. Rock Mass Parameters used in the simulation of domains at the Grace
Mine. ..........................................................................................................................................237
Table 23. Pre-mining stress regime at 731m below surface at the Grace Mine. ............237
Table 24. Grace Mine production (after Eben, 2004 ). ..............................................................241
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xviii
ABBREVIATIONS
BF Bulking Factor
BPM Bonded Particle Model
CPF Cave Propagation Factor
CWFS Cohesion Weakening, Friction Strengthening
DFN Discrete Fracture Network
DPE Dissipated Plastic Energy
FEM Finite Element Model
GSI Geological Strength Index
HOD Height of Draw
HR Hydraulic Radius
ICS International Caving Study
MMT Mass Mining Technology
MRMR Mining Rock Mass Rating
REV Representative Elemental Volume
SJM Smooth Joint Model
SLC Sub Level Cave/ing
SRM Synthetic Rock Mass
TDR Time Domain Reflectometry
UCS Unconfined Compressive Strength
UJRM Ubiquitous Joint Rock Mass

A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses xix
SIGN CONVENTIONS
The following conventions have been used throughout this thesis to present
numerical modelling results:
Direct Stress Positive stresses indicate tension. Negative stress indicates

compression.
Shear Stress A positive shear stress points in the positive direction of the
co-ordinate axis. A right-handed co-ordinate system has
been used. +x is east. +y is north and +z is up.
Direct Strain Positive strain indicates tension. Negative strain indicates

compression.
Shear Strain A positive shear strain points in the positive direction of the
co-ordinate axis. A right-handed co-ordinate system has
been used. +x is east. +y is north and +z is up.
Pressure A positive pressure will act normal to and in the direction

towards the free surface.
Dip, Dip Direction A right-handed co-ordinate system has been used. +x is east.
+y is north and +z is up. Dip direction has been expressed as
positive clockwise from north (+y axis). Dip has been
expressed as positive down from the horizontal plane.
Vector Quantities The x, y and z components of vector quantities such as forces

displacements and velocities are positive when pointing in
the directions of the +x, +y and +z axes.
A Model for Cave Propagation and Subsidence Assessment in Jointed Rock Masses 1
1 INTRODUCTION
1.1 Cave Mining Method
Caving is a mass mining method, capable of high and sustained production rates
and is relatively low cost per tonne when compared to other mining methods. In
general, a uniform grade distribution and rock mass strength is required to assure
that the maximum potential of a deposit is achieved (Brady and Brown, 2006).
Presently, there are approximately twenty operating cave mines around the world
and many more in the planning stage. Figure 1 provides the locations of the most
notable caving mines that have been, or are currently in operation.
Figure 1. Location of some historical and currently operating caving mines around the
world.
The caving process involves undercutting (blasting a horizon of in situ rock mass)
and extraction of the broken rock from drawpoints on a production horizon
located at depth. When the plan area of the undercut footprint/active area reaches
a large enough dimension a self-sustained propagating cave will develop so long as
the ore is continued to be withdrawn. This is generally described as the critical
Hydraulic Radius (HR) which can be calculated through the ratio of the
undercut/active footprint area (m2) to the cumulative undercut/active footprint
perimeter length (m).
1 – Introduction
Cave mines require extensive infrastructure to be in place prior to any production.

Infrastructure usually includes access through a decline or shaft to an undercut
level and an extraction level that is located approximately 15-20 m below that.
Extraction level infrastructure must be designed to be stable throughout the life of
mine, often without previous experience of the large-scale rock mass response.
The three-dimensional nature of typical extraction level geometries, together with
the complex stress-redistribution around a propagating cave make completing an
accurate assessment of cave propagation and subsidence behaviour difficult. A
schematic diagram of a typical block cave mine layout is provided in Figure 2.
Figure 2. Schematic representation of a typical block cave mine (modified after Atlas
Copco, 2011).
There are three variations to the cave mining method that include block, panel and
sub-level caving. In block and panel caving operations, the ore is withdrawn from a
single mining horizon (extraction level). The transition of the ore from an in situ
rock mass to a fully fragmented cave material is achieved without drilling and
blasting after the initial undercut development. The fragmentation of the rock
mass is controlled by natural processes that include the in situ fracturing of the
1 – Introduction
rock mass, stress redistribution, the limiting strength of the rock mass and
progression of the material downward through the muckpile resulting in
autogenous grinding. In block caving, the orebody is fully undercut prior to
production commencing. In panel caving, production commences prior to the
orebody being fully undercut and mining progresses laterally across the orebody.
Sub-level caving typically operates on numerous mining horizons simultaneously

but progresses downwards through the orebody. Sub-level caving requires the
transition of in situ ore into a mobilised state by conventional drilling and blasting.
The material above the orebody is allowed to cave into the void created by mining.
The requirement for drilling and blasting is generally brought about by the high
strength of the rock mass, and/or the small orebody footprint.
1.2 Development of the Cave Mining Method
The cave mining method was developed in the underground coal mines of England
in the late 17th century. At this time it was referred to as the Shropshire Method,
and was developed to induce gravity caving (through undercutting) in highly
fractured, persistent and flat-lying coal seams. The area from which the coal was
extracted was generally back-filled with stone to limit surface subsidence (Energy
Information Association, 1995). As the demand for coal increased during the 18th
and 19th century, the advantages of the cave mining method were noted “… it
enabled a colliery to be opened with less capital expenditure … the yield per acres
(was) greater… (and) shot firing (could) almost be entirely dispensed with (because
the) weight on the face is, in itself, sufficient to bring down the coal…” (Hughes,
1917).
By the late 19th century – early 20th century, coal caving had been adapted to the
metalliferous mines in Michigan, USA whereby the high in situ density of the ore
was exploited to induce a gravity caving mechanism in mining blocks
approximately 60-75 m long, 30-40 m high and 60 m wide, and, after visiting these
caving operations during the 1950s, De Beers commissioned their first diamond
caving operation (Brown, 2003) in South Africa. The ubiquitous application of
cave mining in South African diamond mines since this time has led to the
development of current day empirical cave assessment techniques (Diering and
1 – Introduction
Laubscher 1987, Laubscher 1990, 1994, 2000) that are further discussed in
Section 2.1.2. In the case of South African diamond mining, effective caving is
enabled without dilution problems by the contrast in strength between the weak
kimberlitic diamond pipe and the comparatively strong host rock mass.
As a result of the continued success of the cave mining method in coal, iron ore and
kimberlite operations and in a number of strong, (UCS, greater than 80 MPa and
GSI, greater than 50) jointed rock masses (e.g. Urad Mine, Colorado, USA; 1914-
1960; Philex Padcal Mine, Philippines, 1959-current and El Teniente, Chile,
1920’s-current), during the mid-1990’s a caving mine was planned at Rio Tinto’s
Northparkes E26 Lift 1 Orebody. Feasibility studies for the mine were carried out
using Laubscher’s empirically derived caveability chart and the orebody was
predicted to fall well within the Caving Zone (Ross and van As, 2005). Production
from the block cave commenced in 1996 and cave initiation followed. However,
once the undercut development was completed, the cave stalled at a height of 95
m. Figure 3 provides a chart showing the initial estimate of caveability at the E26
Mine (a) and the stalled geometry (b).
Figure 3. (a) Laubscher’s stability chart showing E26 predicted caveability (b) section
showing cave stall at Northparkes E26 prior to plug failure (modified after
Ross and van As, 2005).
1 – Introduction
Increased production rates failed to induce further caving of the orebody and an
air-gap developed. In order to stimulate cave growth, a hydraulic fracturing
campaign was conducted. Boundary weakening (blasting a sub-vertical slot on the
south-western boundary of the cave) methods were also employed. After a
significant amount of effort over a two year period, caving once again
recommenced. However, on Wednesday, 24th November, 1999, the cave back
advanced rapidly through to the ground surface and generated a wind-blast
through the underground workings. Four workers were killed. It is estimated that
a total amount of 13 MTonnes of material came down in this collapse - a column
height of 200 m. At the time of the accident, the air-gap was in excess of 180 m
(Ross and van As, 2005).
As a result of the unexpected stall and then rapid failure and air-blast at the
Northparkes E26 Lift 1 Mine, the International Caving Study (ICS 1997-2004) and
Mass Mining Technology (MMT I 2005-2008 and MMT II 2009-2012) projects
were initiated. These industry funded research projects have made significant
contributions to the advance of rock mechanics understanding associated with
cave mining methods in hard, jointed rock masses. The following section provides
a review of the current state-of-the-art engineering for cave propagation and
subsidence assessment in hard, jointed rock masses.
1.3 Caving Mechanics
It is commonly understood that all rock masses must cave if they are undercut over
a significant enough area. Caving can occur as a result of two influences – gravity
and stress. The mechanism of caving will depend on the relationship between
induced stresses, geometry of the cave footprint, strength of the rock mass and
joint fabric (Brown, 2003).
Stress caving occurs when the induced stresses in the cave back exceed the
strength of the rock mass causing yielding and fragmentation of the rock mass into
a caved rock state. Gravity caving is characterised by low mining induced stresses
and is often analysed by knowledge of the joint fabric and simple kinematics.
1 – Introduction
Gravity induced unravelling can be expected to occur in the cave back (roof) as a
tensile failure mechanism under low stress conditions. Failure can occur through
slip along pre-existing joints as the rock is unconfined from below, or through
bending/deflection of the rock layers (voussoir beam theory). Gravity caving
usually results in coarser drawpoint fragmentation since little damage is induced
to the rock mass during its mobilisation. Primary fragmentation in this case, is
usually close to the in situ block size. Example stress-paths for both stress and
gravity caving mechanisms are presented in Figure 4. The disintegration and
mobilisation of a rock mass resulting from yield in the compressive regime is
called stress caving. In the tensile regime, it is gravity caving.
Figure 4. Typical caving stress-paths representing stress and gravity caving

mechanisms.
Self-sustained propagation of the cave stalls when a stable arch develops in the
advancing back. In this case, the induced stresses do not exceed the strength of the
intact rock bridges and/or is unable to induce failure along pre-existing joints.
Time-dependent processes (i.e., stress corrosion, ground water etc.) may
1 – Introduction
eventually mobilise this material, but in most cases, artificial cave stimulation is
usually required. Cave propagation behaviour has been described by Brown
(2003) and Pierce (2010). A summary is provided in Figure 5.
Figure 5. Conceptual stress-state development around a propagating cave (modified

after Pierce, 2010).
1 – Introduction
As a result of the two caving processes identified in Figure 5 (stress and gravity),
Pierce, (2010) has defined seven critical factors that impact the cave propagation
behaviour of a rock mass. They are summarised below.
(a) Cohesion and Tension Weakening. During stress caving, the rock mass
undergoes a reduction in strength from its peak in situ value to a much
lower residual value (representative of a caved rock mass state). This
overall response is often termed a “strain-softening” process, and is the
result of strain-dependent material properties. Gravity caving in low
induced stress environments is driven by the ability of the in situ joint fabric
to loosen and mobilise (its tensile strength).
(b) Post Peak Brittleness. The rate at which the rock mass strength drops
from peak to residual is referred to as brittleness. Rocks that maintain their
peak strength with continued loading are referred to as perfectly plastic
(ductile) constitutive models. Rock masses that instantaneously drop to
low residual strength properties when they exceed their peak strength are
referred to as perfectly brittle. In general, brittle rock masses cave more
readily than ductile rock masses.
(c) Deformation Modulus Softening. During caving, the rock mass increases
in volume as intact rock blocks fracture, separate and rotate during the
yielding and mobilisation process. Along with this bulking, a reduction in
the deformation modulus is expected to occur. Representation of the
decrease in deformation modulus is crucial for assessing the evolving stress
state around the cave, since, as the rock mass dilates/bulks its potential to
carry stress decreases.
(d) Dilational Behaviour. Rock mass dilation is the change in volume of a rock
that occurs with shear distortion. An accurate assessment and
representation of the dilation behaviour of a jointed rock mass during cave
initiation and propagation is essential in being able to accurately predict the
correct bulking behaviour and volume increase for air gap assessment.
1 – Introduction
(e) Rock Mass Discontinuities (faults, joints). The influence of geologic

structure on subsidence and caving is recognised to be important. The
presence of a steeply dipping fault can terminate the angle of draw short of
its normal value. If a shallow dipping fault intersects the collapsing rock
column the lateral extent of surface subsidence can increase outward to the
place where the fault intersects the ground surface. Using experimental
data, Brown and Trollope (1970) demonstrate that even if the rock mass
structure and geomechanical properties are well conditioned for slip, the
existence of high horizontal in situ stresses can inhibit rock mass failure and
therefore prevent cave initiation and propagation.
(f) Production Draw Schedule. The rate of draw and shape of the undercut
footprint has previously been identified by Laubscher (1990, 1994) to
impact the cave propagation behaviour of a rock mass. In addition, the cave
propagation behaviour may be affected by the evolving geometry of the
undercut footprint relative to the principal stress direction. If the caved
mass is withdrawn more heavily in the centre of the undercut the back may
be more arched. Induced stresses will be higher in this case, but they will
be confined to a smaller portion of the cave back.
(g) Other Excavations. Induced stresses may be impacted strongly by the

presence of other excavations beside or above the cave being considered.
Adjacent caves may stress shadow (inhibit the build-up of mining induced
stresses around) an orebody, while overlying caves or open pits may
increase the induced stresses, promoting cave growth.
Due to the low operating cost nature of the cave mining method, the demand for
caving operations is increasing. This demand is driven by the potential for
increasing operational profit and possibility of extracting lower grade orebodies.
However, due to the significant capital investment required prior to production,
the sole reliance on empirical methods is no longer sufficient to justify the
expenditure. In addition, as a result of more complex geomechanical conditions at
proposed locations, empirical methods are unable to answer many questions
regarding expected caving behaviour since they are being applied beyond the
limits of their data set. As a result of this, a robust methodology based on
1 – Introduction
fundamental geomechanical principles is required to predict cave propagation

behaviour and surface subsidence prior to the investment of significant capital and
development of infrastructure.
1.4 Terminology
A conceptual model of a self-sustained propagating cave has previously been

developed by Duplancic and Brady (1999) through seismic investigations and
underground instrumentation at the Northparkes Mine, Australia. The conceptual
model includes four main behavioural regions that are shown in Figure 6.
Figure 6. Conceptual diagram of showing the main behavioural regions of a

propagating cave based on underground observations and instrumentation.
The characteristics of each region are defined below:
 Elastic Region - The host rock mass around the caving region behaves
mainly elastically and has properties consistent with an “undisturbed” rock
mass.
1 – Introduction
 Seismogenic Zone - Microseismic (and sometimes seismic) activity is

concentrated in this region primarily due to slip along pre-existing
discontinuities and the initiation of new fractures.
The overall advance rate, thickness and spatial distribution of the

seismogenic zone has previously been determined by the criterion in
Equation [1] (after Diederichs, 1999). This criterion is based on back-
analysis of seismic response in a number of deep Canadian operations.
[1]
Where: and are the Major and Minor Principal Stress Magnitudes
(MPa), is the Damage Threshold Value (%) and is the Uniaxial
Compressive Strength (MPa)
The damage threshold ( ) has been shown to correspond to the point

at which measurable seismicity might be picked up in the mine. The simple
criterion has been successfully applied in back-analyses of the recorded
seismogenic zones at Northparkes E26 Lift 2 (Pierce et al., 2006) using a
value of 35%. Others (Beck et al., 2006) have correlated seismic
potential to Dissipated Plastic Energy (DPE - the energy in joules dissipated
as a result of yield in a mining step). The greater the DPE, the greater the
seismic potential. However, no clear guidelines have been reported to
interpret this value from a numerical model.
 Yielded Zone – The rock mass in this region is fractured and has lost some
or all of its cohesive strength and provides minimal support to the overlying
rock mass. A rock mass within the yielded zone is subject to significant
damage, i.e. open holes are cut-off, Time Domain Reflectometry (TDR)
breakages are expected and cracking is observable in infrastructure. Stress
components within this region are typically low in magnitude. The
reduction in rock block size (compared to the in situ state) due to yielding in
this region can be described as primary fragmentation. Primary
fragmentation from a stress caving mechanism is generally finer than that
from gravity caving (Laubscher, 1994).
1 – Introduction
 Mobilised Zone – This zone gives an estimate of portion of the orebody

that has moved in response to the production draw and may be
recoverable. Although the specific location of the cave back is difficult to
predict precisely, it is estimated to be rock that has experienced a
displacement greater than or equal to 1-2 m (after Pierce et al., 2006). The
reduction in rock block size (compared to the primary fragmentation state)
is described as secondary fragmentation. Secondary fragmentation is
affected by draw height and internal caving stresses (Laubscher, 1994).
Cave propagation behaviour and subsidence are closely linked geomechanical

processes. The surface projection of a cave after break-through can be described
by similar terminology as the underground regions. A conceptual schematic model
of the surface subsidence domains compared to the underground cave domains are
presented in Figure 7.
Figure 7. Conceptual schematic diagram showing the main behavioural regions of a

cave that has propagated through to the ground surface.
The characteristics of each region are described below.
 A crater is a common surface feature of many caving mines; it also is

referred to as the zone of active movement (van As et al., 2003). The crater
consists of irregular blocks of rock, ranging in size from millimetres to
1 – Introduction
several metres in magnitude. It develops as the mobilised zone influence

reaches the ground surface, causing the overlying rock mass and/or side
rock to fall into the mined void. Over time, the surface of the crater may
subside as ore is withdrawn (Lupo, 1998).
 Zone of Large-Scale Fracturing. The ground surface within this zone is

broken and has large open tension cracks, benches, and rotational blocks.
Usually, the primary failure mechanism of surface cracks associated with
cave mines is shear failure of the abutment rock mass, which results in the
development of stepped benches and scarps. Other types of failure
mechanisms, such as toppling and block rotation, also are present, but they
appear to be secondary mechanisms that form after the primary shear
failure develops. A total strain criterion of 0.5% has previously been used
by Sainsbury and Lorig (2005) to calibrate the limits of large-scale
fracturing at the abandoned Grace Mine in Pennsylvania, USA. This total
strain criterion has also been used to calibrate the limit of large-scale
fracturing at the El Teniente block cave mine in Chile (Cavieres et al., 2003).
 Continuous Subsidence, as defined by Brauner (1973), is the response of

the rock mass to a mined void, which results in the formation of a gentle
surface depression. Generally, the continuous subsidence zone forms
between the large-scale surface cracking zone and the undisturbed surface
(elastic zone). Surface buildings, roads, underground power lines, railroads
and other structures can be impacted significantly by continuous surface
subsidence. Lupo (1998) measured continuus subsidence up to 200 mm at a
distance of 250 m from the zone of large-scale fracturing that caused heavy
damage to nearby surface structures. The strain criteria developed by
Singh (2003) has previously been used to define the extent of this region;
horizontal strain > 0.002 (0.2%) and angular distortion > 0.003 (0.3%) at
the abandoned Grace Mine (Sainsbury and Lorig, 2005).
 Elastic Zone. The area beyond the continuous subsidence zone.
1 – Introduction
1.5 Thesis Outline
The research documented by this Thesis has been broken into chapters based on
unique ideas and concepts. A summary of this break-down is provided below.
Chapter 1 – summarises the problem being studied, outlines the motivations for
the research and provides some background information on the cave mining
method.
Chapter 2 – critically reviews the existing techniques being used by the mining
industry for cave propagation and subsidence analyses in hard, jointed rock
masses.
Chapter 3 – outlines the objectives and methodology for the development of a

cave propagation and subsidence assessment techniques in hard, jointed rock
masses.
Chapter 4 – outlines the development of a numerical cave propagation

demonstration model. The model has been used to investigate the effect in situ
geomechanical conditions have on cave propagation behaviour. The model has
also been used to demonstrate the implementation and impact of the research
detailed herein.
Chapter 5 – outlines the development of a new technique that allows accurate

representation of anisotropic strength and deformation modulus (as defined by
SRM testing) responses in large-scale, three-dimensional cave propagation and
subsidence numerical models.
Chapter 6 – outlines the development of new numerical algorithms that allow the
rigorous consideration of the effect of volumetric changes of a rock mass during
cave propagation that include density, deformation modulus and dilation.
Chapter 7 – examines the existing approaches for simulating large-scale

discontinuities in a numerical model and provides a detailed methodology for
incorporating large-scale structure in a cave propagation and subsidence analysis.
1 – Introduction
Chapter 8 – examines the deficiencies of the existing numerical production draw

models and develops new algorithms to simulate production draw through mass
balance calculations.
Chapter 9 – outlines the development of a new numerical algorithm to simulate

crater development in cave propagation and subsidence analyses.
Chapter 10 – outlines the development of a sub-level caving algorithm for the

simulation of production draw.
Chapter 11 – describes the implementation and validation of the numerical cave

propagation and subsidence model on a back-analysis of caving induced pit failure
at the Palabora Mine in South Africa.

propagation and subsidence model on back-analysis of a structurally controlled
cave initiation and propagation at the Henderson Mine, USA.

propagation and subsidence model on a back-analysis of caving induced
subsidence at the abandoned Grace Mine in Pennsylvania, USA.

propagation and subsidence model on a back-analysis of caving induced
subsidence at the Kiirunavaara Lake Orebody sub-level cave.
Chapter 15 – provides a summary of the conclusions and recommendation for

future work arising from this research.
1 – Introduction
2 DEVELOPMENT OF CAVE PROPAGATION AND ROCK

MASS MODELLING TECHNIQUES
2.1 Cave Propagation Assessment
2.1.1 Analytical
Since the application of cave mining methods during the early part of the 20th
century, researchers have sought to understand and predict the nature of cave
propagation. Rice (1934) and later Panek (1984) developed simple one-
dimensional volume based relationships to characterise cave propagation
behaviour according to assumed bulking factors, as illustrated in Figure 8.
Figure 8. Analytical solution: relationship between thickness of caved ore withdrawn

( t c ), thickness of in-place ore extracted ( t i ) and rise of the caving brow
( h ) (after Panek, 1984).
Simple volumetric relations are still used today by many researchers to estimate
cave propagation rates (e.g. Beck et al., 2011) and assume that:
(a) cave initiation always occurs,
(b) cave propagation (and mobilisation of material) is always vertical; and that
(c) the caving rate is constant (based on the use of a constant bulking factor).
2 – Development of Cave Propagation and Rock Mass Modelling Techniques

Through a number of recent published case studies, these assumptions have been
shown to be incorrect. At the Ridgeway Deeps Mine in NSW, Australia, Beck et al.
(2011) documented a scenario in which cave initiation did not occur in a portion of
the undercut footprint. Carlson and Golden Jr. (2008) describe a situation at the
Henderson Mine in which migration of the propagating cave beyond the undercut
footprint was observed along a weak intrusive contact. And, at Northparkes Lift 1
Mine, Ross and van As (2005) have described in detail the highly variable caving
rate during cave initiation, propagation, stalling and rapid plug failure at this
operation.
The documentation of cave propagation rates at numerous operating mines

around the world also shows significant variability. The propagation rate has been
assessed by the ratio of the yield zone height to the average production draw
height measured in solid rock drawn. A summary is provided in Table 1.
Table 1. Documented yield zone propagation rates from caving operations around the
world (after Sainsbury and Sainsbury, 2010).
Yield
Propagation
Operation Method Rate Reference
El Teniente, Chile Panel 5: 1 Villegas (2008)
Henderson Mine, Colorado, USA Panel 7: 1 Board et al. (2009)
Grace Mine, Pennsylvania, USA Panel 8.2: 1 Sainsbury et al. (2005)
Australian Coal Mine Longwall 8.9: 1 Hebblewhite (1995)
DOZ Mine, Indonesia Block 6-10: 1 Szwedzicki et al. (2006)
Kimberley Mines, South Africa Block 6 - 12: 1 Guest (2009)
Lakeshore Mine, Arizona, USA Block 10: 1 Panek (1984)
Questa Mine, New Mexico, USA Block 10: 1 Gilbride et al. (2005)
San Manuel Mine, Arizona, USA Panel 10: 1 Gilbride et al. (2005)
Athens Mine, Michigan, USA Block 14: 1 Boyum (1961)
Palabora Mine, South Africa Block 15: 1 Sainsbury et al. (2008)
Northparkes Lift 2, Australia Block 20: 1 Pierce et al. (2006)
Chinese Coal Mine Longwall 31.3: 1 Liu (1981)
Caving rates in the order of 5:1 up to 31:1 have previously been documented. This
variability can be attributed to variations in the in situ geomechanical conditions
and cave mining method. Based on the values presented in Table 1 it can be seen
that a longwall mining method provides the greatest documented propagation
rate, followed by block caving and then panel caving methods. The ability to
predict and represent this variation in propagation rates between caving methods

and in situ geomechanical conditions is paramount in developing a robust

methodology for forward predictions.
Since there is often very little detailed consideration of the geomechanical

conditions in analytical caving solutions, previous experience of the in situ rock
mass bulking behaviour is required to estimate the cave propagation rate. Based
on the results compiled in Table 1, these rates are highly variable and the selection
of an appropriate value can be problematic in green-field operations, or when
changing conditions are encountered. As a result of these restrictions, empirical
methods are seen as an advancement since they have been developed based on a
range of case history experience.
2.1.2 Empirical
The use of empirical methods to estimate cave propagation behaviour have been,
and still are, widespread. The most commonly used approach for estimating
caveability was developed by Laubscher (Diering and Laubscher 1987, Laubscher
1990, 1994, 2000) and is based on a compilation of caving case histories - largely
derived from low strength, kimberlitic deposits in South Africa. Laubscher’s chart
(Figure 9) defines, three possible caving states that include; “no caving”,
“transitional”, whereby the cave initiates, but propagation is minimal, and “caving”
whereby self-sustained propagation occurs. The states are defined based on a
measure of the rock mass strength, through the Mining Rock Mass Rating (MRMR)
and the hydraulic radius of the undercut footprint. The MRMR was originally
developed by Laubscher (1975) and includes the consideration of intact rock
strength, discontinuity frequency and weathering, joint orientation, stress and
blasting.

Figure 9. Empirical method for predicting caveability: Laubscher’s stability chart

(after Laubscher, 1994).
Many mines still use Laubscher’s caving chart to estimate the undercut dimensions
required to induce self-sustained propagation, and, in most cases, good agreement
is achieved. However Lorig et al. (1995), van As and Jeffrey (2000) and De Nicola
Escobar and Fishwick Tapia (2000) have previously reported instances where
significant differences were observed between the actual and predicted cave
propagation behaviour. A detailed review of these cases by Trueman and
Mawdesley (2003) showed that the biggest variance in actual versus predicted
outcome was associated with strong (MRMR greater than 50) rock masses and
misinterpretation of the application of adjustments in the MRMR rating scheme.
As a result of this review, Trueman and Mawdesley (2003) proposed an alternate
method for the prediction of self-sustained propagation through an extension of
the Mathews stope stability chart (Mathews et al., 1981). Their extended stability
chart is provided in Figure 10.

Figure 10. Empirical method for predicting caveability: Extended Mathews Stability
Chart (after Trueman and Mawdesley, 2003).
Although the development of this method extended the application of the existing
empirical approach to stronger rock masses, by necessity, the method is still
limited by the dataset that it was developed from. Additional limitations of
Laubscher’s and Trueman and Mawdesley’s empirical approaches have previously
been documented by Brown (2003) and suggest that the approaches are only
satisfactory for footprint length to width ratios of three or less. Beyond this, the
technique is unable to account for variations in three-dimensional stress
redistribution around rectangular undercut footprints. In addition the influence of
only one joint set orientation can be analysed. Experience suggests that the critical
joint set orientation may vary around the undercut footprint as the principal stress
direction changes during undercutting and cave propagation. Milne et al. (1998)
also suggest that the determination of adjustment factors can be ambiguous and
subject to personal experience. This means, for the same rock mass data set,
different caving behaviour may be interpreted.

In addition to their limited application, the use of empirical methods do not

provide any indication regarding the rate of cave propagation, nor the extent of the
cave behavioural regions. It can be assumed that the further into the caving zone
your scenario falls, the more rapid the cave propagation. However, the actual
timing or magnitude and the impact to underground and surface infrastructure
cannot be predicted without assumptions regarding the bulking behaviour of the
rock mass. In addition, limitations associated with compiling enough case study
data limit the application of empirical techniques to situations in which large-scale
discontinuities, rock mass strength anisotropy, scale effects, excavations and/or
significant topological relief and heterogeneous rock mass domains do not exist.
2.1.3 Numerical
There are numerous numerical modelling methods (Boundary Element, Finite

Element, Finite Difference, Distinct Element and Hybrid) and approaches available
for performing stress and deformation analysis in geomechanics. The important
aspect of modelling is not necessarily the numerical program itself, but the
methodology for simulating the caving process, and the estimation of input
properties.
In his review of cave mining practices, Brown (2003) reasons that numerical
modelling enables a more fundamental and rigorous assessment of cave initiation
and propagation behaviour than empirical (or analytical) methods, since it may
have advantages in cases for which current experience is lacking.
The following section attempts to review the development of numerical caving

methodologies from 1970 (the time of the first known numerical caving model) to
the state-of-the-art routines that are currently being applied in geotechnical pre-
feasibility and feasibility studies today. Although other emerging numerical
techniques have been applied to the study of cave propagation such as Smooth
Particle Hydrodynamics (Karakel et al., 2011), they are not included in this
literature review since, to date, their application has not been compared to actual
case study data. For this reason, they are considered demonstration tools only.

2.1.3.1 Two-Dimensional Elastic Models
Soon after the introduction of the Finite Element Method (FEM) for the numerical
analysis of stresses and displacements in continuous structures by Clough (1960),
Palma and Agarwal (1973) developed the first known two-dimensional, elastic,
finite element model to study cave propagation behaviour at the El Teniente Mine
in Chile. During this study, they identified the need to consider the nature of the in
situ rock mass fracture network and the impact of principal stress direction in
relation to the undercut dimensions on the cave propagation behaviour.
The high level of fracturing in the El Teniente rock mass was represented by
assigning zero tensile strength to all zones within the model. Although not many
details of the modelling methodology are provided, it is clear that yielding of the
rock mass immediately above the simulated undercut was assumed to propagate
when a tensile stress component was identified within a zone. Figure 11 presents
the results that clearly show the simulated impact of cave height based on in situ
stress and orientation of the undercut footprint.
Figure 11. Impact of principal stress orientation in relation to an undercut as defined

by two-dimensional numerical modelling (after Palma and Agarwal 1973).

Although these simulations assumed that caving only occurred as a result of a

tensile failure mechanism, they were fundamental in understanding that strength
degradation and jointing played an important role in cave propagation behaviour.
In addition, they highlighted the importance of the orientation of a rectangular
mining block with respect to the in situ principal stress direction. It is now
understood that inducing the redistribution of the maximum principal stress over
the shorter footprint axis will promote cave propagation since the cave back
experiences greater stress concentrations when maximum principal stress hits
cave "broadside" as opposed to "end-on". In the case of “end-on”, the cave
presents a larger obstacle to stress. The concept is presented in Figure 12.
Figure 12. Conceptual diagram of effect of principal stress direction and undercut
footprint dimensions on caveability.
The results of Palma and Agarwal’s work are the first known application of
computer-based numerical modelling for cave propagation analysis and
successfully provided a means for a more rigorous analysis of the tensile failure
mechanism that develops in advance of a propagating cave. In addition, they
highlighted the effect of stress field variations around the extraction level
geometry on cave propagation height. However, their consideration of a simple
elastic material model was unable to account for a stress-caving mechanism since
no failure criteria is specified. This meant that the rock mass in the cave back

could deform and become stressed infinitely without failing. This basic assumption
may be appropriate for some mining methods in very strong, massive rocks but
can lead to highly misleading results in weaker rock masses when there is the
potential for shear failure of the rock mass and redistribution of stresses.
2.1.3.2 Two-Dimensional Plasticity Models
Through the application of two-dimensional FEM simulations at the Grace Mine,

located in Pennsylvania, USA, Barla et al. (1980) introduced a softening material
model to represent the degradation of the in situ rock mass strength to a fully
weakened and bulked state during cave propagation. The use of such a material
model highlighted the limitations of elastic modelling completed by Palma and
Agarwal (1973) and the development in understanding that caving may not only
occur due to a gravity mechanism, but also a stress mechanism – as discussed in
Section 1.3.
The softening behaviour in the model was simulated through a periodic review of
the failure states in the numerical mesh. If a zone failed via a compressional or
tensile mechanism, then the strength, density and stiffness were reduced to a
residual value. Production draw was simulated through a force application in the
undercut roof. Figure 13 provides a schematic representation of their modelling
methodology.
The simulations conducted by Barla et al. (1980) do not only account for a
softening material model, but also represent the changes in deformation modulus
and density during cave propagation. In addition, they identified the importance
in being able to accurately represent the mining process within the numerical
model in order to predict the most realistic rock mass response. However no
correlation was made regarding the amount of material withdrawn.

Figure 13. Development of two-dimensional numerical modelling approaches for cave

propagation analysis. (a) Model mesh (b) section through the mining
geometry (c) simulated undercutting process (d) contours of resultant
mobilised strength – the shaded area represents a fully softened/caved rock
mass (after Barla et al., 1980).

During the early 1990’s, Rech and Lorig (1992) conducted two-dimensional, finite
difference analyses in order to reproduce the existing cave conditions at the
Henderson Mine in Colorado, USA and predict the expected cave propagation
behaviour. These are the first simulations that attempted to correlate the
production schedule with the simulated cave propagation behaviour.
The cave zone was initialised within the model through a number of incremental
undercut expansions that corresponded to the historical and planned production
schedule. Vertical draw and a bulking factor were assumed based on the
volumetric equations outlined by Panek (1984). Stresses were reset to zero within
the cave mass and the rock mass properties were reduced to those consistent with
a fully-bulked rock mass. Simulation of residual rock mass properties and reduced
vertical stress conditions within the caved mass ensured that the mining induced
stress magnitude and directions were accurately captured. However, by manually
initialising the caved rock mass within the model, a true and spontaneous cave
initiation could not be predicted. In addition, by artificially reducing stresses
within the cave, mass and energy were not conserved within the system. As a
result of this, the representation of the exact production tonnage simulated within
the model could not be gauged and the real caving induced stress-damage may
have been under-estimated.
The algorithm used by Rech and Lorig to model the undercutting and mining
advance sequence is provided in Figure 14.

Figure 14. Methodology for the application of a continuum based numerical model for
the prediction of onset of caving (after Rech and Lorig, 1992).
During the International Caving Study (ICS 1997-2004), Karzulovic and Flores
(2003) considered the influence of depth, stress, large-scale discontinuities, rock
mass strength and groundwater on caveability through a generalised sensitivity
analysis with a two-dimensional FEM code. The caving methodology employed
assumed vertical cave propagation, similar to an analytical approach. In order to
estimate the potential for cave propagation, it was assumed that cave growth
would equal 10% of the undercut length (i.e. if the undercut length is 100 m, the
vertical propagation of the cave would be 10 m), as illustrated in Figure 15a.

Figure 15. Determination of the Cave Propagation Factor at Northparkes E26 Lift
1 (after Karzulovic and Flores 2003). Here a 480 m block height, and a
rock mass of fair to good geotechnical quality, (HC = 480 m, B = 180 m,
K = 1.50) has been assessed. The chart indicates that the caving will not
propagate through the 480 m block.
Based upon stress redistribution around an imposed cave shape, a Cave

Propagation Factor (CPF) was used to determine if caving is ‘Problematic’,
‘Transitional’ or ‘Self-Sustained’ - much like Laubscher’s ‘Caving’, ‘Transitional’ and
‘Stable’ Zones. The CPF has been defined by Karzulovic and Flores as the ratio
between the average deviatoric stress acting on the cave back and the maximum
deviatoric stress that the rock mass can sustain. The equations used to determine
the value are presented in Figure 15.
Although simple assumptions were used in the representation of geometry, stress

redistribution (two-dimensional), rock mass plasticity and post-peak rock mass
behaviour, assessment of the CPF at the Northparkes Lift 1 Mine provided good
correlation with the actual performance of the cave that stalled in 1999 – as shown
in Figure 15b.

However, the methodology is unable to account for the time-dependent nature of

the stable arch that developed at Northparkes Lift 1 and the subsequent plug-
failure. Limitations associated with the two-dimensional nature of the modelling
and the assumptions regarding cave shape and the propagation window (defined
by W=0.1B in Figure 15a) also make it difficult for the technique to accurately
predict three-dimensional, self-sustained cave propagation, and the cave
behavioural regions presented in Figure 6. In addition, only vertical cave
propagation and homogeneous rock mass properties can be considered which
limits its applicability.
2.1.3.3 Axis-Symmetric Strain-softening Models
In an attempt to include a better representation of the three-dimensional shape of

the propagating cave and surrounding induced stress field, Lorig (2000) (also
reported in Brown 2003) conducted sensitivity simulations in axis-symmetric
models. A cylindrical undercut located at increasing depths was considered. The
initial state of stress within the model was assumed to be lithostatic and stress
boundaries (a support pressure) were imposed at the excavated undercut level to
ensure initial stability. To simulate production draw, the support pressure was
monotonically reduced in the roof of the undercut (similar to the approach of Barla
and Boshcov, 1980) and the extension of the yielded rock mass (represented by a
strain-softening material model) was assessed.
A schematic representation of the modelling methodology used to simulate

production draw from these axis-symmetric models is provided in Figure 16.

Figure 16. Development of axis-symmetric numerical methods for cave propagation (a)
axis-symmetric concept (b) evolution of the undercut pressure and height (c)
stepwise reduction of undercut pressure (d) details of the pressure evolution
with a simulated reduction step (after Brown, 2003).
Through this approach, even though the true three-dimensional geometry and
stress tensor were not accurately represented, Lorig (2000) was able to predict a
hydraulic radius associated with cave initiation that compared well to Laubscher’s
empirical cave stability chart over a range of MRMR values. This technique is
considered to be an advancement on the CPF method proposed by Karzulovic and
Flores (2003) since failure was not limited to a window of rock mass in the cave

back (dictated by the undercut width), but allowed to evolve based on the rock
mass properties and stress state in the model.
Using this axis-symmetric approach, Lorig completed an analysis of the

Northparkes Lift 1 cave using the same parameters as Karzulovic and Flores
(2003). The displacement vector (a) and cohesion (b) results are illustrated in
Figure 17.
Figure 17. Development of cave geometry resulting from a two dimensional strain-
softening numerical simulation, hydraulic radius 42.5m (a) displacement
vectors (b) cohesion softening ; green represents no softening, red represents
fully softened rock mass(after Lorig, 2000).
For an incrementally expanding undercut size, the resulting cave yield height was
assessed in the numerical model. A hydraulic radius of 42.5 m was required to
reproduce the observed stalled yield zone height at Northparkes of 95 m. This is
consistent with the documented stalled undercut geometry by Ross and van As
(2005). However, although the cave yield height was reproduced, it is clear that the
shape of the yielded rock mass volume is not necessarily realistic when the results
are compared to the conceptual model of a cave. It is simulated with a flat back.
The modelling results of Lorig showed that a strain-softening model could be used
with confidence to predict rock mass damage and cave propagation. However, a
model that was able to simulate the failure mechanisms in the back of the cave in
more detail was required to ensure the flat back shape was addressed.

2.1.3.4 Two-Dimensional Distinct Element Models
Two dimensional, Distinct Element Models (DEM) were developed by Lorig et al.
(1995) within the PFC2D code (Itasca, 1995) to provide a greater understanding of
the fracturing of the in situ rock mass and an improved cave back shape based on
the models shown in Figure 17. Conceptual models of cave propagation behaviour
in a high initial stress state were developed and two fundamental failure
mechanisms associated with cave propagation were identified that included; intact
rock block failure and slip along pre-existing joints. The model results are
presented in Figure 18a and Figure 18b. Brown (2003) reports on the extension of
the two-dimensional DEM simulations to three-dimensional axis-symmetric
models which are presented in Figure 18c.
Figure 18. Development of a discrete element model to study cave propagation (a)
particle clusters early in the caving process with superimposed contact force
chains (after Lorig et al., 1995). (b) particle clusters after significant cave
propagation showing internal fractures of blocks in the caving zone chains
(after Lorig et al., 1995). (c) forces arching around the unstable rock mass
(after Brown, 2003).
Within these models, each intact rock block is represented by particles that are
glued together. Upon initiation of the undercut, the bonds between the particles
are broken if the induced force is greater than the bond strength (stress caving).

Unstable particles that have broken their bonds or are unconfined, are free to
dislocate and fall moving into the space created by the undercut. Figure 18C
shows the force distribution in the particles around the cave periphery. The
increased caving induced stresses are seen as arching of these forces around the
cave limits.
These models illustrate the capacity of DEM codes to simulate the two primary
caving mechanisms in a jointed rock mass (stress and gravity). However, at the
present time, the size of these models is limited due to the computational intensity
of the modelling technique.
2.1.3.5 Three-Dimensional Distinct Element Models
In spite of the computational intensity of the modelling technique, Gilbride et al.

(2005) and Sharrock et al. (2011) have most recently used a three-dimensional
DEM approach to model the mechanisms of caving for large-scale subsidence
analyses. Gilbride has calibrated a large-scale rock mass response through rock
mass simulations in a laboratory environment and used the results in a large-scale
model. His results are presented in Figure 19.

Figure 19. Large-scale (mine-wide) discrete element modelling of caving and subsidence
phenomena in three-dimensions. Cross section of subsidence mass movement
from block caving and simulated synthetic rock mass triaxial test of PFC
material (after Gilbride et al., 2005).
Since there is no comparison between actual field data, or presentation of the data
used for the calibration, it is difficult to determine the accuracy of the model
results.
Sharrock et al. (2011) calibrated a response through large-scale observations of

surface subsidence, as presented in Figure 20.

Figure 20. Large-scale (mine-wide) discrete element modelling of caving and subsidence
phenomena in three-dimensions. Plan view cave zones: measured (blue) and
simulated (red) (after Sharrock et al., 2011).
In this case, there is significant difference between the model result and the
subsidence observations on the eastern limits of the crater.
It seems in each case, computational inefficiency of the DEM technique, has limited
the particle size within the models to 13-24 m, and thus restricted the size of the
physical phenomena that can be resolved. As a result of this, the small-scale
cracking / dislocation of the rock mass achieved by Lorig et al. (1995) was unable
to be reproduced with such a large-scale model due to the minimum particle size
required to achieve this scale of model.
As shown by Gilbride et al. (2005) and Sharrock et al. (2011), at present, it is not
practical to simulate the complex large-scale mining/geological processes in DEM
codes due to the computational intensity of the numerical technique. It seems at
the current time, computational constraints continue to limit application of DEM
simulations to small-scale (e.g. <100 m) boundary value problems in densely

jointed rock. As a result of this, continuum methods continue to be used to ensure

that the regional mine scale changes can be captured in the one model.
2.1.3.6 Three-Dimensional Continuum Methods
Based on computational limitations of simulating the large-scale caving process in

DEM codes, Pierce & Lorig (1998) describe an improved methodology developed
in a three-dimensional continuum code compared to the axis-symmetric approach
reported by Brown (2003). In this model, sequential undercuts of constant width
were simulated to reproduce the increasing undercut hydraulic radius during cave
initiation. Production draw was simulated by monotonically reducing stresses at
the undercut level using the same methodology presented in Figure 16. For each
undercut increment, the model was stepped to equilibrium before subsequent
undercut expansions were simulated. In addition to the dynamic nature of the
undercut expansion, Pierce & Lorig implemented a user-defined function that
modified material properties and stresses based on plasticity state and strain
within the numerical model. Through this approach, the point at which self-
sustained propagation (the critical HR) could be determined based on the actual
three-dimensional stresses distributing around the undercut. A diagram that
represents the modelling methodology is provided in Figure 21a along with typical
model results (Figure 21b). It can be seen that the modelling methodology
generally still results in a flat cave back - Figure 21B-7.

Figure 21. Three-dimensional strain-softening, continuum models for cave propagation

(a) logic sequence to simulate caving (b) typical simulation results (after
Pierce and Lorig, 1998).
By reducing the stresses uniformly across the undercut, material was withdrawn
preferentially from the higher stress areas in the undercut. This methodology
highlighted the need for better control on production draw simulation.
As an advancement on this technique, Pierce et al. (2006) simulated production

draw based on the pseudo-static application of small downward-oriented
velocities in the back of the undercut. By doing so, the amount of material being
removed from an area could be controlled. Details of this methodology are
summarised in Figure 22.

Figure 22. Simulation of production draw from large-scale, three-dimensional strain-

softening continuum models based on velocities.
Through this velocity controlled production draw, Pierce et al., (2006) were able to
simulate the evolving cave behaviour at Northparkes Lift 2 cave that was
consistent with the conditions observed in situ. The numerical results presented in
Figure 23 have been validated against measurements of openhole blockages, TDR
breakages and the progression of the seismogenic zone. Material properties
within the model were generated using the Synthetic Rock Mass (SRM) modelling
technique, which is discussed in Section 2.2.3.

Figure 23. Large-scale back-analysis of cave propagation behaviour at the

Northparkes E26 Lift 2 Mine. Progression of predicted mobilised zone
limit (white iso-surface) and overlying yield zone limit (blue iso-surface)
versus TDR breakage locations (blue spheres) and open hole blockage
locations (red squares) (after Pierce et al., 2006).

The numerical model was able to accurately predict the rate and shape of the
mobilised and yielded rock mass zones through a large-scale application.
Similar to this approach, Beck et al. (2007) developed a methodology for the
assessment of cave propagation behaviour using the numerical modelling package
ABAQUS (Simulia, 2011). An example of some of the results from one of these
models is provided in Figure 24.
Figure 24. Example of a mine-wide, three-dimensional, multi-scale simulation (after

Beck et al., 2011).
An attempt was made to review the caving approach that has been developed in
Abaqus based on a number of publications (Beck et al., 2006; Beck et al., 2011).
However, at the present time there are insufficient details provided in these (and
associated) publications to allow for a detailed critical review. As a result of this, it
was not possible to comment on the appropriateness of the underlying
methodology, constitutive behaviours and associated assumptions for simulating
caving using Abaqus.

2.1.3.7 Hybrid Techniques
Cave propagation behaviour of a jointed rock mass is strongly governed by the

unique nature of joints fabric together with the intact strength of rock-bridges that
make up a rock mass (Lorig et al., 1995). Apart from the DEM’s, most of the
current methodologies discussed to date represent the rock mass as an isotropic
material.
Pine et al. (2007) and Vyazmensky et al. (2007) used the combined finite element-
discrete element ELFEN code (Rockfield, 2007) to insert physical fractures into a
continuum finite element mesh that is gradually degraded into discrete blocks
through systemic sampling of the tensile strength and principal stress tensors.
The approach includes an adapted Mohr-Coulomb failure criterion that has been
coupled with a smeared crack model to trace the crack propagation process and
crack interaction. Initially, material is embedded with an initial fragmentation
profile based on a site specific fracture network. Results of recent simulations
conducted with this approach are presented in Figure 25.
Figure 25. Simulation of cave development using a hybrid, two-dimensional approach

(after Rogers et al., 2010).
The modelling of a jointed rock mass using such an explicit technique requires the
precise specification of the joint locations and mechanical properties. It is
considered impractical to suggest that every joint can be defined and modelled in a
deterministic way for a large scale problem such as a caving. In addition, this
approach, although well able to represent brittle failure, still relies on a macro

failure criterion to determine the initiation of new cracks within the continuum
mesh.
In addition, the hybrid finite element / distinct element modelling methodology is

currently limited to two-dimensions. Numerical modelling conducted by Palma
and Argawal (1973) have already shown that cave propagation is a three-
dimensional problem. The two-dimensional plane strain nature of the ELFEN
model, and the over simplified production draw simulation routine (i.e., uniform
draw over the entire footprint) can over-estimate the influence of major structures
and is not well suited to studying the potential for cave propagation to stall, as it
does not allow for the three-dimensional concentration of stresses in the cave
back. Significant research has been conducted that demonstrates that a three-
dimensional analysis is required to accurately account for the influence of the
major principal stress orientation, undercut advance orientation, (Palma an
Agarwal, 1973) and three-dimensional structure (faults and joints) orientation and
persistence during cave propagation (Phillips and Hellewell, 1994).
Although many conceptual block cave models have been documented (Rogers et
al., 2010; Pine et al., 2007 and Vyazmensky et al., 2007) this approach has not
been validated against observed behaviour at an existing mine.
2.1.4 Summary
The understanding of failure and deformation of jointed rock masses surrounding

underground and surface excavations has been a problem for centuries. However,
it was only during the 1970’s and 80’s and the rapid development of computer
technology that enabled numerical methods in rock mechanics to explore these
issues.
As a result of the low cost nature of the cave mining method, very large caving
operations are currently being planned around the world and the empirical caving
design methods which have long served the industry are no longer adequate for
the assessment of complex rock failure and deformation processes expected at
these locations. In addition, these empirical techniques do not satisfy the
increasingly stringent risk-based criteria for approving the large capital
expenditure in caving mines prior to production.

It is clear that the last ten years have seen significant developments in numerical
modelling methodologies that facilitate the ability to simulate the rock mass
behaviour and failure mechanisms associated with large, cave-scale problems.
Through its consideration of fundamental rock mass behaviour and evolving
production draw schedule, the caving algorithm described by Pierce et al., (2006)
is considered to be the current “state-of-the-art in block and panel caving from a
geomechanics perspective (Brown, 2003)” and should be used as the basis for an
expanded numerical model moving forward.

2.2 Rock Mass Modelling Techniques
2.2.1 Introduction
The strength and deformation behaviour of a jointed rock mass is governed

strongly by (a) the intact strength of the rock and (b) the presence of
joints/discontinuities (Brown, 2003).
The strength of the intact rock bridges is generally assumed to be the same as the
intact strength of the rock determined by laboratory testing. However, it is well
established that uniaxial compressive strength of intact rock decreases as
specimen size increases. This kind of change in the mechanical properties of rock
with size is referred to as scale effects. Previous investigations (Hoek and Brown,
1980; Pratt et al., 1972) have shown that this scale-dependent decrease may vary
between 20-80% of the measured intact value in the laboratory. The large-scale in
situ and laboratory scale tests of Pratt et al. (1972) are provided in Figure 26.
Figure 26. Measured rock strength-scale effect including large size specimens of in situ
test (after Pratt et al., 1972).

The strength of rock joints is dependent on their properties, level of interaction

and loading conditions. Bandis et al. (1983) showed through laboratory
investigation of the deformation characteristics of rock joints under normal and
shear loading that, at both low and high stress levels, the deformation of pre-
existing joints dominates the behaviour of a rock mass. It has previously been
shown by Mas Ivars et al. (2011) that preferred joint orientations can induce a
marked anisotropy in deformation modulus, strength and brittleness of the rock
mass. In addition, joint density and persistence must be considered relative to
problem size – as shown in Figure 27.
Figure 27. Applicability of the Hoek-Brown empirical rock mass strength criterion at
different scales (after Li et al., 2008).
It is not possible to derive material properties of a rock mass based on laboratory

tests due to the size requirements of a sample scale large enough to achieve
repeatable strength results (i.e. Representative Elemental Volume, REV). In
addition, the field testing programs of a sufficient magnitude and nature are
expensive and relatively crude. Therefore, the current practice for the
characterisation of jointed rock masses is based on either empirical or numerical
methods.

The following section reviews the current techniques used to represent the
strength response of jointed rock masses that can be used for an assessment of
cave propagation and subsidence behaviour.
2.2.2 Hoek-Brown Strength Criterion
The Hoek-Brown failure criterion is widely accepted as the standard for rock mass
strength estimation and is routinely applied to rock mechanics problems. The
criterion was developed during the 1980’s (Hoek and Brown, 1980) based on the
results of research into the brittle failure of intact rock and simulations of jointed
rock mass behaviour. The criterion uses the properties of intact rock and applies
some reduction factors based on the characteristics of jointing to represent a large-
scale peak strength response.
During 1983, most numerical modelling codes were written in terms of the Mohr-
Coulomb criterion and Bray (1983) developed a solution to relate the non-linear m
and s parameters of the Hoek-Brown strength envelope to the Mohr-Coulomb
strength criterion. This allowed Hoek-Brown strength to be expressed in terms of
cohesion and friction and therefore be easily used in numerical modelling
simulations of rock mass strength.
Using the method developed by Bray, values for cohesion and friction can be
obtained by a least-square fit to the Hoek-Brown failure envelope. A bi-linear,
Mohr-Coulomb fit to the Hoek-Brown curve can be used to more accurately
represent the actual non-linearity of the failure envelope. The range of stress over
which the Mohr-Coulomb properties are fit can be limited to ensure a better match
over the range of expected induced stresses. The impact of applying a bi-linear
versus linear fit can be seen in Figure 28a along with the requirement of estimating
the range of 3 expected (Figure 28b).

Figure 28. Development of equivalent Mohr-Coulomb property estimates from a fit to

the Hoek-Brown curve.
It can be seen that a bi-linear fit provides a more accurate strength estimate over
the entire range of expected stresses. The use of a low 3 value leads to the over-
prediction of rock mass strength at high 3 values if the induced stresses reach this
range.
However, the Hoek-Brown strength envelope only considers the response of the
rock mass up to the point of failure. The post peak response, which exerts a strong
control on rock mass caveability (Pierce et al., 2006) is not described. In order to
do this, separate consideration of the post-peak behaviour is required.
2.2.2.1 Cohesion and Tension Softening
It can be seen from the conceptual model of caving (Figure 6) that the boundaries
between each of the cave regions are not sharp. It is therefore apparent that
during caving, a rock mass undergoes a gradual change in strength from its in situ
state to a caved-rock state. This behaviour is described as strain-softening. The
rate of change of the strength reduction is affected by the intact strength, joint
configuration and the confining stress level causing failure (Tiwari and Rao, 2006).
Experimental stress-strain curves have previously been idealised to represent
different material behaviour typically used in numerical analyses and are

Figure 29. Idealised stress-strain curves representing different material behaviour used
in numerical modelling.
Linear Elastic - An isotropic elastic model provides the simplest representation of

material behaviour. This model is valid for homogeneous, isotropic, continuous
materials that exhibit linear stress-strain behaviour with no hysteresis on
unloading. There is no representation of the post-peak response. In reality, rock
only exhibits elastic behaviour until a certain yield stress is reached, beyond which
it ceases to behave elastically.
Perfectly Plastic - During perfectly plastic straining, plastic strains continue

indefinitely at constant stress. The ratio of plastic strain is related to the yield
stress, which also represents the failure stress. A perfectly plastic rock is
characterised by the assumption that the stress causing the permanent non-
recoverable strain must reach a certain value before any extension or contraction
can take place. When the yield stress is reached, the rock deforms permanently
and continues to yield at this stress.

Linear Elastic-Perfectly Plastic - An assumption is made by an elastic-perfectly

plastic relationship that the stress - strain response can be represented by two
straight lines that describe an initial linear elastic stiffness and the yield stress or
strength at failure during plastic straining. Rock behaves elastically for stresses
less than the yield stress, then deforms without limit at the yield stress. A rock that
exhibits such a response can be considered perfectly ductile.
Perfectly Brittle - Rock that exhibit a stress-strain response similar to Figure 29E
are called perfectly brittle materials. For a brittle rock, the stress strain curve is
nearly linear at all stress levels, up to and including the final fracture stress. Brittle
failure is the process by which sudden loss in strength occurs. During brittle
failure the strength of the rock mass reduces to a residual value instantaneously.
Strain-softening - When stress has exceeded the elastic limit, the rock mass
begins to yield. It continues to yield until the peak strength is reached before
reducing to a residual value.
Strain-softening behaviour best describes a rock mass response during caving

since it is able to represent the progressive nature of the strength reduction. A
strain-softening model has previously been used by Pierce et al. (2006) to simulate
the complex process of the progressive failure and disintegration of a rock mass
from an intact, jointed material to a bulked rock mass during the caving process.
An example of simulated strain-softening behaviour in is provided in Figure 30. In
this instance, both the intact and joint response is considered during simulated
sample straining.

Figure 30. Stages of damage within a three-dimensional, strain-softening specimen.
Within a strain-softening material model the post-peak strength behaviour is a

function of plastic, shear-strain-dependent rock mass cohesion and tensile
strength. Accumulated plastic shear strain, 𝜖 , or more specifically, the second
invariant of the deviatoric plastic strain tensor is a common metric for irreversible
shear strains in geo-materials and, in a more general sense, can be considered a
measure of damage. The plastic shear strain required in going from peak strength
to a fragmented rock mass defined by zero cohesion defines the critical plastic
strain, 𝜖 , or brittleness of the rock mass. In the absence of any experimental data,
the relationship is considered to be linear.

The study of the post-peak behaviour of rock is limited (Hoek and Karzulovic,
2000; Russo et al., 1998; Cai et al., 2007) since few apparatus exist that have the
capacity to test large-scale samples. In addition, it is known that the post-peak
behaviour of rocks tested in the laboratory is dependent on specimen geometry
(height: diameter ratio) (Hudson et al., 1971). As a result of this, the current
understanding of the strain-softening behaviour of jointed rock masses is based
largely upon large-scale numerical back-analyses and physical observations.
2.2.2.2 Estimation of Critical Plastic Strain, 𝜖
Cundall et al. (2005) have previously estimated the FLAC3D (Itasca, 2009), post-
peak softening rate (𝜖 ) of a rock mass based on the numerical back-analyses of a
number of large-scale case studies. The data is presented as the blue data points in
Figure 31.
Figure 31. Summary of FLAC3D critical strain relation and data points used for
fitting.
The results indicated a dependence of 𝜖 on the Geological Strength Index (GSI)

and, by linear regression:
𝜖 ( ) [2]
Where GSI is the Geological Strength Index and 𝜖 is the critical plastic strain of a
10m zone in FLAC3D. It is important to note that this equation is only valid for GSI

values ranging from 0 to 98. The value may be scaled with respect to edge length
by Equation [3].
𝜖 ( ⁄ ) [3]
Where z is the element width or edge length of zone in model.
The 𝜖 values derived from this relation are consistent with the typical stress-
strain relations provided by Hoek and Brown (1997) for strain-softening rock
masses which show that rock masses with higher GSI values are more brittle than
rock masses with a lower GSI.
Since the development of the 𝜖 relation by Cundall et al. (2005), additional case
studies have been completed as part of this research (presented in Sections 11, 12,
13 and 14) and are provided as the red data points in Figure 31. A new linear
relation [4] is proposed to determine a 𝜖 value that fits the combined datasets
and is valid for all GSI values from 0 to 100.
( )⁄
𝜖 [4]
In the case of caving, softening of the tension and cohesion at this rate should be
applied to ensure the rock mass in its post-peak state is represented with the
greatest accuracy.
2.2.2.3 Mesh Dependency
Trueman and Mawdesley (2003) suggest that numerical methods that use a strain-
softening approach are not robust since the post-peak response is highly sensitive
to mesh size. This is true, if mesh dependency is not accounted for in the
development of material property responses.
There are at least two methods that can be used to alleviate the problem of mesh
dependency as outlined by Dawson and Cundall (1995). They include Cosserat
Theory and the Standard Regularisation Method.

2.2.2.3.1 Cosserat Theory
Cosserat theory of elasticity considers material as a continuous collection of

particles that behave like rigid bodies. The response is dependent on the elastic
constants and internal scale length for the material. Sluys (1992) shows that a
cosserat continuum for representation of rock mass failure is only applicable when
frictional slip is dominant and that micro-polar rotations have no influence on
mode I tensile failure. This is a potential weakness of the method, since direct
modelling of mode I tensile failure is important for simulation of the propagation
of the cave back as shown by Brown (2003).
2.2.2.3.2 Standard Regularisation Method
The standard regularisation method relies on the calibration of the softening law
to the results of laboratory triaxial tests. For the calibration, the laboratory test is
simulated numerically for a given zone size, and a best-fit is found by varying the
numerical softening rate. If the same zone size is used for the problem simulation,
no regularisation is necessary. If the zone size is increased, the numerical
softening rate is increased in proportion, and vice-versa. Here, the best-fit can be
determined by Equation [4] if no other information is available.
To account for the mesh dependency, a displacement applied to the boundary of a

body has previously been considered by Detournay (2009). If the strain localises
within the body, the applied displacement appears as a jump across the localised
band. The thickness of the band contracts until it is equal to the minimum allowed
by the grid (i.e., a fixed number of element widths). This is further discussed in
Section 2.2.2.3.4. Thus, the strain in the band, , can be described by Equation [5].
⁄ [5]
Where is a fixed number, is displacement jump and is the element width.
If the softening slope is linear, (i.e., the change in a property value p , is

proportional to strain), the change in the property value with displacement can be
determined by Equation [6].
⁄ ⁄ [6]

Where s is the softening slope, is the change in material property value and
is the change in displacement.
In order to obtain mesh-independent results, a scaled softening slope can be input

such that the slope is dependent on the element width as shown in Equation [7].
[7]
Where is constant and relates to Equation [4] and, in the case ⁄ is

independent of and can be re-written as Equation [8].
𝜖 ⁄ [8]
Using this relationship, for example, if the zone size is doubled, then the critical
strain must be halved for comparable results.
Through this approach Lorig (2000) showed that cave simulation results are
repeatable with different sized meshes. His results are provided in Table 2.
Table 2. Cave height as a function of brittleness (after Lorig, 2000).

Critical Plastic Strain Cave Height
Grid ( ps crit ) (m)
Coarse 0.1 160
Fine 0.02 150
Coarse 0.005 200
Fine 0.01 205
Coarse 0.0025 225
Fine 0.005 250
This scaling approach has also been used by other researchers to account for mesh
dependency in the numerical simulation of other geomechanical processes (Crook
et al., 2003).
2.2.2.3.3 Scale-Dependent Brittleness
It can be shown that for the same numerical zone size, increasing the size of a test
sample will result in a more brittle post-peak response. This is shown in Figure 32
where Unconfined Compressive Strength (UCS) testing has been conducted on a
strain-softening material. The zone size and critical strain have remained

consistent between all simulations, along with the applied loading rate at the
specimen ends.
Figure 32. Post-peak response as a function of zone resolution controlled by sample

width.
It can be seen that, as the sample increases in size, the peak-strength remains the
same, but the post-peak response becomes more brittle. This phenomenon has
been considered by Hajiabdolmajid and Kaiser (2003) and is believed to be the
result of differences in failure mechanisms leading to different material brittleness
– or “scale-dependent brittleness”.
This phenomenon has been described by Detournay (2009) by considering a

triaxial test on a sample of height L. For simplicity Detournay considered that all
the deformation is concentrated along a band, and that the band is at an angle 
with the direction of the axial load. The slope of the axial stress-strain behaviour
at the platen can be then given by Equation [9].

[9]
Where is the axial reaction pressure, is the global axial displacement (both
measured at the platen), L is the sample height and is the axial strain.
With all the deformation being associated with the band, the quantity u can be
considered as a material band property. This may be achieved since softening
behaviour may be understood by considering that, viewed from the outside, the
band behaviour is characterised by a stress-displacement law with the slope (S)
characterised by Equation [10].
⁄ [10]
Where α is the band angle in reference to the direction of the axial load.
Thus, the overall stress-strain law is dependent on sample size, and the larger the
size, more brittle the global behaviour. So, in performing simulated laboratory
tests for calibration of material responses, the observed softening rate will be
proportional to sample size and the observed mode of localisation.
Similar modelling results can be achieved in each of the test samples by:
 Increasing the zone resolution on the smallest sample size to be consistent

with that of the largest samples size. This will yield the same brittle post-
peak response as the largest sample ; or
 Decreasing the zone resolution on the largest sample size to be consistent

with that of the smallest sample size. This will yield the same post-peak
response as the smallest sample.
The post-peak brittleness of the numerical constitutive model must be an

emergent behaviour since, small and large-scale failure mechanisms must be able
to develop naturally to allow the natural cave propagation behaviour. Calibration
of material responses must be undertaken at a scale consistent with the detail of
failure that is required to be resolved within the simulation. The selection of zone
and sample size is further discussed in Section 2.2.2.3.4.

2.2.2.3.4 Localisation and Bifurcation
Zones of localised deformation (shear bands) are a common feature of brittle

jointed rock masses that have failed in compression, both in the laboratory and
naturally as earth faults (Rudnicki and Rice, 1975).
Localisation refers to the “occurrence of strong strain gradients in specific areas of a

material that finally become discontinuities” (Varas et al., 2005). The possibility of
localisation occurs when one or more stress components in an element are able to
decrease with increasing strain. Cundall (1991) suggests that there are three
possible ways that stress in an element can decrease with increasing strain:
 large geometrical distortions (i.e., buckling of a thin beam)
 material softening in which the intrinsic material becomes weaker (i.e.,

dilatant material becomes looser and hence weaker), or a
 change in stress state such that at least one stress component decreases.
It is the latter two of these possibilities that are expected to occur during caving.
Localisation has been shown to occur by Santarelli (1989) and Besuelle (2001) at
stresses levels of between 60% and 90% the peak strength. However, shear bands
are already in place by the time that the material softening commences. The
development of shear bands is triggered by very small local variations in the initial
conditions of the problem – known as bifurcation. Bifurcation is well known in
laboratory settings, where, for example, in a simple shear test, a sample may either
deform uniformly or develop shear bands. Within numerical modelling codes,
bifurcation entails the occurrence of multiple solutions compatible with
equilibrium equations and boundary conditions.
In a previous study, Cundall (1991) shows that the process of shear band
formation is one of crack coalescence rather than propagation. When the
numerical simulation is run for a long time, bands are seen to coalesce, grow both
in length and intensity and finally become dormant. Furthermore, Cundall (1991)
also shows that the formation of one band will inhibit the formation of another
laterally nearby, i.e., two bands cannot form close together. The inhibiting effect

(i.e., build-up of confining stress at given locations due to a mean stress decrease in
the evolving band) is proportional to the band thickness and is related to
instantaneous band length. The localisation band thickness in a numerical test is
equal to about three element widths whereas in reality they would be related to
grain size (Cundall, 1991). By default, the thickness of shear bands in a numerical
mesh reduces to the smallest dimension resolved by the mesh, which is the zone
size (typically one zone width if the band is parallel to the mesh, or three zones
width if the band cuts the grid at an arbitrary angle). Therefore, although
numerical solution schemes are able to reproduce the physics of band formation
accurately, the thickness of the bands and their spacing are strongly dependent on
the resolution of the numerical mesh.
Through a series of numerical simulations unloading a circular excavation in a

strain-softening material, Varas et al. (2005) confirmed that mesh dependency
controls the bifurcation phenomena in numerical models and suggest that
calibration of a material response should be conducted using a mesh size similar to
that required in the large scale model.
Provided that the model mesh is fine enough (compared to the problems size) then
any strain-softening numerical simulation should be able to capture the correct
evolving failure mechanism within the rock mass. Cundall (1991) has shown that,
with enough elements within a mesh, many thin shear bands can be resolved in a
continuum calculation.
2.2.2.4 Independent Tension Softening
The softening of a rock mass can be attributed to strain-based damage manifesting

as intact rock block damage (accumulation of micro-cracks) and joint
slip/dislocation and/or extension. So far, through a standard strain-softening
relation, only shear softening has been considered. Based on the review of fracture
modes within a cave mine conducted by Lorig et al. (2000), shear failure is only
one failure mechanism that causes cave propagation. Tensile failure can also be
expected.
Figure 33 shows how it is feasible that tension can soften independently of

cohesion.

Figure 33. Schematic diagram of a tensile failure mechanism that does not affect
cohesive strength.
This phenomenon has previously been described by Pine et al. (2007). “Crack
growth orthogonal to the direction of dilation in a compressive stress field does not
immediately produce a mechanical instability, as observed in tensile fields…It is this
stable fracture process in compression that results in large differences between
tensile and compressive strengths.”
Palma and Agarwal (1973) have previously considered this mechanism of

independent tensile softening in an analysis of caving at the El Teniente Mine
through a continuous sampling of stresses within a numerical model. The same
approach is proposed for the numerical model of cave propagation. In addition to
softening tension and cohesion at the same rate based on plastic strain (discussed
in Section 2.2.2.2), tension should be allowed to soften independent of cohesion in
the instance of a tensile yield state within the model. This will ensure that a
gravitational mode of caving can accurately be represented within the model and
develop independently of a stress caving mechanism.
The implementation of such a relation suggests that a rock mass is perfectly brittle
in tension. Previous investigations by Cai (2010) suggest that this is a reasonable
assumption.

2.2.2.5 Friction Hardening
If linear Mohr-Coulomb fits are derived for a Hoek-Brown strength envelope at

varying 3 values (see Figure 34), it is clear that; as confinement decreases,
friction increases and cohesion decreases. This suggests some dependency of
cohesion and friction on confinement levels.
Figure 34. Development of equivalent linear Mohr-Coulomb strength parameters based

on a fit to the Hoek-Brown strength envelope.
Through laboratory testing, Schmertmann and Osterberg (1960) showed that the
two strength components, cohesion and friction are not necessarily mobilised
simultaneously during straining and that the cohesive component of strength is
mobilised early in compression tests, while friction (and dilation) requires
additional straining to reach full capacity.
Previous research conducted by Diederichs (2007) has shown that the response of
a rock mass during stress-yielding is predominantly a result of a tensile failure
mechanism, as pre-existing joints propagate. As a result of this, the impact of
friction (and dilation) is limited until a failure plane localises.

Hajiabdolmajid, Kaiser and Martin (2002) have explored this concept in relation to
failure in laboratory specimens and around underground openings. They relate
the mobilisation of the cohesion and frictional strength components to
accumulated plastic strain. A sketch showing their conceptual model is presented
in Figure 35.
Figure 35. Schematic diagram of the mobilisation of the strength components cohesion
and friction (a) in the laboratory (b) around an underground opening; ci and
cr and εcp and εfp represent the plastic strain components when the friction
and cohesion strength components have reached ultimate values (after
Hajiabdolmajid, Kaiser and Martin, 2002).
The simplest Cohesion Weakening Friction Strengthening (CWFS) model has

previously been described by Ettema et al. (1989) as a bi-linear function. “One
(friction) value is taken at the peak of the stress-strain curve … Its value is affected by
initial porosity and confining pressure prior to shearing. The other value is taken
after considerable straining … when further straining occurs without significant
change in either porosity or confining pressure.” This is known as the constant or
residual friction angle. Pierce et al. (2006) have previously reported residual
values of 43-45 degrees for jointed rock masses.

In order to account for the generalisation of frictional strength by Ettema et al.

(1989), a dynamic cohesion and friction model was proposed by Hoek, Kaiser and
Bawden (1995) whereby instantaneous friction and cohesion values were
calculated based on the relationships between normal and shear stress, joint
condition (JRC and JCS) and an initial estimate of a Basic Friction Angle ( ).
Although good modelling results were achieved, limitations associated with the
scale of the problem arise when implementing such an equation when joint
condition and orientation with respect to normal and shear stress is not constant.
Hajiabdolmajid and Kaiser (2002), Diederichs (2002) and Zhao and Cai (2010)
propose a plastic-strain dependent CWFS model whereby the softening
characteristics are related to site-specific case study applications. Within this
constitutive relation, after mobilisation to a peak value, friction is gradually
reduced to a residual value. An example of such an implementation is provided in
Figure 36.
Figure 36. Implementation of the CWFS model in a two-dimensional numerical model

of a tunnel failure (after Barton and Pandey, 2011).

A CWFS model has successfully been used in the back-analysis of two breakouts in
different rock types at the Underground Research Laboratory (Hajiabdolmajiod et
al., 2002) as well as in two slope failures in jointed rock (Hajiabdolmajiod and
Kaiser, 2003; Eberhardt et al., 2002). However, it is noted that in a numerical
back-analysis of one of the tunnel break-outs presented in Hajiabdolmajiod and
Kaiser (2002), the non-uniformity of the mesh (i.e., not aspect ratio zones) and no
discussion of critical strain scaling may be affecting the simulation results.
Pierce et al. (2006) have modelled friction as a strain-softening value whereby the
peak value has been estimated by a linear Mohr-Coulomb fit to the Hoek-Brown
curve. Friction is modified to a residual value in response to plastic shear strain (at
the same rate as cohesion) and volumetric strain. Barton and Pandey (2011)
developed a similar approach; however, their friction value is dynamic and has
been calibrated based on the application at two case study locations.
In order to account for the CWFS behaviour of a rock mass, it is proposed that the
peak friction angle/s be estimated by the bi-linear technique described in Section
2.2.2. The values should be modified to a residual value of 43-45 degrees after a
fully bulked rock mass state is achieved and/or at the same softening rate as
cohesion.
A fully bulked rock mass state (or maximum volumetric strain achievable) can be
estimated based on the maximum porosity (η) of a rock mass previously
determined by Pierce et al. (2006) to be approximately 0.4. Based on this, a
volumetric strain cut-off, , of approximately 66% can be determined by
Equation [11].
( ) [11]
Where η is the rock mass porosity. In this instance, the fully bulked rock mass will
have properties consistent with gravel e.g., frictional strength only.
2.2.2.6 Summary
It is generally accepted by the geotechnical engineering community that the Hoek-

Brown Failure Criterion, is the most widely adopted standard for expressing the
strength of a jointed rock mass. It is therefore considered the starting point for the

representation of rock mass strength for a cave propagation and subsidence

assessment.
As a result of this, a bi-linear CWFS constitutive relation is proposed for the

numerical model of cave propagation. Initial cohesion and friction values are
determined from a bi-liner Mohr-Coulomb fit to the Hoek-Brown curve.
Cohesion and tension should be softened to a residual strength of zero in relation

to 𝜖 which can be determined from Equation [4]. In addition, tension softening
must be allowed to occur independently of cohesion softening through the
constant querying of plasticity states. A residual friction angle of 43-45 degrees is
simulated once a maximum volumetric strain is exceeded and/or at the same rate
as cohesion. This maximum value can be determined by Equation [11].
Provided mesh dependency is accounted for within the calibrated rock mass
response, the simulated laboratory response of a bi-linear, strain-softening rock
mass under uniaxial compression and triaxial compression loading conditions
(illustrated in Figure 37) provides the general elastic, peak strength, post-peak
softening and dilatancy mechanisms expected in an isotropic rock mass as
confinement is increased.

Figure 37. Example of simulated bi-linear, strain-softening response (after Sainsbury

et al., 2010).
However, limitations associated with the application of this failure criterion

include:
 Assumes an isotropic rock mass, i.e. same strength in all directions.

Previous studies by Mas Ivars et al. (2008) shows that most jointed rock
masses display some anisotropic strength characteristics.

 No consideration of the post-peak response. This is required to be

estimated based on an empirical relation developed by Cundall et al. (2005)
and modified herein (Equation [4]).
 Developed by extensive curve fitting. Its extrapolation beyond the data

limits (i.e., in rock masses that exhibit very low or very high GSI values), and
rock-type for which the curve was developed has not been validated.
 Requires the selection of a confining stress for equivalent Mohr-Coulomb

strength parameters. The selection of the stress range poses problems
when conducting large-scale analyses under a number of different
confinement conditions (depths) or stress paths.
 The strength is derived for a specific sample size – in relation to the

problem description. There is no way to simulate the scale-dependent rock
mass strength response within the one analysis.
2.2.3 The Synthetic Rock Mass (SRM) Modelling Approach
Due to the inherent difficulty of testing large-scale rock mass samples in the
laboratory or field, reliance has been placed on empirical classification rules and
systems derived from practical observations (i.e., GSI, MRMR, etc.). Despite the fact
that these systems and relations are in widespread use in engineering design, their
ability to consider strength anisotropy (resulting from a preferred joint fabric
orientation), scale effect (resulting mainly from the combined effect of joint density
and joint persistence), and post-peak strength response remains limited.
Previous investigations (Hoek and Brown, 1980) have shown that the strength and
deformation behaviour of a jointed rock mass is governed strongly by (a) the intact
strength of the rock and (b) the presence of joints/discontinuities. The Hoek-
Brown strength criterion accounts for each of these components implicitly by
using a global strength value and smearing the effect of joints through an isotropic
rock mass response. A technique is required that allows for the independent
failure and degradation of each of these components (joints and intact blocks).
Based on the small-scale failure mechanisms that are required to be resolved in

order to simulate a propagating cave, Synthetic Rock Mass (SRM) modelling was

developed (Pierce et al., 2006) to allow for the detailed consideration of the rock
mass joint fabric in the determination of rock mass response at large scales – i.e. 10
to 100 m. The SRM methodology uses PFC3D (Itasca, 2007) to explicitly represent a
Discrete Fracture Network (DFN) embedded within an intact rock matrix.
The intact rock matrix is simulated using the Bonded Particle Model (BPM). The
BPM represents the rock as rigid particles (grains) glued together at their contacts
by parallel bonds that represent a normal and shear stiffness. As a result of these
bonds, the BPM does not impose theoretical assumptions and limitations on
macroscopic material behaviour, as continuum models do. Micro-cracks are able
to form, interact, and coalesce into macroscopic fractures according to local
conditions. In this manner, macroscopic material behaviours not encompassed by
current continuum theories can be investigated. The BPM has a demonstrated
ability to reproduce many features of rock behaviour, including elasticity,
fracturing, acoustic emission, and damage accumulation producing material
anisotropy, hysteresis, dilation, post-peak softening, and strength increase with
confinement (Mas Ivars et al., 2011); all of which have been validated based on
instrumented laboratory tests. The micro-properties of the intact rock in SRM
samples are chosen via a calibration process based on matching laboratory test
results (intact rock UCS, Young’s modulus, and Poisson’s ratio).
Discontinuities within the rock mass samples are represented via the Smooth Joint
contact Model (SJM) which allows the simulation of a smooth interface between
particles regardless of the local particle contact orientations along the alignment as
shown in Figure 38.

Figure 38. The Smooth Joint Contact Model. (a) Graphical representation of how the
smooth joint contact model can be used to realign the default contact
orientation to one that honours a macroscopic joint orientation. (b) By
using the smooth joint contact model to reorient all contacts lying along the
macroscopic joint plane, sliding along a smooth planar feature can be more
accurately simulated (after Mas Ivars et al., 2011).
With the SJM, macroscopic joints with a given dimension and orientation can be
embedded within the assembly and can experience shearing in the manner of a
smooth, frictional surface without resorting to particle size refinement or particle
relocation along the joint surface (Mas Ivars et al., 2011). It has been demonstrated
that the model can be used to reproduce the extension and coalescence of multiple,
isolated, embedded flaws observed in laboratory experiments (Deisman et al.,
2008).
A SRM sample can explicitly account for the presence of intact rock bridges
between terminating fractures – similar to in situ rock mass conditions. Through
simulated testing of the synthetic samples, it is possible to derive large-scale rock
mass failure mechanisms and properties such as deformation modulus, strength
and brittleness. An example SRM composite sample is presented in Figure 39.

Figure 39. Components of a Synthetic Rock Mass sample. (a) Three-dimensional

DFN, (b) the corresponding three-dimensional synthetic rock mass sample,
and (c) synthetic rock mass basic components. The colours in (b) and (c)
denote intact rock blocks bounded by joints. Notice the internal non-
through-going jointing in the ‘‘intact’’ rock blocks (after Mas Ivars et al.,
2011).
Simulations of uniaxial compression testing on a SRM sample are presented in

Figure 40. For each of the tests, the full stress-strain curve and the percentage of
bond breakages versus total initial bonds have been recorded (percent damage).
These parameters provide an estimate of the pre-peak and post-peak behaviour of
the rock mass along with an estimate of the amount of intact damage occurring
within the sample during loading.

Figure 40. Three-dimensional response of a synthetic rock mass sample tested in three-
opposing directions under unconfined compression; testing directions east-
west, north-south and vertical (after Sainsbury et al., 2009).
The potential power of SRM is that it allows for site-specific consideration of joint
fabric, loading conditions and material property variations. This may be
particularly useful in cases where the joint fabric is highly anisotropic (as shown in
Figure 40) or where the problem is sensitive to post-peak strength. In addition,
large-scale samples can be tested in significant detail – ensuring an accurate failure
mechanism is represented.
Based on a review of the SRM technology, Hoek and Martin (2010) believe that
“there is no doubt that the tools assembled in the SRM approach are the most
advanced available to us today” and that they “believe the physics in the SRM
approach is sound”. In addition they go on to say that “we believe very strongly that
the Bonded Particle Model (BPM), the modelling foundation for the SRM, is the only
commercially available code that can be used to properly capture the behaviour of
intact rock. Sufficient published and independent research using the BPM has been

carried out over the past 10 years to show that the BPM logic can be used to match
the laboratory behaviour of intact rock… We are optimistic that the SRM
methodology should be suitable for studying the behaviour of discontinuous rock
masses; because we believe it has the best representation of the processes that are
active in a yielding discontinuous rock mass… Claims have been made by users of …
ELFEN … that many of the SRM features can be reproduced by these codes. However,
it is difficult to verify some of these claims as critical comparisons are seldom carried
out”. It can be said that this statement rings true for just about all of the existing
numerical techniques for cave propagation assessment (reviewed in Section 2.1),
apart from the technique outlined by Pierce et al. (2006) since, little validation
against real-life case studies has been completed and documented in detail for
verification purposes.
The review by Hoek and Martin endorses the SRM approach developed in PFC3D for
rock mass characterisation above all other SRM approaches that have been
proposed (e.g. Vyazmensky et al., 2007; Beck et al., 2007) and therefore forms the
basis for the characterisation of rock mass strength in the numerical model of cave
propagation behaviour and subsidence assessment detailed herein.
2.2.3.1 Application of Synthetic Rock Mass Modelling for Cave Propagation

Assessment
Previous caving investigations have been conducted using the SRM approach at the
Northparkes Mine (Pierce et al., 2007). Through simulated testing of synthetic
samples and carrying tests through to complete disintegration (residual strength),
both pre-peak properties (modulus, damage threshold and peak strength) and
post-peak properties (brittleness, dilation angle, residual strength and
fragmentation) were developed as shown in Figure 41.
The material response of SRM samples has been validated through a comparison of
fracture orientations produced when a SRM sample is subjected to the same
mining induced stress path and that has seismic monitoring data available (Figure
42a); and a comparison of SRM fragmentation predictions with drawpoint
observations (Figure 42b).

Figure 41. Stress-path dependent Synthetic Rock Mass approach (a) stress path, fitted
peak-strength envelope (two separate stress paths), and damage threshold
observed in SRM sample at Northparkes Lift 2 (b) estimates of brittleness
obtained from SRM testing (after Pierce et al., 2006). Steep lines represent
higher brittleness.

Figure 42. Validation of Synthetic Rock Mass response based on observed and
measured fracture modes and fragmentation (after Pierce et al., 2006).

In the case study of Northparkes (above) it can be seen that the microseismic
response generated in SRM samples matches with seismic observations
underground. In addition the predicted SRM fragmentation, after stress-path
dependent testing, closely matches with fragmentation measured at production
drawpoints.
In order to use SRM strengths in a large-scale analysis, Pierce et al. (2006)

calibrated continuum constitutive model responses. Although good results were
achieved when cave propagation behaviour was simulated for a specific instance
as shown in Figure 43, the SRM simulations were only applicable to those domains
and stress paths immediately surrounding the Northparkes Lift 2 cave.

Figure 43. Development of a large-scale caving model using stress-path dependent

Synthetic Rock Mass strengths (a) SRM sample locations relative to the
main stages of mining Northparkes Lift 2. (b) predicted cave shape; caved
zone - white iso-surface, yield zone - blue iso-surface, TDR breakages - blue
sphere, open hole blockages - red squares (after Mas Ivars et al., 2011).
To conduct further analyses (e.g., undercut footprint expansions, etc.) a whole new
set of SRM simulations were required which makes the methodology inefficient
since the costs associated with developing, testing and implementing a SRM
strength model are significant.

As a result of this, the more generalised SRM testing methodology (described by

Mas Ivars et al. 2008) is considered more universally appropriate for
implementation in caving models. Mas Ivars et al. (2008) describe how the SRM
methodology has been developed to allow testing of a rock mass in all three
opposing loading directions and at a number of different scales. Three industry-
standard tests, (direct tensile test, uniaxial compressive-strength test and triaxial
test) were selected to provide measures of rock-mass tensile strength ( ),
unconfined compressive strength ( ) and compressive strength at several
confinement levels (σ3i, σ3ii, etc). This ensures that the material constitutive
properties derived from this technique are not specific to one particular stress
path, and they may be applied to a number of different large-scale
mining/geological processes.
2.2.3.2 Summary
Based upon the limitations of empirical techniques to consider, strength

anisotropy and scale effects, the SRM technique is considered an advancement in
the determination of rock mass behaviour when compared to the Hoek-Brown
approach. A review of SRM technology (Hoek and Martin, 2010) has provided a
positive response and their assessment concludes the SRM provides “(the best)
representation of the processes that are active in a yielding discontinuous rock mass”.
However, the application of three-dimensional SRMs to large-scale (e.g. >100 m)

boundary value problems in densely jointed rock has been relatively limited to
date. Current limitations of the technique include computational constraints that
continue to restrict application of the technique to smaller-scale boundary value
problems. Pierce et al. (2006) have previously shown how a continuum
constitutive model can be calibrated to SRM strength responses. Cave propagation
analyses conducted using this technique provided numerical results that match
well with the observed conditions during cave propagation at the Northparkes E26
Lift 2 Mine. However their responses were limited to a specific application and a
more generalised calibrated continuum response is required.

In lieu of SRM strength responses, the Hoek-Brown strength criterion can be

considered in instances when strength anisotropy or scale effects are expected to
be minimal.

3 RESEARCH OUTLINE
3.1 Objectives
The ability to assess the advance rate and shape of a propagating cave in response
to a production draw schedule is critical in being able to plan mining infrastructure
since, once production commences, the ability to modify the mine design is limited.
The requirement to make predictions with confidence, from drill-core data, prior
to the rock mass being exposed on a large-scale (i.e., scanline mapping of a drift) is
of paramount importance.
Based on a review of the current literature, it is clear that numerous approaches to

assess cave propagation and subsidence behaviour exist. The objective of this
research is to develop a state-of-the-art numerical model of cave propagation and
subsidence assessment that is rigorous and robust and that can be used to focus
geotechnical studies and provide a greater understanding of cave propagation and
subsidence behaviour.
The numerical model will extend the existing numerical techniques reviewed in
Section 2 through the improvement of:
 the existing height of draw production scheduling technique described by

Pierce et al. (2006) and implementation of a mass based production
schedule as a direct input to the numerical modelling algorithm.
 the simulated rock mass constitutive behaviour that governs the volumetric
response of a material due to caving. This includes the development and
implementation of relationships between volumetric strain and density,
dilation and deformation modulus.
 simulation of rock mass response around a propagating cave due to the

presence of large-scale persistent structures.
3 – Research Outline
And the development of:
 a validated constitutive model that is able to account for rock mass strength
and deformation anisotropy as described by Synthetic Rock Mass testing
outlined by Mas Ivars et al. (2008).
 a generalised production algorithm that is able to simulate all three, block,

panel and sub-level caving scenarios.
 an algorithm that is able to update the ground surface profile as a result of

the development of a subsidence crater. This will ensure that the
subsidence limits predicted by the numerical model will be rigorous.
In addition, the development of the numerical model must:
 ensure that the requirement for geotechnical information is limited to that

which is already contained within a typical geotechnical database.
 not rely on assumptions having to be made regarding initiation, cave shape

and the rock mass dilation (bulking) response to ensure that cave stalls can
be predicted.
 be based on a clear and unambiguous methodology which is thorough,

robust and transparent and that provides results that are clear, easy to
interpret and meaningful. The results should be presented in terms of
evolving cave shape, interaction with other mining areas, location/extent
of cave behavioral domains, bulking factors, propagation rates, expected
magnitude and extent of damage (surface and underground) and potential
for cave stalling.
 consist of well documented and transparent algorithms that can easily be

implemented within any commercially available numerical modelling
package.
 be validated against a number of case study back-analyses.
3.2 Methodology
Based on a review of the current cave assessment techniques, outlined in Section 2,

it is clear that the caving algorithm described by Pierce et al. (2006) represents the
current state-of-the-art in rock mechanics for cave propagation analyses. In
addition to addressing five of the six key rock mass behaviours that affect caving
(defined by Pierce, 2009 and Brown, 2003 – outlined in Section 1.3) the Pierce et
al. (2006) caving model has been extensively documented and can be implemented
in any commercially available modelling code that can support a strain-softening
constitutive model.
The current research aims to further develop the caving model described by Pierce
et al. (2006) – herein described as the 2006 model - through the consideration of
each of the key rock mass behaviours that affect caving – as outlined in Section 1.3.
At the current time the 2006 caving model is limited by:
 Stress-Path dependent calibrated SRM-continuum material inputs,
 Linear modulus reduction relation,
 Constant rock mass dilation relation,
 Consideration of large-scale discontinuities based on an implicit

methodology
 Height of Draw (HOD) based production scheduling technique; and
 Application to block and panel caving methods only. Sub-Level caving is not
considered.
Based on these limitations, the development of a numerical model for cave

propagation can be addressed in two separate streams that include; rock mass
behaviour and production scheduling techniques.
3.2.1 Simulation of Rock Mass Response to Cave Propagation
3.2.1.1 Rock Mass Cohesion/Tension Softening and Post-Peak Brittleness
The SRM methodology uses the Particle Flow Code PFC3D to explicitly represent a
DFN embedded within an intact rock matrix. SRM samples can explicitly account
for the joint fabric and its impact on rock mass strength anisotropy and scale
effects.
Previous investigations (Mas Ivars et al., 2011) have shown that it is possible to
determine generalised cohesion/tension weakening and post-peak brittleness
from the testing of SRM samples through a standard suite of laboratory tests. A
procedure is required to allow for the accurate representation of the standard
suite of SRM responses in cave propagation models.
3.2.1.2 Rock Mass Dilation
An accurate assessment of the peak rock mass dilatancy within and around an
evolving cave is crucial to ensure that the bulking behaviour of the caved mass is
accurate. Few of the documented cave propagation assessment models discuss the
consideration of this parameter. A critical review of the current understanding of
rock mass dilational behaviour and existing relations is required. This review will
form the basis of the formulation of a rock mass dilational response for cave
propagation and subsidence assessment.
3.2.1.3 Rock Mass Deformation Modulus Softening
During cave propagation, the rock mass increases in volume as intact rock blocks
fracture, separate and rotate during the yielding and mobilisation process. Along
with this bulking, a reduction in the deformation modulus is expected to occur.
Representation of the decrease in deformation modulus is crucial for assessing the
evolving stress state around the cave, since, as rock mass softens its potential to
carry stress decreases.
The rate at which the deformation modulus decreases from an in situ state to a
fully bulked state in response to production draw has previously been simulated as
a linear decrease based on volumetric strain in the 2006 cave propagation model.
However, Hoek and Diederichs (2006) report that deformation modulus softening
is non-linear in nature. A thorough review of published laboratory test data

associated with granular material is required to develop a relationship between
volumetric strain and deformation modulus for use in the numerical model of cave
and subsidence assessment.
3.2.1.4 Simulation of Large-Scale Discontinuities
The influence of large-scale discontinuities on subsidence and cave propagation is

recognised to be important by many researchers. In situ observations have shown
that the impact of structure can be varied based on persistence, strength and
orientation relative to the undercut footprint and major principal stress direction.
There are various numerical techniques available to simulate large-scale

discontinuities within a three-dimensional numerical model. A thorough
investigation of all approaches is required to ensure the most accurate method is
implemented within the numerical model of cave propagation and subsidence
assessment.
3.2.2 Production Draw Simulation
The development and implementation of state-of-the-art numerical techniques for

a more accurate and adaptable production draw simulation is implemented within
the numerical model of cave propagation and subsidence assessment.
3.2.2.1 Mass-Based Production Draw Algorithm
The rate of production draw and shape of the undercut footprint has previously
been identified by Laubscher (1990, 1994) to impact the caveability of a rock mass.
Evolution of the footprint shape and evolving hydraulic radius can be simulated
through the constant updating of the active production area in the model. At the
current time, most of the methodologies control production draw by a Height of
Draw (HOD) schedule that is estimated based on a pre-determined bulking factor
for the in situ rock mass.
A mass-based production draw algorithm is required to ensure the production

schedule is accurately represented in the numerical model of cave propagation.
This is most important during cave initiation when the production tonnes are low
and caving rates will be at their maximum. By allowing a cave to develop as a
result of minor variations in the production schedule, cave initiation and stalling
can be predicted with rigour.
3.2.2.2 Development of an Algorithm to Update Ground Surface Profile
When a cave breaches the ground surface a crater is formed. The development of
the crater in the numerical model of cave propagation needs to be simulated to
ensure that the subsidence limits are assessed correctly and that toppling
instability around the crater slope may be predicted and incorporated into
mobilised zone for infrastructure planning purposes.
3.2.2.3 Sub-Level Caving Algorithm
In block and panel caving, mobilisation of the ore is achieved without drilling and
blasting. Sub-level caving requires the transformation of in situ ore into a mobile
state by drilling and blasting. There is currently no documented methodology that
considers a sub-level caving production draw schedule. An algorithm is required
to be developed in the numerical model of cave propagation that considers this
cave mining method.
3.2.3 Validation
An outline for the development and validation of each of the caving algorithm
components (rock mass constitutive behaviour and production scheduling) is
provided in Figure 44. In addition, a number of case study applications are
required to be completed to validate the model response and ensure that the
methodology described herein can be applied with some confidence.
Figure 44. Research methodology plan. Red tasks indicate developments that consider
rock mass behaviour and its impact on caving. Green tasks indicate
developments that are associated with numerical modelling techniques
required to simulate caving.
4 DEVELOPMENT OF A CAVE PROPAGATION

DEMONSTRATION MODEL
To investigate the effect of the in situ geomechanical conditions and imposed

production schedule on cave propagation behaviour, a demonstration model has
been developed within the numerical modelling code FLAC3D (Itasca, 2009). The
model has been constructed to demonstrate the capabilities of the 2006 cave
propagation model and demonstrate the implementation of new developments
detailed herein for the numerical assessment of cave propagation and subsidence
behaviour.
4.1 Geomechanical Conditions
A typical one lift block cave has been developed. The undercut footprint has
dimensions of 120 m x 120 m (Hydraulic Radius (HR) of 30) and is located at a
depth of 600 m below the ground surface. The maximum principal horizontal
stress has been simulated at twice the vertical stress. Production has been
simulated as a total Height of Draw (HOD) of 10 m. A schematic representation of
the conceptual scenario is provided in Figure 45.
4 – Development of a Cave Propagation Demonstration Model

Figure 45. Development of a numerical demonstration model: geomechanical conditions.
4.2 Production Draw Simulation
To simulate the undercutting and production process at the start of each mining
advance increment, undercut zones are deleted and the support it provided to the
surrounding rock mass is replaced with reaction forces. Draw is simulated by
applying a small downward-oriented velocity to all gridpoints in the back of the
undercut. This velocity is set low enough to ensure pseudo-static equilibrium
throughout the model. Displacements that match the production height of draw in
the back of the undercut are induced in the back of the undercut zones.

4.3 Cave Propagation Sensitivity Studies
4.3.1 Effect of Rock Mass Peak Strength on Cave Propagation
The effect that rock mass strength has on propagation behaviour has been
explored in the Cave Demonstration Model. Four different strength rock masses
have been defined (RM1, RM2, RM3 and RM4) and bi-linear, strain-softening
material responses have been developed for each property set based on the
methodology described in Section 2.2.2. A summary of the measurable material
properties and Mohr-Coulomb strength estimates for each of the rock masses are
provided in Table 3 based on a minor principal stress fit of 15 MPa.
Table 3. Estimate of Hoek-Brown and equivalent Mohr-Coulomb rock mass strength

properties for four simulated domains in the numerical demonstration model.
Seg. 1 Seg. 2
UCS Erm Tens. Coh.  Coh. 

(MPa) GSI MRMR mi (GPa)  (MPa) (MPa) (Deg.) (MPa) (Deg.)
RM1 120 48 52 12 8.9 0.25 0.2 1.7 50 5.8 35
RM2 120 55 58 14 13.3 0.25 0.3 2.2 52 6.8 38
RM3 145 59 65 16 16.7 0.25 0.4 3.0 54 8.1 42
RM4 170 70 75 20 31.6 0.25 0.9 5.2 57 11.4 47
The Hoek-Brown failure envelopes for each of the rock masses and stress-strain
material responses at different confinement levels are provided in Figure 46.
Figure 46. Hoek-Brown failure envelopes and simulated rock mass stress-strain curves
for the rock mass domains in the numerical demonstration model.

The empirical estimates of caveability (after Laubscher, 2000) for each of the rock
masses are provided in Figure 47 for a HR of 30. Each of the rock masses are
classified as having a different caveability potential based on their MRMR values
ranging from ‘caving’ to ‘stable’.
Figure 47. Empirical estimates of rock mass caveability for four rock mass domains
simulated in the numerical demonstration model.
The RM4 rock mass falls within the ‘Stable’ region which suggests that cave
initiation and propagation may be problematic. Caving in the RM1 and RM2 rock
masses is not expected to be problematic. The caveability of RM3 is unable to be
determined since it falls within the ‘Transitional Zone’. The numerical simulation
results are provided in Figure 48. Propagation rates for both the mobilised and
yield zones have been calculated based on their simulated height above the
undercut level and the simulated HOD.

Figure 48. Predicted cave propagation behaviour for variable peak strength rock masses
in the numerical demonstration model.

It can be seen that as the rock mass strength increases, the propagation rate
decreases (from 18:1 in the case of RM1 to 1:1 in the case of RM4). Self-sustained
cave propagation in the lower strength RM1 and RM2 rock masses is predicted.
Stalling is seen in the simulation of caving in RM3 (depicted by the co-incident cave
and yield zones in Figure 48A) at a height of approximately 25 m above the
undercut. The cave fails to initiate in the strong rock mass RM4.
The evolution of the bulked caved mass for each rock mass is also presented for
each of the rock mass strengths through a review of the evolving rock mass
density. It can be seen that bulking is not uniform within the caved mass. Higher
bulking (lower densities) can be seen at the edges of the cave. Results of the RM4
model provide a full-bulked caved rock across the entire undercut footprint, while
the RM1 and RM2 rock masses only reach a maximum bulked rock mass density
around the cave periphery - where the shearing stress is at a maximum. These
numerical results are considered to represent more closely the actual response of a
rock mass during caving than assuming a constant reduced density for production
calculations (such as Beck et al., 2011)
Maximum principal stress magnitudes in the mining abutments of the lower

strength RM1 and RM2 rock masses reach larger values than the higher strength
RM3 and RM4. This shows that minimal stress redistribution has occurred in the
high strength cases. It is expected with the additional simulation of production
draw, abutments stresses will continue to increase in the RM1 and RM2 models
until the cave intersects the ground surface.
The Damage Threshold ( ) values plotted for each of the rock masses provide
two important pieces of information that include if damage is occurring in the rock
mass; and where this damage is occurring. This is important in understanding the
ongoing caveability of the rock mass and where the cave is likely to evolve.
The simulation of caving in each of the rock mass types agrees with the initial
estimates of caveability based on the Laubscher (2000) caveability chart. In
addition, in the case of RM4 the inability of the cave to initiate and problematic
caving was predicted in the case of RM4 and RM3 respectively are a significant
result.

The prediction of the critical hydraulic radius and the potential for cave stalling is
imperative, and, at the current time there are no other documented numerical
methodologies that are able to do this.
4.3.2 Effect of Post-Peak Softening Rate on Cave Propagation
The rate at which the degradation from the in situ to caved state occurs within a
rock mass is referred to as the post peak brittleness. Lorig (2000) has shown,
through numerical simulations of caving in two-dimensions, that cave height is
strongly dependent on the brittleness of the rock mass. In lieu of SRM testing
results to provide an estimate of post-peak brittleness, the response can be
estimated based on a relationship with GSI and zone size within the numerical
model presented in Equation [4].
The effect of variable post-peak brittleness on cave performance has been studied
in the Cave Demonstration Model by modifying the post-peak response of the RM1
rock mass. The stress-strain results of computer simulated, large-scale, laboratory
tests that represent a ductile, average, and brittle post-peak response for RM1 are
provided in Figure 49A. The actual 𝜖 values used are documented in Figure 49B.
Figure 49. Simulated variable post-peak softening responses for the same peak strength
rock mass.
In each case, the rock mass responses have been developed based on the same
peak strength properties. The only difference is the 𝜖 value required to reduce
the strength of the rock mass from a peak to residual value. The effect of each of
the post-peak responses after 10 m draw have on cave propagation behaviour in

the demonstration model have been simulated. The results are provided in Figure
50.
Figure 50. Variation in cave propagation behaviour based on variable post-peak

softening rates simulated in the numerical cave propagation model.
It can be seen that the cave propagation rate is strongly influenced by the post-
peak response of the rock mass. Problematic cave initiation and propagation is
simulated with a ductile post-peak response. Similar propagation rates are
simulated for the Average and Brittle post-peak responses, however, it can been

seen that, as the brittleness increases, the advance of the yield zone in front of the
mobilised zone also increases from 10 – 30 m up to 40 – 50 m. The lateral extents
of the yield zone (beyond the undercut footprint) are also increased in the case of
the more brittle response suggesting ‘easier’ and more vertical cave propagation
behaviour.
4.3.3 Effect of Estimation of mi value on Cave Propagation
The Hoek-Brown mi constant is an intact property that varies with rock type and is
used to estimate rock mass strength. It is a measure of strength gain with
confinement and can be determined from laboratory triaxial testing. This
parameter can be difficult to define, and is often estimated based upon empirical
values which have significant variance. The effect of the assumed mi value on the
simulated caving response has been studied in the Cave Demonstration Model by
modifying the failure envelope for the RM1 rock mass to reflect mi material
constant values of 5, 12 and 25. The resultant Hoek-Brown failure envelopes and
equivalent bi-linear Mohr-Coulomb strength properties are presented in Figure 51.
Figure 51. Hoek-Brown curves and equivalent bi-linear Mohr-Coulomb property

estimates for varying mi values.
The effect of estimated mi values on cave propagation behaviour in the

demonstration model has been simulated after 10 m draw. The results are
provided in Figure 52A.

Figure 52. Effect of estimates of mi on predicted cave propagation behaviour in the

numerical demonstration model.
It can be seen that an increase in the estimated mi reduces (marginally) the

propagation rate of a cave. This decrease can be attributed to the increased stress
at which failure must occur with increasing mi values. In each case, the bulked
cave mass profiles (Figure 52B) are similar. Increased seismic potential – based on
the empirically derived equation developed by Diederichs (2000) within the cave
back is predicted with an increased mi value (Figure 52D). This reflects the stress

caving mechanism required to be associated with the propagation of strong rock

masses.
4.3.4 Effect of Stress/Depth on Cave Propagation
During cave propagation, in situ stresses are redistributed around the evolving
cave mass as it propagates. The effect that the in situ stress magnitude has on the
evolving cave has been investigated by simulating mining at different depths. The
RM1 rock mass properties have been simulated. It can be seen from Figure 53A
and Figure 53D that, the cave propagation rate, and seismic potential increases
with increased stress/depth. In addition, as the stress magnitude increases, the
rate of the advance of the yield zone in front of the mobilised zone also increases
(Figure 53A). It can be seen that stresses increase at the extraction level, the cave
propagation rate increases. Significant increases are seen between the 600m –
650m simulations as the in situ stress at the extraction level approaches the rock
mass peak strength.

Figure 53. Cave propagation results for increasing stress /depth in the numerical
demonstration model. Significant increases are seen between the 600m –
650m simulations as the in situ stress at the extraction level approaches the
rock mass peak strength.

4.4 Summary
Through the development and application of the demonstration model, it can be

seen that the numerical approach developed by Pierce et al. (2006) allows cave
propagation to evolve as a function of the constitutive behaviour of the rock mass
and, induced stress conditions.
Based on these results of the numerical demonstration model it is clear that the
rate and shape of cave propagation through the in situ rock mass will be affected
by the geomechanical conditions at the site (rock mass behaviour and in situ
stress). However, as discussed previously (Section 3) limitations to this model
exist. These will be explored in the following chapters and developments to the
existing model will be implemented to address these limitations. The
demonstration model will be used to exhibit and validate responses for the
extended numerical model of cave propagation and subsidence assessment.

5 DEVELOPMENT OF THE UBIQUITOUS JOINT ROCK

MASS MODEL (UJRM)
At present, it is not practical to simulate large-scale mining/geological processes

using the SRM methodology due to the computational intensity of the numerical
technique. For this reason, continuum codes are required.
Continuum models of jointed rock masses are routinely used in rock mechanics;
Salamon (1968), Singh (1973), Chappell (1975), Gerrard (1982), Fossum (1985),
Cai and Hori (1993), Sitharam and Latha (2002), Zhu and Tang (2003), Yoshida
and Hori (2004), Samadhiya et al. (2004), Liang et al. (2004) and Nicieza et al.
(2006). Within these models the rock mass is considered as an isotopic continuum
with equivalent material properties, with the effect of joints accounted for
implicitly. Although the impact of joint frequency and persistence on strength is
considered by this approach, the joint orientation and its impact on strength
anisotropy is not.
The Subiquitous constitutive model in FLAC3D (Itasca, 2009) is routinely used to

represent laminated materials that exhibit non-linear material hardening or
softening. Clark (2006) used FLAC (Itasca, 2005) to demonstrate that the
assignment of ubiquitous joint orientations at the zone level (from a known joint-
orientation distribution) results in realistic rock mass behaviour and can yield
properties that are consistent with empirical techniques. Pierce et al. (2006)
calibrated stress-path dependent SRM results to a continuum constitutive
response and used the results in a successful analysis of cave behaviour at
Northparkes E26 Lift 2. Details regarding the continuum calibration are not
known.
Within the Subiquitous constitutive model, both matrix and joint properties are
specified, as illustrated in Figure 54. For each of the Mohr-Coulomb strength
properties, softening tables are defined. The softening tables provide the peak and
residual strength values along with the rate at which softening from the peak to
the residual value will occur. A full description and validation of the Subiquitous
constitutive model can be found in Itasca (2009).
5 – Development of the Ubiquitous Joint Rock Mass Model (UJRM)

Figure 54. Subiquitous constitutive model in FLAC3D; assignment of matrix and

joint properties.
An example of the damage evolution within a subiquitous sample can be seen

through the progressive degradation of matrix cohesion and ubiquitous joint-
failure plots at various stages of sample loading – illustrated in Figure 55.

Figure 55. Stages of damage within a simulated UCS test on a subiquitous sample.
Since both the rock block and joint conditions can be modelled with this
constitutive model, the calibration of anisotropic SRM sample responses can be
achieved. The development of a calibrated SRM rock mass response within FLAC3D
using the subiquitous constitutive model has been termed a Ubiquitous Joint Rock
Mass (UJRM) model. The SRM standard laboratory testing environment described
by Mas Ivars et al. (2008) has been used as the basis for the development of this
methodology.

5.1 Establishment of a Standard Laboratory

Environment
5.1.1 Sample Geometry and Generation
To match the SRM testing environment developed by Mas Ivars et al., (2011) a
rectangular shape has been used for the generation of each test sample.
This shape:
 matches the shape and volume of the SRM samples that have been tested;
 ensures ease associated with the scaling of critical strain values during the
subsequent modelling process;
 ensures ease associated with applying stresses in all three axial directions
(σxx:E-W, σyy:N-S, σzz:vertical) during the laboratory testing process; and
 minimises end effects.
Examples of sample geometry are provided in Figure 56.

Figure 56. Development of a UJRM sample (a) variation in sample size with equal
zone sizes; (b) joint assignment as a function of sample size.
Note that in Figure 56, the zone size in each sample remains constant. However,
the resolution decreases in the small scale samples - increasing the impact of the
joint assignment and ensuring sample scale bias is honoured. Figure 57 illustrates
the simulated axial loading conditions for each of the UCS, triaxial and direct
tension tests, and the three sample loading orientations that are completed for
each of the simulated loading conditions (UCS, triaxial and direct tension).

Figure 57. UJRM sample testing geometry (a) sample loading conditions (b)
orientation for anisotropy tests completed for each sample loading condition.
5.1.2 Sample Zone Resolution
A standard zone size is selected for the generation of each sample geometry. To
minimise the zone resolution dependency, described in detail in Section 2.2.2.3, the
UJRM material should be calibrated with the same zone size and shape to be used
in the large-scale model. This size will be dictated by the dimensions of the
undercut footprint to be modelled to ensure a reasonable zone resolution. In
practical terms, at least 8-10 zones are required across the shortest footprint

dimension (or undercut increment) to achieve the most accurate results.

Examples of poor and good mesh resolutions are provided in Figure 58.
Figure 58. Examples of (a) poor (b) low and (c) good mesh resolution required for
large-scale analysis of cave propagation.
5.1.3 Sample Loading Conditions
The loading and end conditions in analytical solutions, physical laboratory tests
and simulated numerical loading experiments can have a significant effect on the
test result. The problem is discussed by Brady and Brown (2006) and illustrated in
Figure 59.

Figure 59. Discontinuity shear strength in a triaxial cell (after Brady and Brown,
2006).
It is clear from Figure 59a, that if relative shear displacement of the two parts of
the sample is to occur, there must be lateral as well as axial relative translation.
Laboratory testing is often conducted with spherical seats (Figure 59b and Figure
59c) which can cause rotation of the sample during loading. An alternative end‐
condition involves the use of lubricated discs, as illustrated in Figure 20d. This
laboratory technique allows the lateral component of translation to freely occur,
however, this unrestrained end‐condition is not encountered in situ.
In reality, laboratory UCS tests conducted on intact samples are performed by

loading the sample between two steel platens. These platens provide a small
amount of confinement to the test specimen due to frictional resistance. In a
numerical sample loading conditions may be modelled by (a) allowing the sample
ends to move in all directions perpendicular to the loading direction, (b) fixing the
sample ends in all directions, or (c) modelling the steel platens above and below
the sample, and installing an interface between the two materials that is assigned a
stiffness and frictional resistance. The effect of each of these loading conditions on
the sample response is shown in Figure 60.

Figure 60. Simulation of different boundary loading conditions on the response of

UJRM material in the numerical UCS test environment.
To be consistent with the end conditions used throughout the SRM testing
conducted by Mas Ivars et al. (2008), fixed end conditions have been applied in the
UJRM testing environment on samples with a length: diameter ratio of 2:1.
However, it is important to note here, the varied strength response that is
generated based on the simulated end-conditions in the tests. By fixing the end
conditions in the model, artificial confinement conditions are simulated, resulting
in higher rock mass strength estimates that what is achieved in situ.
5.1.4 Large Strain/Small-Strain Calculation Mode
Some numerical modelling packages are able to calculate numerical solutions in

both a large-strain and small-strain mode. In small-strain mode, gridpoint
coordinates are not updated during mechanical calculations; in large-strain mode,
gridpoint coordinates are updated at each time step. The application of small-
strain mode is most useful when controlling boundary and applied conditions
when large displacements are expected in relation to the grid size. For large-scale

cave analyses, the small-strain calculation mode is required due to the large
displacements (>100 m) induced in the model. The effect of each of these
calculation modes on the behaviour of a UJRM sample has been investigated by
conducting a series of UCS tests on a material that has varying joint orientations.
The results are summarized in Figure 61.
Figure 61. Investigation of UJRM response as a result of small-strain/large-strain

calculation modes.
It can be seen that small-strain and large-strain calculation modes yield the same
results for those samples that have horizontal and random joints sets. However,
significant differences in the peak and residual strengths are apparent in the
sample that has a vertical joint set. The small-strain sample yields much higher
peak strength. The large-strain sample yields in tension along the joint surfaces
and, as expected, progresses to a fully degraded state with the continued
application of load. In order to allow for very large displacements, the UJRM rock
mass is calibrated using the small-strain calculation mode. Provided the same
calculation mode is used for the SRM calibration and the large-scale cave
propagation analysis, this should not cause any issues.

5.2 Calibration of UJRM Response
The following section outlines the methodology for the selection of the input
parameters for calibration of a UJRM to SRM responses. The methodology has
been based on four rock mass units detailed in Mas Ivars et al. (2008).
For the calibration of UJRM samples, it is assumed that all input properties can be
estimated based on measurable rock mass parameters. By modifying the input
strength parameters (defined in Figure 54), the calibration of deformation
modulus, unconfined compressive strength (UCS), tensile strength and the
softening behaviour of different sample sizes and in different loading directions
can be completed. In addition, SRM failure mechanisms are also honoured through
the monitoring of progressive matrix degradation, joint slip and joint dislocation
within the sample during failure.
5.2.1 Summary of SRM Responses
As discussed in Section 2.2.3, the SRM methodology has been developed to define
generalised stress-strain curves of a large-scale sample of a rock mass in three
opposing loading directions, and at a number of different scales. This ensures that
the material properties derived from the technique are not specific to one
particular stress path (as in the case of Pierce et al., 2006) and may be applied to a
number of different numerical modelling applications (i.e. cave analysis, slope
stability).
Commensurate with the development of the SRM standard suite of laboratory

tests, a testing environment used to calibrate the response of a subiquitous sample
(direct tensile test, uniaxial compressive strength test and triaxial test) has also
been developed.
5.2.1.1 Intact Calibration
The standard suite of laboratory tests, as discussed in Section 2.2.3 have been
carried out on four rock mass domains at the Palabora Mine in South Africa. A
detailed description on the testing program and results can be found in Mas Ivars
et al. (2008). The measured rock mass UCS strength used for the intact calibration
is provided in Table 4.

Table 4. Mean target intact rock block properties for the lithology at Palabora.
Foskorite Carbonatite Dolerite Pyroxenite
Mean measured UCS (MPa) 139 320 90 63

Estimated rock-block strength (MPa) 111 256 72 50
Mean modulus of Deformation (GPa) 58 90 72 78
Mean Poisson’s ratio, v 0.33 0.30 0.35 0.27
5.2.1.2 Discrete Fracture Network
A Discrete Fracture Network (DFN) for each of the rock mass domains was
developed by Mas Ivars et al. (2008) based on borehole and scanline data (from
both underground and open pit exposures). The DFN models encompassed
statistical distributions for the fracture persistence and orientation, coupled to a
density parameter. The DFN realisation is calibrated both to fracture frequencies
and orientation distributions observed. A summary of the mapping data is
compiled in Table 5.
Table 5. Measured joint frequencies and persistence from mapping at Palabora (after
Mas Ivars et al., 2008)
Lithology Mean joint frequency (min.–max.) Mean joint dia. (min.–max.)
Open Pit
Carbonatite 0.77m-1 (0.21–3.33) 15m (10–354.7)
Dolerite 2.26m-1 (0.35–16.00) 7.5m (5–658)
Pyroxenite 0.37m-1 (0.12–0.73) 15m (10–246)
Underground
Carbonatite West 0.83m-1 (0.16–3.33)
Carbonatite South 0.53m-1 (0.04–3.64)
Dolerite West 1.88m-1 (0.02–4.80)
Dolerite South 2.54m-1 (0.16–10.00)
Pyroxenite West 0.39m-1 (0.04–0.94)
Joint orientation data for each of the domains is provided in Figure 62.

Figure 62. Joint orientations considered in the development of the DFN for (a)
carbonatite (b) micaceous pyroxenite (c) dolerite (d) foskorite (after
Sainsbury et al., 2008).
Fractures are represented within the sample as ubiquitous joints. The assignment
of the joint dip, dip direction and radius is achieved via a random sampling
procedure from the DFN developed for the rock. The persistence of joints can be
honoured throughout the grid via extrapolating the joint dip and dip direction to
adjoining zones, which honours the fracture radius. To ensure complete
randomness in the model, a random list of all the zones is generated for the
importation of the fracture network and the presence of existing joints is honoured
(i.e., not overwritten) when importing the DFN.
An example of the ubiquitous joint orientations represented by this sampling

procedure is provided in Figure 63b. The actual joint orientation data is provided
in Figure 63a.

Figure 63. Representation of DFN in a UJRM sample (a) actual DFN (b) DFN
represented in numerical model (after Sainsbury et al., 2008).
5.2.1.3 Estimated Joint Strength
The joint properties were estimated from the roughness and hardness of joints
measured in the open pit at Palabora. The joint properties are listed in Table 6.
The friction and cohesion values were considered as direct inputs for the
ubiquitous joint strength.

Table 6. Estimated joint properties for the rock mass domains at Palabora (after Mas
Ivars et al., 2008).
Carbonatite Dolerite Pyroxenite Foskorite
Normal Stiffness (GPa/m) 150 250 150 150

Shear Stiffness (GPa/m) 20 30 20 20
Friction Angle (Degrees) 30 26 34 26
Cohesion (MPa) 0 0 0 0
5.2.1.4 SRM Simulation Results
Results of the SRM testing are provided in Table 7. The results show significant
strength anisotropy and scale effects in all testing directions.
Table 7. SRM-derived strengths for the rock mass domains at Palabora - triaxial 5-
MPa confinement (after Mas Ivars et al., 2008).
N-S E-W Vertical
Sample Width Sample Width Sample Width
40 m 20 m 10m 40 m 20m 10m 40m 20m 10m
Carbonatite (MPa) 65 60 69 38 42 61 45 38 46
Pyroxenite (MPa) 56 39 41 47 58
Foskorite (MPa) 38 38 36 39 35 33 27 28 29
Dolerite* (MPa) 66 70 78 68 79 104 61 62 100
* Dolerite samples were tested at 10 m, 5m and 2.5m sample sizes
5.2.2 Calibration of Deformation Modulus and Poisson’s ratio
The SRM testing considered an intact rock block Young’s modulus based on
laboratory testing of intact samples. When the joints were embedded within the
SRM sample, the simulated modulus decreased. In order to match the SRM-
derived deformation moduli for each domain, the deformation modulus in the
UJRM samples was reduced by 30% and 50% of the laboratory measured intact
Young’s Modulus. A summary of the calibrated modulus values is provided in
Section 5.2.2.
5.2.3 Calibration of Matrix Friction, Cohesion and Tension
Rock mass friction, cohesion and tension were selected based on a calibration of
the UJRM matrix material to the intact rock block strength (80% of laboratory UCS
strength) of each rock mass (Mas Ivars et al., 2008). A summary of the cohesion,
friction and tension values required to calibrate the intact material response for
each rock mass domain are provided in Table 8. Independent tension softening of

the matrix material based on plasticity state has also been assumed for the
calibration as discussed in Section 2.2.2.4.
5.2.4 Calibration of Ubiquitous Joint Properties
Joint cohesion was varied to achieve a match in peak strength between the SRM
and UJRM materials. Joint cohesions between 0.1% and 1% of the matrix cohesion
were required to calibrate the UJRM response. A summary of the calibrated joint
cohesion values for each lithology is provided in Table 8.
Joint friction angles were set for each lithology based on estimates determined for
SRM testing. For simplicity, it was assumed that joint friction did not soften. A
summary of the friction values used in each UJRM calibration is provided in Table
8. Joint tension was assumed to be zero. This is consistent with the description of
an open fracture.
5.2.5 Calibration of Critical Plastic Strain, 𝜖
Both matrix and joint critical strain values were varied to achieve calibration of the
UJRM sample. Matrix critical strains, (𝜖 ) between 0.01 and 0.15 were required
and joint critical strains (𝜖 ) less than 1% of the matrix values were required to
achieve calibration with SRM results. A summary of the calibrated critical strain
values is provided in Table 8.

Table 8. Calibrated UJRM properties for the rock mass domains at Palabora.
Carbonatite Foskorite Pyroxenite Dolerite
Matrix
Deformation Modulus (% Intact value) 50% 50% 30% 30%
Cohesion (MPa) 15 7 10 37
Tension (% of Cohesion) 40% 40% 36% 56%
Friction (Degrees) 40 35 49 47
𝜖 0.15 0.1 0.015 0.025
Joint
Cohesion (% of Matrix Cohesion) 1% 0.1% 0.2% 0.1%
Friction (Degrees) 30 30 34 26
𝜖 (% of 𝜖 ) <1% <1% <1% <1%
5.2.6 Calibrated Laboratory Stress-Strain Curves
Calibration of a UJRM for each of the four lithologies has been completed in three
different testing environments, in three loading directions, and at three different
scales. Results of simulated UCS and triaxial tests at 5 MPa confinement for the
Carbonatite domain are provided in Figure 64.
Figure 64. UJRM sample stress-strain responses (a) calibrated 40x80m carbonatite
UCS UJRM rock mass samples showing strength anisotropy (b) calibrated
40x80m carbonatite triaxial UJRM rock mass samples showing strength
anisotropy.
Calibrated UCS results from all the domains are provided in Figure 65.

Figure 65. Calibrated UJRM: SRM results at 5 MPa confinement for each lithology
at Palabora in three testing directions.
As shown in Figure 64 and Figure 65, UJRM samples are capable of reproducing the
anisotropy in peak strengths as well as the ductile stress-strain response observed
during testing of the equivalent size SRM samples. In addition, the average
deformation modulus has also been generally reproduced.
When the result of UCS tests at other scales are compared (Figure 64), it is evident
that the impact of scale on the average strength and on the variability in strength
(shown by the error bars) is stronger the smaller the sample size is.

Figure 66. UJRM UCS results for the carbonatite domain at Palabora compared to
SRM results at three different sample sizes in three loading directions.
Result of the testing show that, as sample size increases the rock mass strength
decreases and becomes less variable and that, in this case the calibrated numerical
UJRM/SRM responses are only valid for resolving failure mechanisms that have a
volume greater than 64,000m3 (40m x 40m x 40m).
It is interesting to note the UJRM sample responses in comparison to the

traditional isotropic Hoek-Brown strength estimates in Figure 66. In this case,
only two co-incident comparable values are obtained between the UJRM and
equivalent estimated Hoek-Brown strengths.

5.3 Application and Validation of the UJRM

Methodology
5.3.1 Calibrated SRM-UJRM in Laboratory Environment
In order to test and validate the methodology developed for the calibration of
UJRM-SRM samples, three additional UJRM materials have been calibrated using
the same procedure developed for the lithology at Palabora discussed in Section
5.2. Two of the units exhibit distinct foliation as a result of metamorphism. A
summary of the rock mass characteristics and SRM testing results are provided
below. Details of the SRM testing can be found in Sainsbury, Mas Ivars and Darcel
(2008).
5.3.1.1 Calibration of Intact Response
The values that have been used for the intact calibration have been determined
from laboratory testing. The presence of foliation in two units (Domain 1 and 3)
was considered to be an intact property since the spacing between foliation planes
is in the order of millimetres. This is due to the fact that in order to adequately
define a rock block within PFC, at least 5 particle diameters are required and thus,
representing the foliation in a real-life scale would produce a SRM sample that
would be impossible to test with our current computer efficiency.
As a result of this intact strength anisotropy, the intact strength calibration has
been completed in four directions (parallel, perpendicular and at angles of 30
degrees and 60 degrees to the foliation). Values for the intact and foliation failure
have been defined by the description of the failure mechanism in the laboratory
testing database. A summary is provided in Table 9.

Table 9. Summary of laboratory test results for three rock mass domains.
The intact rock calibrations have been completed on samples sizes of 2 m, 5.2 m
and 2 m height (2:1 height: width ratio) for the Domain 1, 2 and 3 rock masses
respectively. Figure 67 illustrates the calibrated UCS response for each of the rock
mass domains.

Figure 67. Calibrated stress-strain curves within PFC for three rock mass domains.

The calibrated PFC3D micro-properties required to simulate the strength and

deformation behavior of the intact rock for each of the domains are provided in
Table 10.
Table 10. Calibrated PFC micro-properties for three rock mass domains (after Sainsbury,
Mas Ivars and Darcel, 2008).
Domain 1 Domain 2 Domain 3
Minimum particle radius (m) 6.18e-2 1.37e-1 6.18e-2

Particle radius ratio* 1.66 1.66 1.66
Particle density (kg/m3) 4109 4109 4109
Particle E (Pa) 154e9 105e9 78e9
Particle friction 2.5 2.5 2.5
Particle k ratio 4.6 4.5 4.6
Parallel bond E (Pa) 154e9 105e9 78e9
Parallel bond k ratio 4.6 4.5 4.6
Parallel bond mean normal strength (Pa) 94e6 84e6 204e6
Parallel bond normal strength St. Dev. (Pa) 18.8e6 16.8e6 40.8e6
Parallel bond shear normal strength (Pa) 94e6 85e6 204e6
Parallel bond shear strength St. Dev. (Pa) 18.8e6 16.8e6 40.8e6
The properties of the foliation planes that have been used in the calibration of the
intact responses for the domains are summarised in Table 11.
Table 11. Calibrated intact foliation strength properties in PFC3D (after Sainsbury, Mas
Ivars and Darcel, 2008).
Kn (GPa/m) 10 n/a 10
Ks (GPa/m) 1 n/a 1
Friction (Deg.) 15 n/a 15
Cohesion (MPa) 0 n/a 0
Tension (MPa) 0.2 n/a 0.2
5.3.1.2 Selection of Joint Properties
The joint properties for each of the domains have been completed based on an
assessment of the roughness and planarity of the joints (roughness, waviness,
planarity, JRC, JCS, Js, Jw, Ja, aperture, infilling, weathering etc). A summary of the
estimated joint strength properties for each of the domains is provided in Table 12.

Table 12. Estimated open joint strength properties for simulation of joints in SRM
sample (after Sainsbury, Mas Ivars and Darcel, 2008).
Joint Kn (GPa/m) 150 150 150

Ks (GPa/m) 15 15 15
Friction (Deg.) 30 30 27
Cohesion (MPa) 0 0 0
Tension (MPa) 0 0 0
5.3.1.3 Development and Validation of a Discrete Fracture Network
For each of the rock mass domains, a DFN has been developed based on borehole
and scanline data. A complete description can be found in Sainsbury, Mas Ivars
and Darcel (2008). A summary is provided below.
A REV for Domain 1 has been estimated to be a cubic volume with a side length of
18 m. Scanline traces along theoretical boreholes in the SRM rock mass yield P 10
values of 0.55 - 0.59. This relates to an average fracture spacing of 1.7 - 1.8 m or
block volume of approximately 6 m3. A graphical representation of the Domain 1
DFN is provided in Figure 68.
Figure 68. Domain 1 fracture network views: 18m REV edge length (after
Sainsbury, Mas Ivars and Darcel, 2008).
A REV for Domain 2 has been calculated as a cubic volume with a size length of 40
m based on a review of the fracture spacing and persistence. Scanline traces along
theoretical boreholes in the SRM rock mass yield P10 values of 0.94. This relates to
an average fracture spacing of approximately 1 m – 1.42 m or block volume of

approximately 3 m3. A graphical representation of the Domain 2 DFN is provided

in Figure 69.
Figure 69. Domain 2 fracture network views: 40m edge length (after Sainsbury, Mas
Ivars and Darcel, 2008).
A REV for Domain 3 has been calculated as a cubic volume with a side length of 18
m. Scanline traces along theoretical boreholes in the SRM rock mass yield P 10
values of 0.77 - 0.6. This relates to an average fracture spacing of 1.7 - 1.3 m or
block volume of approximately 3 m3- 6 m3. A graphical representation of the DFN
is provided in Figure 70.
Figure 70. Domain 3 Fracture Network views : 18m edge length (after Sainsbury,
Cai et al. (2007) has previously developed a GSI chart that is based on quantitative
properties of the rock mass jointing. It is provided in Figure 71. When the field
estimates of joint strength properties and block sizes of Domain 1, 2 and 3 are used
to estimate GSI based on this chart, the results are consistent with the GSI values

developed from synthetic scanline mapping for each of the domains (Sainsbury,
Figure 71. Quantification of GSI chart (after Cai et al., 2007).
5.3.1.4 Calibrated Continuum Responses
The calibrated material input properties used for the UJRM sample are provided in
Table 13 along with the original dataset used to develop the technique for
comparison to provide validation of the developed methodology.

Table 13. Calibrated continuum material properties for seven rock mass domains.
1 2 3 Carbonatite Foskorite Pyroxenite Dolerite
Zone size Calibrated (m3) 1 8 1 1000 1000 1000 100
Matrix
Deformation Modulus (% Intact) 70% 70% 70% 50% 50% 30% 30%
Poisson Ratio 0.29 0.33 0.29 0.33 0.35 0.27 0.3
Cohesion (MPa) 19 12 48 15 7 10 37
Tension (% Cohesion) 48% 50% 50% 40% 40% 36% 56%
Friction (Degrees) 36 39 33 40 35 49 47
Dilation (Degrees) 6 6 6 10 10 10 10
𝜖 0.4 0.01 0.4 0.15 0.1 0.015 0.025
Joint
Cohesion (% of Matrix) 5% 30% 5% 1% 0.1% 0.2% 0.1%
Friction (Degrees) 30 30 27 30 30 34 26
𝜖 1.0e-06 1.0e-01 1.0e-02 1.5e-03 1.0e-03 1.5e-04 2.5e-04
The stress-strain results of the calibration for Domains 1, 2, 3 are provided in

Figure 72, Figure 73 and Figure 74.
Figure 72. Domain 1 SRM test results and UJRM response represented in FLAC3D
: 1 MPa triaxial tests (after Sainsbury, Mas Ivars and Darcel, 2008).

Figure 73. Domain 2 SRM test results and UJRM response represented in FLAC3D
: 1 MPa triaxial tests (after Sainsbury, Mas Ivars and Darcel, 2008).
Figure 74. Domain 3 SRM test results and UJRM response represented in
FLAC3D: 1 MPa triaxial tests (after Sainsbury, Mas Ivars and Darcel,
2008).

The peak strength and post-peak behaviour of the UJRM samples provide good
correlations with the SRM test results. In addition, it can be seen from Table 13
that the material input parameters are consistent across the entire suite of seven
calibrated SRM-UJRM samples – the four from Palabora developed in Section 5.2
and the three Domains (1,2 and 3) described here. As a result of this it can be said
that the methodology developed for deriving calibrating UJRM samples to SRM
responses has been validated for peak strength and post peak responses. The
simulation of more realistic deformation modulus values is described in Section
6.3.
5.3.2 UJRM Large-Scale Response
In order to demonstrate the importance of detailed consideration of the in situ

joint fabric in a cave propagation analyses and how a UJRM can capture a varied
response, four jointed rock masses have been simulated using the Subiquitous
constitutive model in FLAC3D. Each of the rock masses has the same commonly
defined rock mass properties (UCS, mi, GSI). However, in each of the rock masses,
the persistence and orientation of the joint fabric have been modified.
The jointed rock mass scenarios include:
 Random joints; i.e. isotropic rock mass.
 Horizontal joints.
 Vertical joints.
 Joints orientated at 45°.
Using the rock masses described above, production draw has been simulated in the
demonstration model (as described in Section 4). The cave propagation results are

Figure 75. Cave propagation behaviour for varying joint orientations simulated in the
Compared to the empirical approach for cave analysis, where the rock mass is
considered to be an isotropic material (Figure 75(a)), the consideration of jointing
(albeit extreme) allows significant variation (Figure 75(b)–(d)) in the cave shape
and rate of propagation to emerge as a result of production draw.
Based on the simplistic Cave Demonstration Models presented, the following

conclusions can be made:
 Joints that are orientated perpendicular to the direction of draw (i.e. in

most cases horizontal joints) are most favourable for cave propagation. The
mobilised zone advances vertically at the most rapid rate. In this case, the
rate at which the mobilised zone progresses far exceeds the production
draw rate.
 Joints that are orientated parallel to the direction of draw (i.e. in most cases
vertical joints) are not favourable for cave propagation. Minimal
displacement of the rock mass is achieved above the mining footprint. In

this scenario, shear failure of the rock bridges must occur to enable this
rock mass to yield and the cave to propagate.
 Joints that are orientated at an angle to the direction of draw result in a

preferred cave propagation direction. The principle stress direction can
either promote shear and tensile failure along the existing joints, causing
displacement of the rock mass beyond the lateral extents of the mining
footprint, or, cause clamping of the joints and result in hang-ups in the cave
back.
In this respect the consideration of joint orientations and rock mass fabric is
essential in determining cave propagation behaviour and the UJRM approach is
able to capture this variability.
5.4 Summary
Calibration of a UJRM assumes that the SRM testing is an accurate representation

of the rock mass strength and deformation behaviour in the simulated tested
loading directions and sample scales. Based on the calibration of SRM test results
of seven different rock mass domains, it has been shown that the UJRM method can
reproduce accurate failure mechanisms and strength anisotropy as well as the
expected scale effects shown by SRM testing. In addition, the implementation of a
UJRM approach in the Cave Demonstration Model highlights the potential impact
that a rock mass which exhibits significant strength anisotropy can have on cave
propagation behaviour.
By nature, the development of SRM responses is limited by the complex nature of

inputs required that include; a more thorough laboratory (intact) testing program
that is required to characterise variability in intact strength as a result of micro-
defects and or/veining as described by Pierce et al. (2009).
In addition, the development and validation of a characteristic DFN is difficult

since a more detailed level of fracture characterisation is required to develop a
robust model. The development of DFN’s are also limited to a few specialist
scientists in this area at the current time.

The time required to develop and implement SRM technology is generally greater
than the current time constraints put on geotechnical feasibility studies. However,
it is considered state-of-the-art and should be applied in instances where
significant strength anisotropy and/or scale effects are expected. In lieu of SRM
testing, the Hoek-Brown approach for strength estimation is considered a
reasonable starting point for cave propagation and subsidence assessment.

6 CONSIDERATION OF THE VOLUMETRIC CHANGES

THAT ACCOMPANY CAVE PROPAGATION
Due to the current computation constraints, the numerical model of cave

propagation must be implemented using a small-strain calculation mode (as
discussed in Section 5.1.4). As a result of this, modification of the density and
bulking/dilational behaviour of the rock mass during volumetric expansion
associated with cave propagation is required.
6.1 Rock Mass Density
Volumetric expansion of a rock mass occurs during caving as a result of intact rock
block fracture, separation and rotation during dilation under shear failure and or
expansion under tensile failure. As a result of this volumetric expansion the
density of the rock mass must decrease and porosity must increase. A typical
simulated porosity profile in a block cave during cave propagation is provided in
Figure 76.
Figure 76. Simulated porosity profile during propagation of a block cave.
Based on this profile, porosity tends to be highest in actively flowing regions and
lowest in areas that have been stagnant for some time and may have been subject
to re-compaction.
6 – Consideration of the Volumetric Changes the Accompany Cave Propagation

Within the caving model described by Lorig (2000), density of the rock mass
varied in proportion to plastic strain accumulation. This process is not physically
correct since mass is lost due to volumetric expansion (and mobilisation) and not
strength reduction. For example, a rock mass may have yielded due to an increase
in major principal stress but still be confined, and thus, no density changes should
have occurred.
In the Pierce et al. (2006) caving algorithm, the evolving rock mass bulked density
was related to increases in porosity through Equation [12].
⁄[ ( ⁄ )] [12]
Where is the rock mass bulked density (kg/m3), is the rock mass
undisturbed in situ density (kg/m3) and is porosity.
To prevent bulking of the rock mass to unrealistic levels within the model, a
maximum volumetric strain was set within the model that could not be exceeded
as discussed in Section 2.2.2.5. To implement this density reduction relation in the
numerical model of cave propagation, Equation [13] can used.
[ ⁄( )] [13]
Where is the volumetric strain increment as defined in FLAC3D.

6.2 Rock Mass Dilation
Based on research conducted by Hill (1950), it is known that plastic deformation of

a rock mass must be accompanied by an increase in volume. This phenomenon is
known as dilation and can be described by the sliding of micro-cracks (joints) or
particles (intact rock blocks) when subjected to shear strain. This mechanism is
represented in Figure 77. It is most commonly described by an angle,.
Figure 77. Conceptual diagram of dilation associated with sliding along micro-cracks
and particles (after Zhao and Cai, 2010).
Dilation is generally considered with an associated flow rule ( ) or a non-

associated flow rule ( ).
A typical stress-strain curve displaying the essential features of brittle rock

behaviour under triaxial compression is presented in Figure 78A based on
Rudnicki and Rice (1975).
Figure 78. Typical stress-strain curve for uniaxial compression of brittle, crystalline
rock (after Rudnicki and Rice, 1975).

The curve can be divided into four regions:
A. Slightly convex up portion that is characterised by the initial mobilisation

of rock mass strength through clamping of joints.
B. A nearly linear portion that is characterised by random joint shear failure
C. A non-linear region of decreasing slope characterised by minor intact

yielding and uniform joint shear failure; and
D. A maximum is reached the curve decreases representing intact yield

localisation and contiguous joint shear failure.
The corresponding stress/volumetric strain curve is presented in Figure 78B. The

stress – volumetric strain curve can also be divided into the four regions displayed
on the stress-axial strain curve in Figure 78A.
A. Non-linearity is due to the elastic closing of cracks,
B. Unloading of this region causes little hysteresis,
C. Initiation of dilatant volume increase and non-linearity due to micro-crack

growth and frictional sliding on micro-crack surfaces, and
D. The initiation of this region is less clearly defined, but is characterised by

accelerated micro-crack growth and rapid increase of dilatant volume
change leading to failure.
It is clear, based on these curves that dilation is not constant and it must be varied
as the rock mass yields and mobilises during cave propagation.
According to Detournay (1986), dilatancy depends on the porosity of a rock mass,

plastic strain (yielding) and the confining stress – all of which vary during cave
propagation. Relations with scale dependency are also expected (Sterpi, 1999).
Within the Lorig et al. (1995) caving algorithm, the dilation angle of the rock mass
was increased above the friction angle to induce fracturing parallel to the cave
back (i.e., perpendicular to 3) as the primary mode of failure. In this model, active

yielding was confined to the surface of the material since dilation causes build-up
of isotropic stress in the interior elements, thus causing failure.
Results of SRM testing conducted by Pierce et al. (2006) indicate that alternative
and/or additional fracture modes are likely within the cave back. As a result,
Pierce et al. (2006), simulated dilation as a constant value that was reduced to zero
when the maximum bulking potential was reached.
Based on the published information it appears that the constant dilation angle
assumed by most models is not realistic but should vary with rock mass damage
(decreasing GSI values) and confinement. This is confirmed by triaxial testing of
intact samples undertaken by Medhurst (1996) and Ribacchi (2000) and Zhao and
Cai (2010).
Table 14 presents a summary of the dilation angles calculated from large-scale

triaxial tests conducted on granular material by Marachi et al. (1972).
Table 14. Dilation angle in large-scale triaxial tests on rock fill material (after Marachi
et al., 1972)
Maximum
Particle Confining Dilation
Size Stress Angle
Source [mm] [MPa] [Degrees]
Oroville Dam 50 1 6.5
Oroville Dam 150 1 4
Oroville Dam 50 0.2 13
Oroville Dam 150 0.2 11
Crushed basalt 50 0.2 7.5
Crushed basalt 150 0.2 9
Pyramid Dam 50 0.2 7.5
Pyramid Dam 150 0.2 6.5
Vermeer and de Borst (1984) conclude that the dilation angle is at least 20o less
than the friction angle.
Other researchers have provided guidelines for the selection of a dilation angle.
Hoek and Brown (1997) suggest that dilation is greatest in competent rock masses,
and tends to zero as damage accumulated:

 GSI = 75 dilation angle is 25% of the friction angle of the rock mass ; 11o –
16o
 GSI = 50 dilation angle is 12.5% of friction angle of the rock mass; 6o – 8o
 GSI ≤ 30 dilation angle is zero.
Based on laboratory test results of Duncan-Farmer, (1993) Medhurst (1996) and

Ribacchi (2000); Alejano and Alonso (2005) developed a relationship for the
estimation of peak dilation angle based on confinement, friction angle and UCS. It
is presented in Equation [14].
[14]
Where is the peak dilation angle (o), ø is the angle of friction (degrees), is
the Intact Unconfined Compressive Strength (MPa) and is the Minor Principal
Stress magnitude (MPa).
It is also assumed, that dilation must tend to zero at zero confinement (after
Barton and Bandis, 1982) and once the maximum volumetric strain has been
reached (estimated by Equation [11]).
The implementation of the Alejano and Alonso (2005) relation is considered the
most appropriate for cave propagation and subsidence analysis since it is based on
the reinterpretation of previously published compressive test results for a range of
rocks. The relationship reflects dependencies on confining stress, plasticity and
indirectly on scale through UCS estimates.
The relation does not increase the number of parameters needed to model the
strain-softening rock mass and can be easily implemented with the existing
information in the subiquitous constitutive model.
The implementation of this equation on a conceptual rock mass with a UCS of 100
MPa and friction angle of 45 degrees is provided in Figure 79.

Figure 79. Evolution of peak dilation estimate on a rock mass during cave propagation
using the Alejano and Alonso relation.
Based on Figure 79, it is clear that in this relationship dilation decreases with
increasing confinement and peak dilation occurs at low confinement levels when
rock blocks are free to bulk and rotate. Dilation does not exceed friction angles at
its maximum and is close to the Hoek-Brown estimate of dilation at increased
confinement levels.
6.2.1 Implementation of Non-Constant Dilation in the Cave

Demonstration Model
The result of implementing the Alejano and Alonso (2005) dilation relation within
the Cave Demonstration Model – compared to a constant dilation angle of 20o is
presented in Figure 82. The impact on the evolving dilation, deformation modulus
and propagation rate is noted.

Figure 80. Implementation of a non-constant dilation relation and its impact on cave
propagation behaviour in the numerical demonstration model compared to
the simulation of a constant dilation angle.
Increased bulking (lower densities) immediately above the extraction level in the
demonstration model is seen with the implementation of the Alejano and Alonso
dilation relation. The height of the cave back will be controlled by the confinement
exerted by the bulked rock at the cave periphery.

6.3 Deformation Modulus
It is generally assumed that the modulus of deformation is equal in all directions

regardless of confinement and/or failure mode. This assumption has previously
been questioned by Fairhurst (1961), Adler (1970), Haimson and Tharp (1974),
Passaris (1977), Sundaram and Corrales (1980), Khan and Yuan (1988) and Chen
and Stimpson (1993).
A depth (confinement) dependent deformation modulus has previously been

described by Barton and Pandy (2011) in the context of open stope performance
and Hutchinson and Diederichs (1996) via observations of tunnel failure. In both
these cases, the greater the distance from an excavation face (or the greater the
confinement) the greater the estimated rock mass deformation modulus.
In the case of cave mining, the reverse can be considered. As the rock mass bulks
during the caving process, the point-to-point contacts that are created are
inherently softer than the face-to-face contacts when the rock mass is sitting in
situ.
The rate at which the deformation modulus decreases from an in situ state to a
fully bulked state in response to production draw has previously been simulated as
a linear decrease based on volumetric strain in the 2006 cave propagation model.
The methodology is based on the implementation of Equation [15].
( ) [15]
Where is the Bulked Deformation Modulus (GPa), is the in situ rock

mass deformation modulus (MPa), is the fully bulked rock mass
deformation modulus (GPa) and is an Expansion Factor, determined by
Equation [16].
⁄( ⁄ ) [16]
Where is the volumetric strain increment and is rock mass porosity.

A value of 250 MPa has previously been determined for a fully bulked deformation
modulus (Pierce et al., 2006). A schematic representation of the linear softening
relation (defined in Equation [15]) is provided in Figure 81.
Figure 81. Schematic linear relationship for rock mass deformation modulus reduction
based on Pierce et al. (2006) relation.
However, Hoek and Diederichs (2006) report that the deformation modulus is
non-linear in nature and can be related to GSI of the rock mass as shown in Figure
82.
Figure 82. In situ rock mass deformation modulus versus GSI for Disturbance Factors
of 0, 0.5 and 1.0 (after Hoek and Diederichs, 2006).

A compilation of laboratory test data completed on particulate matter is presented

in Figure 83. Here, the softened modulus (Ecave) has been presented as a fraction of
its initial value (Ein situ) and related to porosity. Porosity is considered in this case
since in situ density and deformation modulus will vary between rock mass
domains. It is assumed, for computational ease, that the in situ pre-mining state of
the rock mass represents a porosity of zero.
Figure 83. Softened deformation modulus versus porosity for particulate matter
determined by laboratory testing.
A second degree polynomial with a maximum porosity of 0.4 (as previously

determined by Pierce et al., 2006) is considered as the best-fit for this data and is

Figure 84. Best-fit deformation modulus softening equation to compiled laboratory test
data.
An equation to describe the non-linear modulus softening relation is presented in

Equation [17].
⁄ [17]
Where is the Bulked Deformation Modulus (GPa), is the in situ rock

mass deformation modulus (MPa) and η is the rock mass porosity (%).
To implement this in the numerical model of caving porosity can be related to

volumetric strain through Equations [18] and [19].
( ) [18]
Where initial pre-softened modulus and is the equivalent porosity
computed by Equation [19].
[19]

Where 𝜖 is the volumetric strain increment, 𝜖 is the maximum volumetric

strain achievable (%) defined by Equation [11] and maximum porosity able to
be achieved by a rock mass (%). Default values of 0.4 and 0.66 are recommended
for and respectively.
Using the equations of Hoek and Diederichs (2006) to estimate the initial in situ
deformation modulus, the non-linear softening curves (based on the non-linear
relation) of four different rock mass domains at increasing levels of porosity are
Figure 85. Typical deformation modulus softening curves of caving rock masses using
the non-linear softening relation.
Fully softened deformation modulus values are predicted that range between 200
MPa and 500 MPa which is consistent with those values previously reported by
Pierce et al. (2006) for a fully bulked rock mass.

6.3.1 Implementation of Non-Linear Deformation Modulus Softening

in the Cave Demonstration Model
The results of implementing the non-linear deformation modulus relation in the

Cave Demonstration Model are presented in Figure 86. Cave propagation
behavior is compared to the linear relation developed by Pierce et al. (2006). The
impact on the density, modulus and propagation rate has been presented.
Figure 86. Impact on cave propagation behaviour by implementing the non-linear

modulus softening relation in the cave demonstration model.
A greater reduction in the deformation modulus is seen with the implementation

of the non-linear relation. To further compare the modeling results, the evolving
bulk modulus has been tracked immediately in the cave back and at heights of 10
m and 20 m above the undercut. The results are presented in Figure 87.

Figure 87. Simulated evolution of the bulk modulus in the back of demonstration
model undercut; the linear and non-linear relations compared in the cave
demonstration model.
Significant variations in the simulated modulus values within each of the linear and
non-linear model relations are presented. The non-linear relation provides results
that soften the rock mass at a greater rate than the linear model. These simulated
differences between the two relations will affect the capacity of the caved rock
mass to carry stress/load and impact propagation rates.
This relation will provide a more rigorous estimate of the cave propagation rates
and allow more realistic bulking factors to develop within the cave.

7 IMPACT OF LARGE-SCALE DISCONTINTIES ON CAVE

PROPGATION AND SUBSIDENCE BEHAVIOUR
7.1 Subsidence Behaviour
Cave propagation behaviour and subsidence are closely linked geomechanical

processes. Mining-induced subsidence is the lowering of the ground surface
following the underground extraction of ore. To a greater or lesser extent, it results
from all forms of underground mining, but it is particularly pronounced in caving.
As the orebody caves and is extracted progressively, the overlying cap rock also
caves and moves downward, with the remaining ore producing a characteristic
surface depression (Brown, 2003).
Brady and Brown (1993) classify subsidence into two types, continuous and
discontinuous. Continuous (or trough) subsidence refers to the formation of a
smooth surface subsidence profile that does not have step changes as illustrated in
Figure 88a. This type of subsidence is the result of the extraction of a thin orebody
such as coal when a longwall mining method is used (Brady and Brown, 1993;
National Coal Board, 1975 and Peng, 1992). Discontinuous subsidence involves
large surface displacements and the formation of steps (or discontinuities) of the
surface as illustrated in Figure 88b. This type of subsidence may be associated
with several mining methods including sub-level, block and panel caving. In these
cases, the subsidence crater can be very large.
7 – Impact of Large-Scale Discontinuities on Cave Propagation Behavioiur

Figure 88. Conceptual models of subsidence a) continuous subsidence (after Kratzsch,

1983) b) discontinuous subsidence (after Whittaker and Reddish, 1989).
Subsidence associated with the extraction of coal seams has been studied in detail
since the late 19th Century. The mechanisms and associated terminologies used in
the analyses are now well understood and largely standardised. However, in
metalliferous mining and especially in massive orebodies, the mechanisms are not
as well established and the terminology used is not standardised. Consequently
there is still some debate over the terminology used and the key parameters used
to analyse or predict discontinuous subsidence. Numerous researchers have
proposed conceptual models of subsidence which have used diverse terminology
as presented in Table 15.

Table 15. Summary of terminology used to define discontinuous subsidence (after Flores
and Karzolovic, 2004).
Author (s) Key terminology

This model defined four zones and two angles:
Zones: Angles:
- a crater area - an angle of break
Hoek (1974)
- an unstable area - the dip of a subsequent failure plane
- a partially stable area
- a stable area
Brown and
Ferguson This model used the terminology defined by Hoek (1974).
(1979)
This model established three zones and two angles:
Zones: Angles:
Kvapil et al.
(1989) - devastation area (crater) - an angle of break
- transition area - an angle of sliding
- stable area
Karzulovic
This model used the same terminology defined by Brown and Ferguson’s model.
(1990)
This model defined three zones and two angles:
Zones: Angles:
Herdocia
- a crater area - an angle of caving
(1991)
- an unstable area - an angle of fracturing
- a stable area
This model defined four zones and four angles:
Zones: Angles:
Singh et al. - a crater area - an angle of caving
(1993) - an unstable area - an angle of unstable fracture
- a stable fractured area - an angle of fracture initiation
- a deformed area - an angle of deformation
Zones: Angles:
Lupo (1996) - a caved rock area - an angle of cracking
- a cracking area - an angle of deformation
- a deformation area
Zones: Angles:
Karzulovic et al.
- a crater area - an angle of break
(1999)
- a cracked zoned - an angle of influence
- a stable area
This model defined four zones and three angles:
Zones: Angles:
van As et al. - a crater area - an angle of break
(2003) - an fractured zone - a fracture initiation angle
- continuous subs. zone - an angle of subsidence
- a stable area
This model defined three zones and one angle:
Flores and Zones: Angles:
Karzulovic - a crater area - an angle of break
(2004b) - an influence zone
- a stable area

There is particular confusion regarding the “Angle of Break” term. Kvapil et al.
(1989), Karzulovic et al et al. (1999) and van As et al. (2003) define the Angle of
Break as the angle between the edge of the undercut and the limit of the crater
wall. However, Flores and Karzulovic (2004) propose the angle of break to be the
angle between the edge of the undercut and the limit of the discontinuous
deformation, or fractured zone.
van As et al. (2003) proposed the terminology presented in Figure 89 to

standardise the description of subsidence features related to block and panel
caving. This terminology has been adopted for the discussion of results in the cave
propagation and subsidence model developed herein.
Subsidence Zone of Influence
Large-Scale Small-Scale
Small-Scale Surface Displacement
Displacement Zone Cracking Zone
Stable Stable
(Continous Subsidence) Crater (Fractured (Continous
Zone Subsidence) Zone
Zone)
Tension Cracks
Caved (Broken) Material B

A
Local Geology Affects C Angle of Subsidence
Zone of Influence
Angle of Break / Cave Angle

(Angle of Draw = 90-C)
Fracture Initiation Angle
Figure 89. Terminology used to describe subsidence features for block- and panel-cave
mines (modified after van As et al., 2003).
Historically, mining engineers have defined the extent of subsidence features using
angles measured from the base of the undercut. However, extreme caution should
be used when using such quoted angles to predict subsidence, because factors such
as the mining depth and rock mass properties can have a significant impact on the
angles of break. The use of angles measured from the base of the undercut implies

that the failure mechanism develops along a plane, although the actual failure
surface may take any shape in situ.
7.2 General Characteristics of Caving Induced

Subsidence
A number of orebody, local geologic and topographic features can influence the
nature and extent of subsidence. Some of these factors include:
 the dip of the orebody
 the shape of the orebody in plan
 the depth of mining and the associated in-situ stress field
 the strengths of both the orebody and host rock mass
 the slope of the ground surface
 major geological features such as faults and dykes intersecting the orebody
and host rock
 prior surface mining
 the placement of fill in a pre-existing or newly produced crater, and
 nearby underground excavations.
In order to accurately predict subsidence behaviour, the numerical model of cave

propagation and subsidence assessment must include all of these factors.

7.3 Conceptual Models of Caving Induced Subsidence
7.3.1 Block Caving
A conceptual model describing the block caving subsidence process has previously
been described by Abel and Lee (1980) and is provided in Figure 90. It can be
described as follows:
 The collapse of rock progresses upward from the mining horizon as ore is
extracted (Figure 90A).
 The ground surface does not begin to measurably subside until caving has
so thinned the overlying cap rock so that it cannot transfer the load into the
adjacent solid cave walls. The overlying rock will thus begin to deflect
downward (Figure 90B).
 Further extraction of ore results in increased subsidence of the ground

surface above and adjacent to the area of extraction. The overlying rock is
progressively thinned by further propagation of the cave (Figure 90C).
 Continued extraction of ore will result in breaching of the ground surface.

The initial breach is typically in the form of a circular pit, commonly
referred to as a chimney-cave breakthrough. The breakthrough is roughly
centred over the area of greatest production draw, although offsets may
occur if geologic discontinuities are present (Figure 90D).

Figure 90. Conceptual model of the development of block caving subsidence (after
Sainsbury and Lorig, 2005).
If ore extraction continues, the surface breach will grow laterally near the surface.
The rock adjacent to the subsided crater either slides along geologic weaknesses,
such as joints or faults, or topples into the open crater.
7.3.2 Chimney Caving
Chimney caves are secondary draw-collapse structures that may develop over the
mined area. Chimney caves form when the flow channel of the drawpoint(s)
reaches the surface. At this point, the caved ore-flow ellipse changes into
cylindrical flow (Kvapil, 1982), as illustrated in Figure 91.

Figure 91. Conceptual model of chimney cave development (Betourney et al., 1994), b)
surface expression of a chimney pipe in a kimberlite caving operation (after
van As et al., 2003).
van As et al. (2003) suggest that chimney caves are usually the result of poor cave
management, in that, excessive draw from an isolated drawpoint is allowed to
occur that causes these features.
7.3.3 Plug Caving
Plug caving (or plug subsidence) is a form of chimney caving that occurs suddenly
rather than progressively and is controlled by one or more major structural
features which provide low strength surfaces on which the plug of undercut rock
may slide under the influence of gravity. In this case, the rock will undergo
essentially rigid body displacement without breaking up or dilating if the vertical
distance through which it falls is restricted (Brown, 2003). Figure 92 illustrates the
observed plug subsidence controlled by intrusive dykes at the Athens Mine, in
Michigan, USA.

Figure 92. Plug subsidence mechanism at the Athens Mine in Michigan USA (after
Obert and Duvall, 1967).
Plug caving resulted in a fatal air-blast at the Northparkes E26 Lift 1 Mine in 1999
as discussed in Section 1.2. The plug cave resulted in the formation of a circular
subsidence crater at the ground surface as illustrated in Figure 93.
Figure 93. Geometry of Lift 1 cave a) before and b) after plug caving (after Pierce,
1999).

7.4 Subsidence Features Related to Cave Mines
The time associated with subsidence resulting from mining is composed of two
distinct phases: (1) active and (2) residual. Active subsidence refers to all
movements occurring simultaneously with the mining operations, while residual
subsidence is that part of the surface deformation that occurs after the cessation of
mining. The duration of residual subsidence is of particular importance from the
standpoint of evaluating the extent of liability of underground mine operators and
developers for post-mining subsidence and land use.
7.4.1 Description of Active Subsidence Features
A review of large-scale surface disturbances from block and panel caving mines
was conducted by Lupo (1998), who found that the primary surface features that
develop as a result of block and panel caving include the following zones:
• a caved rock zone
 zone of large-scale fracturing;
 a small-scale surface displacement (continuous surface subsidence) zone,

and
 a stable (elastic) zone.
The following section provides a discussion on their distinct features.
7.4.1.1 Caved Rock Zone
The caved rock zone is a common surface feature of many caving mines; it also is
referred to as the zone of active cave movement (van As et al., 2003) or the crater.
Caved material consists of irregular blocks of mobilised rock, ranging in size from
millimetres to several metres. The caved rock zone develops as the underground
caving influence reaches the ground surface, causing the overlying rock mass
and/or side rock to fall into the mined void. Over time, the surface of the caved-
rock zone may subside as ore is continued to be withdrawn (Lupo, 1998). Figure
94 illustrates the formation of a crater in steep terrain at the Henderson Mine in
Colorado.

Figure 94. Photo showing crater and caved rock zone at Henderson Mine (after Lupo,
1998).
7.4.1.2 Zone of Large-Scale Fracturing
The zone of large-scale fracturing consists of an area in which the ground surface is
broken and has large open tension cracks, benches, and rotational blocks. The
primary failure mechanism of surface cracks associated with cave mines is shear
and tensile failure of the side rock, which results in stepped benches and scarps.
Other types of failure mechanisms, such as toppling and block rotation, are also
present, but they appear to be secondary mechanisms that form after the primary
shear failure develops. Figure 95 illustrates the typical scarp and cracking features
observed at Northparkes E26 Lift 1 Mine, in NSW, Australia.

Figure 95. Photo showing large-scale surface cracking at Northparkes E26 Lift 1
Mine (after van As et al, 2003).
Sainsbury et al. (2010) report that a total strain criterion of 0.005 (0.5%) can be
used to assess the limits of the large-scale fracturing at the Grace Mine. This total
strain criterion has also been used to calibrate the limit of large-scale fracturing at
the El Teniente block cave mine in Chile (Cavieres et al., 2003).
7.4.1.3 Small-Scale Displacement Zone (Continuous Zone of Subsidence)
Continuous surface subsidence, as defined by Brauner (1973), is the response of

the rock mass to a mined void, which results in the formation of a gentle surface
depression. Generally, the continuous subsidence zone forms between the large-
scale fractured zone and the undisturbed surface.
Surface buildings, roads, underground power lines, railroads and other structures
can be impacted by continuous surface subsidence. Lupo (1998) reports measured
subsidence up to 200 mm in a continuous subsidence zone at a distance of 250 m
from a large-scale fractured zone that caused heavy damage to nearby surface
structures.

Sainsbury et al. (2010) report that the limit of measured small displacements at
the abandoned Grace Mine can be derived by generating a contour line that
encompasses all areas of horizontal strain > 0.002 (0.2%) and angular distortion >
0.003 (0.3%). These strain criteria are based on the surface subsidence required to
cause damage to a masonry structure during active subsidence (Singh, 2003).
Figure 96 illustrates a small tension crack within the small-scale displacements
zone at the Kiirunavaara Mine.
Figure 96. Photo showing tension crack within small-scale displacements at the
Kiirunavaara Mine (after Villegas, 2008).
7.4.1.4 Stable (Elastic) Zone
The area outside the small-scale displacement zone is termed the stable zone; it
usually is defined as the area in which mining-induced surface displacements are
insufficient to cause any architectural damage.

7.4.2 Long-Term Time-Dependent Subsidence
7.4.2.1 Residual Subsidence
Almost all of the limited research conducted on residual mining-induced

subsidence is associated with underground coal-mining methods. The time span
during which surface subsidence occurs varies considerably with the mining
method used. Longwall coal mines generally induce subsidence beginning almost
immediately after the commencement of mining. With room and pillar systems,
major occurrences of surface subsidence may be delayed for decades until the
support pillars deteriorate and collapse.
The duration of reported residual subsidence movements above longwall coal

mines is relatively short, typically varying between a few weeks and about 5 years.
Singh (2003) reports that the magnitude of these movements rarely exceeds about
10% of the total subsidence. The time spans reported in the literature are
summarised in Table 16.
Table 16. Observed residual subsidence duration over longwall mines (after Singh, 2003).
Reference Country Residual Subsidence Duration
Institution of Municipal Engineers UK 2 - 10 years

(Anon., 1947)
Orchard and Allen (1974) UK 3 - 6 years (strong overburden)
Collins (1977) UK 2 - 4.5 years
Grard (1969) France 6 - 12 months
Brauner (1973) Germany 1 - 2 years
Brauner (1973) USSR 2 years (shallow mines)
4 - 5 years (deep mines)
Shadrin and Zomotin (1977) USSR 0.2 - 2 years
Gray et al. (1977) US 0.3 - 3 years
Hood et al. (1981) US 1 year
Luo and Peng (2000) suggest that the main cause of residual subsidence for
longwall coal-mining operations is the compaction in the overburden strata that
was disturbed during the active subsidence process. This mechanism is captured
in the cave propagation model via the non-linear deformation modulus function
that is related to volumetric strain.

The only observation of residual subsidence at a panel-caving mine noted in the

literature is continued crater expansion at Henderson Mine between 1982 and
1983. Stewart et al. (1984) postulate that the continued crater expansion was
caused by compaction of the caved material.
7.4.2.2 Sub-Surface Erosion
Several years, or even decades, after mining-induced subsidence has stabilised, pot
holes have been observed on the surface, mainly around the perimeter of the
subsided area where tension cracks existed at the time of subsidence at the
abandoned Grace Mine (Sainsbury and Lorig, 2005).
Coincident with subsidence, deep-seated tension cracks develop in the rock and
the overlying soils on the surface. van der Merwe (1999) suggests that, due to
surface erosion, the cracks within the soil are filled with loose soil and become
healed, but the deep-seated cracks within the bedrock remain open and become
natural conduits for percolating surface water. As water percolates through the
soil into the cracks, the surface soils are eroded and form a cavity that eventually
breaks through to the surface. A conceptual model of a sub-surface erosion
mechanism is presented in Figure 97.
Figure 97. Simplified subsurface erosion mechanism (after Van der Merwe 1999).

In reality the surface expression of these mechanisms look like pot holes as shown
in Figure 98.
Figure 98. Photos of subsurface erosion pot holes (after Van der Merwe, 1999).
A similar mechanism has been observed at the abandoned Grace Mine panel cave,
whereby a sinkhole was observed outside the limit of large-scale surface many
years after the cessation of mining, as illustrated in Figure 99.
Figure 99. Photo of sinkhole located outside the limit of large-scale cracking at the
abandoned Grace Mine (after Sainsbury and Lorig, 2005).

The time of development of sub-surface erosion cannot be predicted reliably.

However, van der Merwe (1999) reports that it is conceivable that the mechanism
may take decades to centuries to fully develop. As a result of this time dependency,
the ability to capture such a process in a numerical model is difficult at the current
time. As a result of this, the numerical model of cave propagation and subsidence
assessment with focus on predicting the active subsidence limits only.
7.5 Effect of Large-Scale Discontinuities on Subsidence

Limits
The influence of large-scale discontinuities on cave propagation and subsidence is

recognised to be important by many researchers (Crane 1929, Heslop 1974,
Boyum 1961, Fletcher 1960, Parker 1978, Laubscher 1990, Mahtab 1976, Hoek
1974, Holla and Buizen 1990, Shadbolt 1978, Shadbolt 1987 and Hellewell 1988).
In situ observations by Crane (1929), Parker (1978), van As et al. (2003), Blodgett
(2002) and Hatheway (1968) have shown that the impact of discontinuities can be
varied based on persistence, strength and orientation relative to the undercut
footprint and principal stress direction. Quantifying the effects of large-scale
discontinuities on cave propagation is complicated by the fact that many features
have not reacted adversely when subjected to subsidence and the results of
scientific investigations are in some instances, contradictory.
Laubscher (2000) suggests that major discontinuities must have sufficient

continuity to influence the caveability of the rock mass. Butcher (2005), after
Stacey and Swart (2001) reports that major discontinuities can modify or enlarge a
crater perimeter by further break back as illustrated in Figure 100.

Figure 100. Schematic diagram of how crater shape can be modified by major geological
structure (after Stacey and Swart, 2001).
Based on Figure 100, the presence of a steeply dipping fault can terminate the
angle of draw short of its normal value. Whereas, if a gently dipping fault intersects
the collapsing rock column the lateral extent of surface subsidence can increase
outward to the place where the fault intersects the ground surface. Abel and Lee
(1980) report that whether or not this takes place depends primarily on the shear
strength of the fault zone.

Many observations of the influence of discontinuities have been made; however,

only a modest amount of research work has been carried out to qualify and
quantify their influence.
Crane (1929) carried out extensive measurements of caving at iron ore mines in
Michigan, USA and developed a system for predicting angle of draw based on joint
measurements. Crane’s observations led him to conclude that rock breaks
according to a systematic arrangement of planes of weakness (joints) with slight
irregularities due to breaking between joints, and, in the absence of faults and
dykes, joint dip determines the angle of break. This is consistent with the results
of the Cave Demonstration Models presented in Figure 75.
Parker (1978) also notes that geological structure is a major controlling factor in
subsidence. In weak rocks there may be no significant geologic structure, hence the
reported cave angles are usually consistent and can be predicted with reasonable
confidence. In stronger rocks however, the cave angle is usually controlled by
geological structures. A well-defined fault plane, which is parallel to a mining face
and steep to moderately inclined, will result in a cave which propagates to surface
fairly rapidly and is defined on the surface by the trace of the fault plane. If the
predominant joints and faults are roughly perpendicular to the mining front,
caving may be inhibited and negative cave angles (overhangs) may occur.
van As et al. (2003) report that in most cases when a mining face encounters a
significant discontinuity with a moderate to steep dip, movement will occur on the
fault regardless of the cave angle. A stepped crack will result where the fault
daylights at the surface. If mining is only on the hangingwall side of the fault there
will only be surface movements on the one side. If the fault dip is steeper than the
cave angle, the extent of surface subsidence will be reduced, conversely, if the fault
dip is less than the cave angle the extent of surface subsidence will be increased.

7.5.1 Fault Impacted Caving
The following section provides documented case histories of when large-scale

discontinuities have impacted cave growth. These documented case studies
provide the basis for validating an appropriate numerical modelling technique for
representing cave propagation behaviour in the numerical model of cave
7.5.1.1 San Manuel Mine
Faults at the San Manuel Mine are important factors in causing and forming
boundaries to surface subsidence (Blodgett, 2002). Hatheway (1966) reported that
vertical cave propagation was halted and then deflected by the shallow dipping San
Manuel fault, as illustrated in Figure 101.
Figure 101. Conceptual development of surface subsidence at the San Manuel Mine
(after Hatheway, 1966).

In addition, steeply dipping north-west trending faults at the San Manuel Mine
were observed to exert a major influence on the orientation of tension cracks and
development of the subsidence crater. The Cholla fault arrested the development
of the subsidence crater in a north-easterly direction for several years as
illustrated in Figure 102.
Figure 102. Plan view, section view of subsidence crater at the San Manuel Mine (after
Hatheway 1966).
Based on this case study, it can be seen that sub-vertical faults limit the subsidence
crater beyond their extent and faster propagation rates can be expected as the
mobilised rock mass propagates unravelling along them.

7.5.1.2 Ridgeway Deeps Sub-Level Cave
At the Ridgeway Mine located in NSW, Australia, the vertically orientated North
Fault significantly altered the propagation rate of the sub-level cave and rapid cave
propagation rate was caused by the weak fault strength, as illustrated in Figure
103.
Figure 103. Photos showing cave propagation controlled by weak vertical fault at the
Ridgeway Mine (Brunton, 2009).
Rapid cave propagation occurred along this fault and surface break-through
occurred much sooner than was anticipated. The cave back advanced at three-
times its previous rate, unravelling along this structure (Brunton, 2009).

7.5.1.3 Questa Mine
The crater at the Questa Mine in New Mexico, USA, has been affected by sliding
along large scale structures (Gilbride et al., 2005). Slide features include
escarpments, fresh cracks, block toppling, surface rubblisation, tree tilting, and
disturbance to hillside vegetation. While no sub-surface measurements of
movement exist, the gross surface expression of the east wall slide suggests that
the slide is relatively shallow-seated (<60 m deep) and is occurring along a planar
or near-planar surface. Sliding originally occurred along a high-angle, southeast-
dipping fault in mid-1997, forming a large head scarp. The head scarp currently
measures more than 60 m in height. The slide and head scarp are shown in Figure
104.
Figure 104. Photo of Goathill Crater at the Questa Mine (after Gilbride et al., 2005).

7.5.1.4 Henderson Mine
Geological contacts have also been reported throughout the literature to effect the
formation of surface subsidence features. Crane (1929) reports that igneous
intrusions such as dykes and sills within the iron ore deposits of the Michigan
Upper Peninsula have a significant effect on cave propagation.
Carlson and Golden Jr. (2008) report that a weak intrusive contact at the edge of
the 7210 Production Level at the Henderson Mine was observed to cause a low
angle of break during initiation of the 7210 cave as illustrated Figure 105.
Figure 105. Irregular cave growth along a weak intrusive contact at the Henderson
Mine 7210 Level (after Sainbury et al., 2011)
Further investigation of this phenomena is considered in Section 12.

7.5.1.5 Kimberly Mine
Laubscher (2000) has previously reported on structurally controlled cave shapes

at the De Beers Kimberly Mine. In this case, the structure has impacted the cave
shape at depth, and overhangs that result in negative break angles can form in
structurally unfavourable areas or along geological contacts. Figure 106 shows a
section through the De Beers Kimberley Mine that shows an overhang against the
contact between the host rock and kimberlite pipe.
Figure 106. Section through Kimberly Mine showing over-hang (after Laubscher,
2000).
7.5.1.6 Summary
Based on these case studies, it is clear that in order for a geologic feature to be
considered significant (i.e. to influence the cave angle) two conditions must be met:
 movement on the feature must be kinematically feasible; and
 the forces driving movement must be greater than the forces resisting
movement. These forces are dependent upon many factors including: dip of
the feature, cohesive strength of infilling material, roughness and planarity
of feature, water pressure in feature and stress.

Parametric studies have previously been conducted by Vyazmensky et al. (2010) to

study the effect of joint fabric and large-scale discontinuities on block caving
induced surface subsidence. Figure 107 illustrates the predicted crater formation
for a single fault that intersect the base of a 200 m deep conceptual block cave at a
number of different locations.
Figure 107. Simulation of subsidence crater formation for different two-dimensional fault
orientations (modified after Vyazmensky et al., 2010).
It is clear from these simulations that the greatest impact on caving is seen when a
sub-vertical fault forms the boundary of mining limits. In this case rapid cave
growth can be expected along its extent. However, this study does not discuss the
strength of the discontinuities used in the simulation which are known to affect
caving behaviour.

7.6 Fault Properties
One of the first problems associated with analysing the effects of geologic structure
on ground movements is characterising the properties of the discontinuities. These
properties are highly variable and include: orientation, infill, previous
displacement, planarity and shear strength.
There is limited information available regarding the large-scale shear strength of

fault structures. Figure 108 illustrates the results of direct shear tests carried out
to determine the peak friction angle and cohesion of filled discontinuities as
reported by Wyllie and Mah (2007).
Figure 108. Estimated shear strength of filled discontinuities (after Wyllie and Mah,
2007).
It can be seen that the range of fault properties derived from numerical back
analyses is highly variable, and, at the current time, there is no real way to
determine these properties from large-scale in situ testing. As a result of this,
sensitivity studies are required to determine the range of caving and subsidence
behaviour expected when large-scale discontinuities are present within and
around a propagating cave.

7.7 Numerical Simulation of Large-Scale Discontinuities

in the Cave Demonstration Model
There are various numerical techniques available to simulate geological structures

within a three-dimensional numerical grid. The most common approaches for the
representation of faults within a numerical mesh include an implicit and explicit
technique. An example of each approach is provided in Figure 109.
Figure 109. Conceptual mesh of implicit versus explicit technique for fault
representation.
An investigation into the performance of each of these approaches has been

conducted using the Cave Demonstration Model detailed and discussed in Section
4.
In order to study the effects of geological structures on cave propagation

behaviour and compare results from the two techniques, a series of idealised fault
structures were developed. Figure 110 illustrates the orientation of five test fault
structures.

Figure 110. Conceptual geological structures simulated in numerical demonstration

model.
As discussed, whether or not a discontinuity affects cave propagation and

subsidence depends primarily on its shear strength. Based upon the range of shear
strength parameters, provided in Figure 108, three property sets have been
defined for analysis as presented in Table 17.
Table 17. Conceptual fault shear strength and stiffness parameters represented in
Coh.  Tens. Kn Ks
(kPa) (Deg.) (kPa) (GPa) (GPa)
Strength 1 0 20 0 1 0.1
Strength 2 75 30 0 10 1
Strength 3 200 40 0 50 5
7.7.1 Implicit Fault Representation
Implementation of an implicit scheme for fault representation can be considered

by the same methodology as joints within a UJRM sample (discussed in Section 5).
Figure 111 illustrates the results of a simulated direct shear test conducted on a
rock mass with a horizontal discontinuity embedded along a group of zones
(coloured red) at the block contacts. The fault has been simulated with a
ubiquitous joint cohesion of zero and friction angle of 30 degrees. Matrix
properties are consistent with a rock mass of UCS 100, GSI 60 and mi of 25. A
normal stress of 30 MPa has been applied to the sample during test simulation.

Figure 111. Simulated direct shear test; normal stress 10 MPa using ubiquitous joints
in FLAC3D.
The same ubiquitous joint technique has been used to simulate the presence of
fault structures within a large-scale cave model, as illustrated in Figure 112.
Figure 112. Ubiquitous joint faults used to simulate faults within a cave-scale model.

It can be seen that a significant disadvantage with this approach is the need to
specify relatively thick fault zones that extend across at least two zones in the
model to ensure strain localisation is possible in these elements (as discussed in
Section 2.2.2.3). Depending on the size of the zones this leads to fault thickness of
20 – 40 m within a large-scale model which is not realistic.
Figure 113 illustrates the predicted mobilised zone on a cross-section through the
conceptual cave models using the implicit fault scheme. The rock mass properties
have only been modified in the zones that the fault passes through.

Figure 113. Cross-section of mobilised zone (2m displacement) – implicit, ubiquitous

joint approach used to simulate conceptual discontinuity surfaces.

A minor increase in vertical cave propagation is associated with Fault A and the
weakest fault property set. However, all of the other fault orientations and fault
properties show only a very minor effect on the predicted mobilised zone. Little
impact on the cave growth rate and shape is noted with this approach, unless a low
strength fault is orientated sub-vertically at the cave periphery. These results are
not consistent with the observations from previous documented case studies
presented in Section 0 and therefore the implicit approach is not considered
adequate for the representation of faults in the numerical model of cave
7.7.2 Explicit Fault Representation
Most modelling codes have the capability to simulate interfaces that are
characterised by Coulomb sliding and/or tensile separation. Interfaces have the
properties of friction, cohesion, and dilation, normal and shear stiffness, as
Figure 114. Schematic diagram showing interface logic and how it can be used to
represent a discontinuity in a numerical model of caving.

Although there is no restriction on the number of interfaces or the complexity of

their intersections, historically, interfaces have not been used to simulate
geological structures due to difficulties in creating complicated geometries.
Advances in mesh generation through the programs such as Kubrix (Simulation

Works, 2012) and more specifically the tetra-split meshing option have made it
possible to create models with many irregular and intersecting interfaces, as
illustrated in Figure 114b. In addition, a new methodology developed within
FLAC3D allows the specification of peak and residual strength properties along the
interface.
Figure 115 illustrates the predicted mobilised zone when interfaces are used to
simulate faults within the Cave Demonstration Model. The model results display
much greater influence of the fault structures than what was predicted with the
implicit (ubiquitous joint) approach. Significant over break is predicted with Fault
B and the weakest fault properties, whilst the effect of increasing fault strength
and stiffness is clearly observed.

Figure 115. Cross-section of mobilised zone (2m displacement) – explicit, interface

approach used to simulate conceptual discontinuity surfaces.

The simulation of explicit faults in a numerical model of caving provides results

that indicate faulting has a significant impact on cave shape growth. The numerical
simulation results provide cave shapes that are intuitive and match with those
observations made at various mine sites around the world documented in Section
7.5.1.
7.8 Summary
Based on the simulation results provided for both an implicit and explicit
numerical modelling technique, it is recommended that the explicit fault approach
be adopted within the numerical model of cave propagation and subsidence
assessment to ensure the most rigorous assessment of fault behaviour is achieved.
Most of the available information on the effect of geological structures on cave

propagation behaviour throughout the literature is qualitative in nature. In order
to validate the modelling methodology with the use of interface elements to
simulate large-scale geological structures, the observed and monitored structurally
controlled cave propagation behaviour at the 7210 Level of the Henderson Mine
has been studied. A summary of the modelling results are provided in Section 12.

8 DEVELOPMENT OF A PRODUCTION DRAW

ALGORITHM
8.1 Influence of Production Schedule on Cave

Propagation Behaviour
The impact of the production draw schedule on cave propagation behaviour has
previously been documented by Laubscher (1994). Previous experience at the
Northparkes Mines suggests that the success of a mine plan in hard, jointed rock
masses will rely on the ability of the rock mass to cave at a rate greater than the
production draw rate to ensure continuous cave propagation without creation of
an air-gap. The rate of caving can be slowed by controlling the draw as the cave
can only propagate if there is space into which the rock can move. The rate of
caving can be increased by advancing the undercut more rapidly but problems can
arise if this allows an air gap to form over a large area (Laubscher, 2000).
From simulations conducted in the Cave Demonstration Model thus far, it is clear
that the cave propagation rate can vary significantly between orebodies, working
panels and adjacent drawpoints. As a result of this, the realistic and accurate
representation of a mining schedule is essential in being able to accurately assess
the bulking/dilation behaviour and cave propagation rate. As seen previously in
Table 1, the production methodology can have a significant effect on the cave
propagation behaviour as a result of the induced bulking behaviour internal to the
cave mass.
The magnitude and orientation of the regional stress on and around the mining
footprint also plays a significant role in caving (Laubscher, 2000). Caveability is
promoted when the major principal stress direction is perpendicular to the short-
axis of a cave footprint as discussed in Section 1.3.
The redistribution of stresses around a propagating cave causes the cave volume to
evolve into the most stable shape. In most cases this is circular or elliptical. This
phenomenon was observed at the abandoned Grace Mine as illustrated in Figure
116.
8 – Development of a Production Draw Algorithm

Figure 116. Plan view of subsidence limits at the Grace Mine determined by
observations.
As a result of this, an accurate representation of the evolving cave shape and draw
rate is required to ensure that stress redistribution around, above and below the
yielded rock mass is an accurate representation of the evolving in situ conditions.
8.1.1 Impact of Production Draw Strategy in the Demonstration

Model
When planning a cave mine, there are a number of production controls that can be
implemented to ensure optimum cave performance. The effect that a production
schedule has on cave performance has been considered within the Cave
Demonstration Model by simulating a) uniform draw block caving, b) variable
draw block caving and c) incremental draw (panel caving). For each of these
scenarios an average height of draw of 10 m has been simulated across the entire
undercut footprint. The simulation results are provided in Figure 117.

Figure 117. Effect of draw strategy on the caveability of a rock mass in the numerical
demonstration model.
By simulating perfectly uniform draw (zero variability in drawbell production), an

increase in the propagation rate is seen in Figure 117A since; shear stresses within
the caved mass will be decreased, resulting in a reduction of the bulking behaviour
of the rock mass. The effect that a uniform draw strategy has on a cave
propagation rate has been documented by Pierce et al. (2006) during a numerical
back analysis of the Northparkes E26 Lift 2 caving behaviour. In addition, by
staggering production and simulating an incremental draw strategy (where mining

areas are progressively brought online such as in panel caving) the resulting cave
propagation rate is decreased as this induces greater shear stresses within the
caved mass, resulting in increased rock mass bulking.
8.2 Influence of Rock Mass Bulking Behaviour on Cave

Propagation Behaviour
When caving occurs, due to the translation and rotation of rock blocks, the rock
mass bulks/dilates and occupies a greater volume than its in situ state. The term
‘Bulking Factor’ (BF) defines the ratio of the volume of broken rock, to solid
rock, . It can be calculated based on Equation [20].
⁄ [20]
Where BF is the Bulking Factor (%), is the volume of broken rock (m3) and is
the volume of solid rock (m3).
Highly laminated rock masses, such as mudstones fall in flat slabs and have a low
BF (typically 0.1 – 0.2). Stronger and more massive rock masses such as
sandstones fall in larger irregular shaped blocks. These blocks rotate, giving rise to
a high void ratio and BF’s in excesses of 0.4 are not uncommon. In general, the
higher the BF the lower the propagation rate.
8.2.1 Impact of Bulking Factor in the Cave Demonstration Model
The effect that an assumed BF has on the cave shape has been investigated by
reducing the maximum allowable bulking factor in the Cave Demonstration Model.
The results of the simulations are provided in Figure 118.

Figure 118. Cave simulation results for variable maximum bulking rates in the
It can be seen from Figure 118A that as the maximum BF of a rock mass decreases,
the cave propagation rate increases. The results of these simulations identify the
problematic nature of assuming bulking rates with analytical cave propagation
assessment techniques.

8.3 Limitations to Height of Draw Scheduling
Within the Pierce et al. (2006) cave model, production is simulated by inducing a
small downwards velocity on gridpoints that are located in the undercut. The
induced gridpoint displacement corresponds to the scheduled Height Of Draw
(HOD). Although the 2006 cave model is able to control the production draw in the
model, the HOD production schedule is not ideal.
At the present time most cave mines use the commercially available block cave
production scheduler PCBC (GEMCOM, 2012) for production planning and
operations control. In instances when PCBC HOD’s are used for scheduling
purposes for geomechanical caving simulations, such as Pierce et al. (2006), this
may lead to over-draw during the early stages of the model simulation. However,
during the later stages of draw, production may be underestimated as a result of
the uniform bulking factors applied to estimate the tonnes withdrawn in the PCBC
schedule. A schematic example of the PCBC-HOD approach is provided in Figure
119.
Figure 119. Schematic representation of a HOD based schedule interpreted for

numerical mesh.
In addition, using this technique, production draw is simulated based on

production increment (groups) that are defined by generalised areas of draw for
any given time period. Prior to production simulation, the extent of the mining
footprint must be defined. The mining footprint is divided into groups that
represent the progressive nature of the expanding hydraulic radius, or different

draw rates during cave initiation. These groups are difficult to change during the
production simulation, and future mining is controlled by average heights of draw
in these initial groupings of model zones. For each production increment, an
average draw rate is assigned based on production tonnes. In general, mining
increments are fixed for the life of mine and defined by the initial undercut
strategy. An example of a mining increment schedule is presented in Figure 120b.
The actual production schedule it may represent is provided in Figure 120a.
Figure 120. Representation of (a) typical production schedule (b) mining increment
schedule (c) improved drawpoint scheduling method.
Using this technique it is known, based on the simulations presented in Figure 117,
that cave propagation rates may be over predicted due to the lack of bulking /
shear induced within the cave mass. The majority of the shear strain will
accumulate at the cave periphery and the internal caved massed will be withdrawn
uniformly as an ‘intact plug’.

8.4 Development of a Mass-Based Production Schedule
Modifications to the production draw algorithm described by Pierce et al. (2006)

are required to allow mass balance calculations in the numerical model of cave
propagation to control production draw based on tonnes extracted at the
drawpoint resolution rather than estimates of average HOD over a mining
increment. The following section provides details of this development.
8.4.1 Simulation of Undercutting
Simulation of undercutting in the numerical model of cave propagation is

consistent with the model developed by Pierce et al. (2006). However,
modifications have been made to ensure that the numerical mesh gridpoints
coincide with the production drawpoint locations (easting, northing and
elevation). In order to achieve this, the model mesh must be developed with the
extraction level layout in mind. I.e., the mesh gridpoint spacing should correspond
to the drawpoint spacing.
The zones immediately below these gridpoints are deleted and replaced with
bulked rock mass properties and stepping of the model is completed to allow the
rock mass to relax. The bulked rock mass is then deleted and the support it
provided to the surrounding rock mass is replaced with equivalent boundary
forces on the floor and walls of the undercut. A summary of the technique is
provided in Figure 22.

8.4.2 Simulation of Production Draw
The perimeter of the undercut footprint is identified through a function that

determines internal (good) and perimeter (bad) gridpoints as shown in Figure
121.
Figure 121. Identification of perimeter gridpoints for production draw simulation in a

numerical mesh.
The identification of perimeter zones is important in distinguishing production

areas from “static” areas in the numerical model, since gridpoints/nodes above a
boundary pillar should not be disturbed.
Production draw continues to be simulated the same way as the Pierce et al.
(2006) model by applying a small downward-oriented velocity to gridpoints in the
model that correspond to drawpoint locations. However, rather than being
generalised over the entire footprint, the velocity of draw (Vdraw) is scaled at the
individual drawpoint level – based on the planned production schedule tonnes.
For example, if the maximum tonnes extracted from a drawpoint over a production
period is 100, and another drawpoints has 50 tonnes scheduled to be withdrawn
from it, then it would be assigned a Vdraw of 50%.

To determine the actual draw velocity within the model, the Vdraw is multiplied by a
maximum draw velocity (Vmax). The maximum velocity should be set low enough to
ensure pseudo-static equilibrium throughout the model.
Representation of the variable draw assigned to production gridpoints is provided

in Table 18.
Table 18. Example of gridpoint velocity scaling based on variable production draw.
# Easting Northing Elevation Tonnes DrawVelocity
1 1 1 0 1 1/3 Vmax
2 1 2 0 2 2/3 Vmax
3 1 3 0 3 3/3 Vmax
8.4.3 Selection of Maximum Draw Velocity (Vmax)
The selection of the Vmax within the model to simulate mining production should be
based on the deformation modulus of the rock mass and the zone size within the
model. Simulated laboratory tests (UCS and direct tension) with varying applied
loading velocities are shown in Figure 122.
Figure 122. Simulated large-scale laboratory tests at different applied loading velocities
and the impact on the sample strength response.

It can be seen from these simulations that as the loading rate (draw velocity)
increases, the simulated compressive strength of the rock mass increases and the
tensile strength decreases.
Implementation of a Vmax that is too high results in a tensile failure mechanism in

the undercut roof that prohibits cave initiation. This has been shown through
production draw simulation in the Cave Demonstration Model where the velocity
of draw applied to a rock mass has been varied. The results are presented in
Figure 123.
Figure 123. Impact of selection of draw velocity on cave propagation behaviour in the

It can be seen that when the applied Vmax is greater than the maximum draw
velocity to maintain pseudo static equilibrium, then a tensile failure develops
immediately above the undercut and the cave stalls as a result of numerical
conditions and not the rock mass strength.
8.4.4 Development of a Tonnes Based Production Cut-Off Algorithm
Production in the numerical model of cave and subsidence assessment is

controlled by mass balance calculations that are performed regularly during model
stepping after an initial pre-mining value is established. As mass is removed from
the system (as result of mesh deformation and the resulting density changes
according to the description in Section 6.1), the mining tonnes can be calculated by
the difference in the pre-mining and current mass states. It is important to note
that production draw should only be calculated based on the mass removed from
the system above the extraction level elevation in the model. This ensures that
material that is removed from the system to facilitate the numerical draw
algorithm is not included in the cumulative production tonnes.
8.5 Summary
A schematic diagram that outlines the implementation of the mass-balance

production draw algorithm as described in this section is presented in Figure 124.

Figure 124. Schematic diagram of the mass-based production draw algorithm developed.
An example of an evolving cave shape based on the mass-balance production

algorithm is provided in Figure 125.

Figure 125. Example of evolving mobilised zone based on drawpoint tonnes algorithm.
It can be seen that the evolution of the mobilised zone is strongly affected by small
variations in production draw from adjacent draw points. The implementation of
the drawpoint tonnes scheme will allow the bulking behaviour associated with
differential draw over the undercut footprint (and between drawpoints) to evolve.
By drawing all drawpoints by the same velocity over the mining increment,
minimal bulking behaviour is simulated and higher propagation rates may be
predicted. The algorithm assumes that the isolated movement zones from
drawpoints overlap at a height just above the drawpoint. This assumes a low
height of interaction and makes it difficult to represent isolated draw conditions
within the model. The coupling of results with a flow program such as REBOP
(Itasca, 2012) is recommended to predict these conditions.

9 DEVELOPMENT OF AN ALGORITHM TO CONSIDER

EVOLVING GROUND SURFACE PROFILE
9.1 Impact of Topography on Subsidence Limits
A crater is a common surface feature of many caving mines; it is also referred to as

the zone of active cave movement (van As et al., 2003). Caved material consists of
irregular blocks of rock, ranging in size from millimetres to several metres in
magnitude. The crater develops as the mobilised zone influence reaches the
ground surface, causing the overlying rock mass and/or side rock to fall into the
mined void. Over time, the surface of the crater may subside as ore is withdrawn
(Lupo, 1998).
Toppling of the crater slopes often occurs (Laubscher, 2000) and this is generally
more pronounced when the crater intersects the side of a slope. At the Questa
Mine, Gilbride et al. (2005) report that large-scale sliding/toppling of the west-
facing hillside above the D Orebody occurred during cave breakthrough as
illustrated in Figure 126. Figure 127 illustrates the surveyed displacements
surrounding the crater.
9 – Development of an Algorithm to Consider Evolving Ground Surface Profile

Figure 126. Photo showing the effect of topography on subsidence crate at the Questa
Mine (after Blodgett, 2002).
Figure 127. Survey displacement map above Questa Mine D Orebody (after Gilbride et
al., 2005).

9.2 Toppling Failure Mechanism
Toppling failure has been defined by Hoek and Bray (1981) as the ‘rotation of
columns or blocks of rock about some fixed base’. Three types of primary toppling
modes have previously been defined by Goodman and Bray (1976) that include;
block, flexural and block-flexure. A schematic diagram of their modes is presented
in Figure 128.
Figure 128. Schematic diagram showing the three primary modes of toppling (after
Goodman and Bray, 1976).
In each of the toppling cases, it is clear that an unconfined face is required to allow
the failure mechanism to develop through the rotation of the blocks.
9.3 Limitations of the Small-Strain Numerical Approach
As discussed previously, due to the large displacements associated with cave

mining; numerical simulations are run in small-strain calculation mode. I.e.,
displacements are accumulated but the mesh is not updated. As a result of this
small-strain calculation mode, the maximum extent of the cave crater and zone of
large-scale fracturing may be under-estimated, since there is limited ability for a
toppling failure mechanism at the crater edge to develop - since the ground surface
profile is not updated, and, a bulked/yielded rock mass is simulated in the crater
which provides support to the fractured rock mass walls – as shown in Figure 129.

Figure 129. Schematic diagram showing the simulation of evolving surface crater in
small-strain calculation mode.
As a result of this small-strain calculation mode, an algorithm is required that

updates the ground surface profile to reflect the evolving crater and allow toppling
failures to develop at the crater limits.

9.4 Development of an Algorithm to Simulate Crater

Development
The development of the subsidence algorithm can be defined by a series of queries

that are performed periodically during model stepping. They are defined below.
1. Identify surface zones
2. Calculate average displacement of gridpoints associated with surface zones
3. If average displacement is greater than the surface zone edge length then
the zone is deleted to form a free face.
The algorithm is presented as a schematic in Figure 130.

Figure 130. Schematic diagram of subsidence algorithm logic.
It is recommended that the algorithm be run at least once per simulated

production increment to ensure the evolving ground surface is represented.

The implementation of the subsidence algorithm has been completed through a

function that is called periodically during production simulation. Its
implementation has been assessed through a subsidence assessment of a block
caving scenario. The validation model geometry is provided in Figure 131.
Figure 131. Geometry and undercut footprint of test model used to validate crater
development algorithm.
Results with and without the algorithm are provided in Figure 132 below.

Figure 132. Subsidence limits predicted with/without surface update algorithm. The
darker (more bold) lines represent the subsidence limits predicted with the
surface update algorithm switched on, and the lighter (fainter) lines represent
the predicted subsidence without using the surface update algorithm.
The implementation of the algorithm shows that the subsidence limits are
increased when a toppling failure mechanism is allowed to develop at the crater
edges. Additional model results are provided in Figure 133 that shows the
updated surface elevation within the model as a result of the algorithm. Any
changes to the topography shown in the figure are due to the implementation of
the algorithm since the small-strain calculation mode has been used. The
development of a crater (depicted by the blue coloured zones) is clearly seen after
the simulation of mining.

Figure 133. Updated surface elevation in the model after the simulation of mining with
the surface update algorithm.
The removal of this material from the model is accounted for in the production
tonnes calculation. The ‘nulled’ zones are depicted by the grey zones in Figure 134.

Figure 134. Vertical displacement simulated in the test model and the surfaces zones
that have been nulled to represent the development of the surface crater.

10 DEVELOPMENT OF A SUB-LEVEL CAVING

ALGORITHM
10.1 Sub-Level Caving Mining Method
blasting. The disintegration is brought about by natural processes that include the
in situ fracturing of the rock mass, stress redistribution, the limited strength of the
rock mass and gravitational forces. Sub-level Caving (SLC) requires the
transformation of in situ ore into a mobile state by conventional drilling and
blasting. This may be a result of a high rock mass strength or strategy to reduce
dilution.
The sub-level caving method is thought to have evolved as an up-scaling technique

to the top slicing mining method (Peele, 1918). Block caving, in turn, was the
logical scale-up from sub-level caving. In the first application of sub-level caving,
the ore was not drilled and blasted completely between two sub-levels, but only
parts were broken by induced caving; hence the name sub-level caving (Janelid,
1972). At current day SLC operations, the ore mass between the sub-levels is
blasted. As a result of this, the primary concern with SLC mining methods is not
the strength of the orebody itself but the competency of the hangingwall material
(for subsidence and dilution predictions) and prediction of fragmentation and
gravity flow of the blasted ore material through the SLC rings.
Existing caving algorithms described by Pierce et al. (2006) have been developed
based on a block and panel caving scenarios only. In order to simulate sub-level
caving, some modifications are required.
10 – Development of a Sub-Level Caving Algorithm

10.2 Simulation of Blast Damage
It is very difficult to estimate the effects of blasting of an in situ rock mass without
site specific trial and/or calibration of numerical results. Empirical evidence from
traditional de-stress blasting in highly stressed mine pillars indicates that blast
patterns with high energy explosives and closely spaced drill holes have been
shown to reduce the strength and deformation modulus of the pillar to some
degree (Andrieux and Hadjigeorgiou, 2008).
The simulation of blasting and damage to the rock mass within the orebody for the
SLC algorithm has been achieved through the modification of joint orientations
within the limits of the blasted sub-level ring. Previous numerical modelling
(Figure 75) has shown that the re-orientation of joints perpendicular to the
direction of draw (in most cases horizontally) will reduce the tensile strength of
the rock mass and promote cave propagation. In order to do this a constitutive
model that allows the specification of joints within the matrix has been used (i.e.,
Subiquitous model in FLAC3D). To ensure an accurate representation of the rock
mass strength, properties were estimated based on the approach discussed in
Section 2.2.2 with the ubiquitous joints glued up and orientated vertically. To
simulate blast damage within the SLC, the joint properties within active sub-level
limits were reduced to zero cohesion and zero tension and the joint orientations
redefined as horizontal. In doing this the rock mass strength was reduced in situ
without causing any damage to the surrounding rock mass material. A schematic
representation of this implementation is presented in Figure 135.

Figure 135. Schematic diagram of sub-level caving algorithm logic.

10.3 Mobilisation of a Previous Sub-Level
The sub-level caving technique requires the development of draw points within
the ore zone and the subsequent mobilisation of this infrastructure into the caved
mass as production continues on progressively lower levels.
The numerical model of cave propagation, as described in Section 8, requires the

deletion of zones immediately below the mining horizon in order to induce
displacement in the material above (see Figure 135). As a result of the production
on the active sub-level that has been completed, these “nulled” zones are required
to be mined as the active sub-levels get progressively deeper. This requires their
re-instatement into the mesh with properties consistent with the in situ rock mass.
However, since they were previously used to induce displacement in the sub-level
above, a significant amount of compaction of their volume has occurred. As a result
of this, when this material is reinstated it has a negative volumetric strain (due to
its compression) as shown in Figure 136.
Since production in the numerical model is tied to density decreases in a zone, and
this is dependent on increasing volumetric strain (expansion), production from
this re-instated material is impossible unless the displacements are reset to zero.
Doing so transfers the negative volumetric strain to the overlying sub-level
allowing the easy volumetric expansion of material from the currently active sub-
level. A schematic implementation of this algorithm is provided in Figure 136.
As a result of these modifications to the displacements within historical sub-levels,

simulated production tonnes are mainly withdrawn from the active sub-level. This
may impact the effect of being able to simulate secondary/tertiary draw since it is
assumed that all tonnes are produced during primary draw. This will impact the
timing of subsidence in the model and, in reality the subsidence may be delayed.

Figure 136. Conceptual model of the volumetric changes in the sub-level caving algorithm
logic.
10.4 Incremental Mass-Based Calculation
Since production in the numerical model of caving is controlled by the gradual

reduction of densities within the model, the spontaneous deletion of mass and re-
introduction of mass within the SLC system must be accounted for. As a result of
this, the mass-balance within the numerical model should be completed on a
mining-increment basis rather than from the initial pre-mining state and mass
density calculations are completed based on the state immediately prior to mining
on the sub-level.

11 CASE STUDY VALIDATION: CAVING INDUCED

FAILURE OF THE PALABORA OPEN PIT
11.1 Background
The Palabora mine began operations as an open cut copper mine in 1964. Today
the pit is approximately 450 m deep and measures nearly 2 km in diameter. A
change in mining method to caving was implemented in the year 2000. Soon after
the breakthrough of the cave to the base of the open pit, a significant pit slope
failure occurred on the north wall, as illustrated in Figure 137.
Figure 137. Photo of the failure in north wall at the Palabora open pit.
Based upon initial numerical back-analyses, the failure mechanism has been
attributed to a persistent joint set that intersects the cave volume at depth
(Brummer et al., 2006).
11 – Palabora Mine Case Study

11.2 Geomechanical Conditions
For each of the caving rock mass domains at Palabora, calibrated continuum
responses have been developed based on the SRM results of Mas Ivars et al.
(2008). The development and calibration of the UJRM’s is discussed in Section 5.
The spatial distribution of each of these domains is provided in Figure 138.
Figure 138. The spatial location of each of the rock mass domains and faults throughout
the Palabora model mesh.
A Hoek-Brown strength estimate has been developed for the granite rock mass
domain that is outside the caving column. The strength properties used to
represent this domain are provided in Table 19.
Table 19. Rock mass properties used for the representation of the granite domain.

Four faults have been represented in the numerical model as presented in Figure
139. Each of the faults has been represented in the model via interfaces as
described in Section 7.7.2. They have been assigned the properties; cohesion = 1.5
MPa, tension = 0.0 MPa and friction = 34 Degrees. These values are consistent with
the joint properties derived for the rock mass and have been developed based on a
sensitivity analysis and calibration to the existing conditions. The joint normal and
shear stiffness’s have been estimated at 1.5 GPa and 0.15 GPa respectively based
on the stiffness of the rock mass and are zone size dependent.
Figure 139. Location of large-scale structure simulated in the Palabora numerical mesh.

11.3 In situ Stress
There is high uncertainty associated with the magnitude and orientation of the in
situ stress regime at the Palabora Mine. Previous stress testing studies have
resulted in significant variation in the principal stress magnitudes and the
orientation of the principal stress direction has previously been measured at
orientations varying between 0o and 340o. A schematic diagram representing the
variance in magnitude and orientation in the principal stress directions are
Figure 140. Estimated in situ stress orientation and magnitude at Palabora based on
back-analysis of pit slope failure and stress measurement testing.

The horizontal stress magnitudes and and orientation =

340o have been simulated for the numerical analysis. This in situ stress condition
was selected based on a sensitivity analysis of the existing north wall slope failure.
This principal stress orientation is consistent with:
(a) The orientation of the large-scale faults mapped at Palabora that strike
north-south and east-west.
(b) The orientation of the major joint sets that strike WNW and NNE.
(c) It is also consistent with the magnitudes derived from a calibration the
existing conditions conducted during 1991 – although the orientation for
this calibration is unknown.
11.4 Production History
Production draw within the model has been simulated based on the methodology
outlined in Section 8. The historical production schedule simulated in the
numerical is provided in Figure 141. The historical heights of draw are provided.
Figure 141. Historical mining record at the Palabora block cave mine.

During the simulated production draw, each of the cave behavioural regions were
tracked and compared to physical observations made at the Palabora mine during
that time period.
11.5 Simulation Results
11.5.1 Cave Initiation
The location and magnitude of seismic events recorded during the early stages of
production at the Palabora mine are illustrated in Figure 142(a). Based on this
data, the location of the yield zone (or aseismic zone) has been inferred to extend
approximately 55–83 m beyond the mobilised zone. As illustrated in Figure
139(b), the predicted yield zone within the numerical model extends
approximately 50–80 m above the cave zone, providing a good correlation with the
monitoring data.
Figure 142. Observed seismicity at the Palabora Mine during cave initiation and
propagation (a) observed mobilised, yield and seismogenic zones during
production (after Glazer and Hepworth, 2004); (b) numerical prediction of
mobilised and yield zones during production simulation at the Palabora
mine.

The seismogenic zone has been predicted at various stages of production in the
numerical model using the methodology described by Diederichs (1999). A
value of 0.42 has been used based on a calibration to SRM test results (Mas Ivars et
al., 2006). As illustrated in Figure 143, the seismogenic zone is seen to manifest
immediately beneath the floor of the open pit in the early stages of mining. As
mining progresses and the crown pillar fails, seismicity migrates to the lateral
extents of the undercut footprint, prior to progressing beneath the extraction level
as production continues. This sequence is consistent with the seismic record
collected on site and discussed by Glazer and Hepworth (2004).
Figure 143. Numerical prediction of seismogenic zones during early production

simulation at the Palabora mine.

11.5.2 Yielding of the Crown Pillar – Q4 2002
During the fourth quarter of 2002, the yielded rock mass (aseismic) zone connects
through to the open pit. Prior to this time, an un-yielded crown pillar remains.
The yielding of the crown pillar drives seismicity (high induced stresses) beneath
the extraction level. The results of the numerical simulation for this period in time
are provided in Figure 144.
Figure 144. Numerical simulation - yielding of the crown pillar during Q4 2002.
The simulated yielded rock mass zone within the numerical model breaks through
to the open pit at the same time as the aseismic zone on site that was interpreted
by Glazer (2006).

11.5.3 Cave Break-Through – Q1 2004
During the first quarter of 2004, the mobilised zone connects through to the pit
floor. Immediately after this first instance of mobilisation, a crater emerges.
Mobilisation within the numerical model occurs initially along pre-existing fault
traces in the base of the open pit as presented in Figure 145.
Figure 145. Numerical simulation – cave breakthrough during Q1 2004.
The simulation results show that the mobilised zone intersects the pit floor during
Q1 -2004. This is consistent with the interpretations made by Glazer (2006). In
addition, immediately prior to the development of the crater within the pit, the
shape of the cave back in the numerical model (Figure 146b) is consistent with the
shape of the cave back interpreted by Glazer (2006) that is provided in Figure
146a.

Figure 146. (a) Cave profiles at the Palabora Mine; April 2002 to December 2003
(after Glazer, 2006) compared to the simulated cave profile (b).

11.5.4 North Wall Failure – Q4 2004
During the fourth quarter of 2004, mining on the extraction level extends further
west. As material is withdrawn from these drawbells the mobilised zone intersects
a number of faults that act, along with localised jointing, as a failure surface.
During this increment, the mobilisation and fracturing of the ground surface up to
the top of the pit in the area of the north wall failure occurs in the model. The
simulated model state at the end of 2004 is provided in Figure 147.
Figure 147. Numerical simulation – north wall failure during Q4 2004.
The timing of this event in the numerical model is consistent with the observations
made on site.

11.5.4.1 North Wall Failure Mechanism
An analysis of the location of the north wall failure within the numerical model in
relation to the observed location of the large-scale failure on site is presented in
Figure 148. A good match is observed between the numerical model and the
observed limits of the north wall failure at the ground surface.
Figure 148. North wall failure: observed versus simulated limits.
Evolution of the pit slope failure mechanism is illustrated in Figure 149. An

increase in production draw from the western portion of the undercut footprint is
observed to cause a rapid vertical advance of the yield zone which intersects the
base of the pit. Breakthrough of the cave volume is observed to mobilise the sub-
vertical joints in the pit slope immediately above the yielded zone.

Figure 149. Development of the pit slope failure mechanism at the Palabora Mine at
various stages of production.
Based on the current model results, the north wall failure manifests as a result of
mining on the extraction level that progresses west and undermines the slope. The
presence of intersecting faults and an unfavourable joint orientation provides
release surfaces for the slope to unravel along. A view of the model results in
relation to the simulated faults is provided in Figure 150.

Figure 150. Development of the Palabora block cave between 2003 and 2004 in
relation to fault structure.
11.6 Summary
The numerical model used for the simulation of historical conditions at Palabora
are able to provide a propagation rate and cave shape that fits with the interpreted
cave behaviour during 2002-2004.
This suggests that:
(a) The geomechanical property estimates are reasonable assessments of

the rock mass strength at Palabora. This verifies the SRM testing
behaviour, and the implementation of the calibrated continuum
responses in the model.
(b) The production schedule simulated provides an accurate assessment of

the bulking/dilation of the caved rock mass. This verifies the
implementation of the new production scheduling techniques as well as
the rock mass dilatational and softening constitutive behaviour.

(c) The effect of structure on the evolution of the cave has been accurately
accounted for within the numerical model through the implementation
of the explicit fault technique.
(d) The criteria for assessing the cave behavioural regions based on the
caving and subsidence criteria outlined in Section 1.4 are valid at the
Palabora Mine site.

12 CASE STUDY VALIDATION : STRUCTURALLY

CONTROLLED CAVING AT THE HENDERSON MINE
12.1 History of the Henderson Mine
A weak geological contact has been observed to effect cave propagation and cave
shape on the 7210 Level of the Henderson Mine (Carlson and Golden Jr., 2008).
Climax Molybdenum Company’s Henderson Mine is an underground panel-caving
mine located in Clear Creek County Colorado and is 14 km west of Empire,
Colorado, USA. Figure 151 provides a general cross-section of the mining
geometry.
Figure 151. Cross section of the Henderson Mine (after Rech, 2001).
During undercutting of the 7210 Level in 2005, cave initiation was observed with a
relatively small undercut hydraulic radius (22 m) compared to what has been
observed during other cave initiations at the Mine (HR 35 m). Carlson and Golden
Jr. (2008) report that the presence of intrusive contacts along the northern
boundary of the 7210 Level undercut are responsible for the premature cave
12 – Henderson Mine Case Study

initiation (Figure 152a). Past experience at the Henderson Mine has shown that
intrusive contacts are weak zones that fail quickly.
During early August 2006, migration of the propagating cave beyond the undercut
footprint on the north and west sides was observed along the weak intrusive
contact. Figure 152b illustrates the shape of the 7210 Level yield zone during
December 2007.
Figure 152. Geological domains at the Henderson Mine a) plan view of weak contact;
b) 7210 Level yield zone during December 2007.
12.2 Model Geometry and Production Schedule
A large-scale FLAC3D model was constructed to simulate the regional extents of the
Henderson Mine, as illustrated in Figure 153a. The existing cave volumes (8100
and 7700), developed prior to the 7210 Level were initialised within the model
based upon historical mining records, as illustrated in Figure 153b.

Figure 153. Development of the numerical model of the Henderson Mine a) regional
extents of model; b) existing cave volumes.
The weak contact between the Seriate and Urad Porphyry was simulated with an
interface element as illustrated in Figure 154.
Figure 154. Interface used to simulate the weak Seriate contact at the Henderson Mine.
The evolution of production draw height simulated within the model (based upon
the actual production records) is illustrated in Figure 155. It is represented as
solid rock HOD bars.

Figure 155. Simulated production schedule (cumulative solid rock height of draw) based
on actual draw heights.
12.3 Material Properties and Pre-Mining Stresses
Multiple geotechnical domains have been mapped throughout the 7210 Level. The
geomechanical properties used to simulate the main caving domain, the Urad
Porphyry, are presented in Table 20.
Table 20. Rock mass geomechanical properties of the porphyry at the Henderson Mine.
Seg.1 Seg.2
ci Erm Tens. Coh  Coh. 
(MPa) GSI mi (GPa) v (MPa) (MPa) (Deg.) (MPa)
(Deg.)
Porphyry 118 55 10 11 0.2 0.4 2.3 48 6.3 35
Table 21 presents the interface material properties used to simulate the weak
geological contact.

Table 21. Conceptual fault shear strength and stiffness parameters estimated for the seriate
contact at the Henderson Mine.
Coh.  Tens. Kn Ks
(kPa) (Deg.) (kPa) (GPa) (GPa)
Strength 1 0 20 0 1 0.1
Due to the significant topographic relief and complex previous mining history at
Henderson, it is difficult to estimate the pre-mining stress regime. A stress
calibration exercise has previously been conducted at Henderson whereby nine
stress measurements taken from 1970 to 1989 were calibrated determining the in
situ tectonic stresses that result in a best-fit of model predicted stress to those
measured by over coring. The results of the stress calibration exercise, which
indicated a major principal stress oriented at 155 degrees, were directly applied to
the back-analysis model.
12.4 Simulation Results
The evolution of the model-predicted yield zone is illustrated in Figure 156. The
modelled yield zone is observed to provide a close match to the TDR breakages
(blue spheres) monitored during cave propagation. Shear failure along the weak
contact can be observed to develop along the interface - coincident with vertical
propagation of the yield zone. After initial breakthrough of the yield zone to the
overlying 7700 Level in January 2006, the yield zone is observed to follow the
weak contact outside the northern and western limits of the undercut footprint.
The modelled cave breakthrough timing and dimensions match closely the
observed 7700 level breakthroughs as shown Figure 156.

Figure 156. Simulated evolution of the cave yield zone at Henderson compared to
underground instrument observations.

Figure 157 illustrates a comparison between the shape modelled and actual yield
zones after September 2006. The model provides a close match to the actual
observed conditions.
Figure 157. Comparison of modelled verses actual cave shape at break-through to the
overlying lift at the Henderson Mine.
In order to demonstrate the effect of the weak contact on cave initiation, an

analysis was conducted without the structure. As illustrated in Figure 158, the
yield zone height is significantly impacted by the weak interface with height
reduced by approximately 30 m at the June 2005 state. Clearly, the cave shape and
growth and the ability to predict this behaviour depends on the presence of the
weak interface.
Figure 158. Comparison of cave initiation at the Henderson Mine simulated with and
without a weak structure.

12.5 Summary
The numerical model used for the simulation of historical conditions at the
Henderson Mine are able to provide a propagation rate and cave shape that fits
with the interpreted cave behaviour for the 7210 Level.
This suggests that:
(e) The use of a bi-linear Mohr-Coulomb fit to the Hoek-Brown Curve

provides geomechanical property estimates that are reasonable
assessments of the rock mass strength and softening responses.
(f) The production schedule simulated provides an accurate assessment of

the rock mass dilational and softening constitutive behaviour.
(g) The effect of structure on the evolution of the cave has been accurately
accounted for within the numerical model through the implementation
of the explicit fault technique.
(h) The criteria for assessing the cave behavioural regions based on the
caving criteria outlined in Section 1.4 are valid at the Henderson Mine
site.

13 CASE STUDY VALLIDATION : CAVING INDUCED

SUBSIDENCE AT THE ABANDONED GRACE MINE
PANEL CAVE
13.1 Background
Iron ore mining began in south-eastern Pennsylvania before the American

Revolutionary War (1775–1783) and reached the peak of activity during the
1880s. In 1948, airborne magnetometer surveying conducted by the Bethlehem
Steel Corporation led to the discovery of a large, deep magnetite deposit located
approximately 3.2 km north of Morgantown, in Berks County, south-eastern
Pennsylvania. Diamond drilling from 1949 to 1951 delineated the orebody, which
was named the Grace Mine.
Development of the Grace Mine commenced in 1951 with construction of two

vertical shafts to gain access to the orebody. Shaft A was developed to a depth of
784 m, and shaft B was developed to a depth of 938 m with a diameter of 5.3 m.
The two shafts and all surface infrastructures were constructed outside an area
formed by a conservative 45o subsidence angle (measured from the base of the
underground workings). None of the existing Grace Mine buildings have shown
signs of subsidence induced damage to date. The Grace Mine orebody contained an
estimated 107 million tonnes of magnetite averaging 40% iron ore. During the
active mining period 1958–1977, approximately 33 million tonnes of ore were
mined primarily with the inclined panel caving method.
Figure 159a illustrates the extent of the underground development at Grace Mine
together with the approximate undercut outline. Figure 159b illustrates a
schematic of the inclined panel caving method employed at the Grace Mine. Mining
operations ceased in 1977 due to an influx in foreign steel imports, increased costs
of environmental regulation and increased costs of underground mining.
13 – Grace Mine Case Study

Figure 159. Geometry of the Grace Mine a) extent of underground production drives
(plan view) b) two-dimensional schematic of panel caving at Grace Mine
(after Stafford, 2002).
At the completion of mining, mining induced subsidence had significantly altered

the topography above the mine workings. Mine dewatering continued until 1981-
2. Upon recovery of the water table, a lake formed over the subsided area, as
Figure 160. Photo showing present day subsidence lake at the Grace Mine.

13.2 Geomechanical Properties
Currently, the only core information available from the Grace Mine site is drilling
logs from a diamond drilling programme conducted in 1998 to investigate the near
surface conditions for construction of a large scale industrial facility near the
western edge of the subsidence lake. Owing to the location of the drill holes above
the mining horizon, the intersected rock mass has been disturbed by the caving
process and does not represent the in situ (pre-mining) rock mass condition.
The following descriptions of the rock units at the Grace Mine have been compiled
from the aforementioned drilling logs, mapping of surface outcrops and other
available literature.
13.3 Local Geology
Three rock types are associated with the Grace Mine: diabase footwall, replaced
limestone and Triassic sediments. Sims (1968) suggested that the magnetite
deposit occurs in a lens of Cambrian limestone that is overlain unconformably by
Triassic sedimentary rocks. The magnetite deposit was formed by replacement of
contact metamorphic minerals in the limestone lens, caused by the intrusion of an
underlying diabase sheet. The orebody is roughly tabular in shape, strikes
approximately 60o and dips 20–30o to the northeast. It is approximately 1067 m
long and 213–457 m wide, and ranges from less than 15 m to more than 121 m in
thickness. Figure 161a illustrates an isometric view of the original ground surface
and orebody shape. The surfaces were reconstructed from the original mine
geological cross-sections.

Figure 161. Isometric view of orebody and surface topography at the Grace Mine.
Of interest to the investigation of mining induced subsidence is the nature of the

sedimentary rocks that overlie the orebody. Sims (1968) and Basu (1974) identified
the Triassic sedimentary rocks that formed the hangingwall of the ore deposit as
the Stockton Formation. The formation consists of inter-bedded sandstone, shale
and conglomerate. The Stockton Formation is overlain by the Brunswick
Formation, which also consists of poorly sorted inter-fingered sandstone, shale
and conglomerate layers. The top of the Brunswick Formation has been eroded to
form the ground surface directly above the orebody. Surface outcrops of the
Stockton Formation near the Grace Mine are thinly bedded and highly fractured,
with very closely spaced (<100 mm) vertical and open joints. The conglomerate
layers of the Stockton Formation are thickly bedded (0.5–1.5 m) and moderately
fractured (0.5–2 m), with vertical and open joints.

Table 22. Rock Mass Parameters used in the simulation of domains at the Grace Mine.
UCS
Material (MPa) GSI mi
Sediments 60 45 12
Limestone 90 55 12
Magnetite 40 45 9
Diabase 200 60 25
13.4 Pre-Mining Stress State
The pre-mining stress was measured directly at Grace Mine using the United States
Bureau of Mines (USBM) Deformation Gage Technique. Agarwal et al. (1973)
detailed the measurement procedure for 36 deformation measurements from
three separate boreholes. The boreholes were located to ensure that the measured
stress regime was representative of the pre-mining stress regime (unaffected by
nearby excavations). The stress regime presented in Table 23 is an average of
measurements made at a depth of 731m below the ground surface.
Table 23. Pre-mining stress regime at 731m below surface at the Grace Mine.
Stress Magnitude (MPa) Dip (Deg.) Azimuth (Deg.)
1 51.5 16 027
2 29.0 22 103
3 26.2 85 181
13.5 Caving Induced Subsidence
13.5.1 Evolution of Subsidence Crater and Trough
From 1959 to 1969, engineers from the USBM monitored the evolution of surface
subsidence at the Grace Mine. No USBM report of investigation was published on
the extensive subsidence monitoring programme. However, during the course of
this investigation, several hand written memorandums from USBM Engineers to
management of the Grace Mine were recovered from former Grace Mine
superintendent, Mr Charles Taylor’s private collection. Goodman (1970) reported
the evolution of the subsidence trough as follows:
 10 December 1962: first indication that subsidence was beginning

 18 February 1963: cracks on the surface were noted, and a slumped zone
widened and steadily descended
 16 December 1963: the underground caved to the surface
 17 April 1964: cracks extended to circle the crater and extended in a

concentric pattern
 8 June 1965: the subsidence trough was observed to progress to the

northeast
 3 June 1969: the subsidence trough moves to the northeast following the
development and extraction pattern.
Figure 162 illustrates different aerial views of the subsidence trough. The initial
cave breakthrough appears to have been facilitated by the presence of a steeply
dipping joint/fault structure oriented at approximately 60o. As the subsidence
trough progressed to the northeast, the actual cave did not break through to the
surface. In the south-western section of the subsidence trough, large concentric
surface cracks can be observed, while two sets of cracks oriented at approximately
60o and 110o are observed in the north-eastern section. The ground within the
approximate extent of surface cracking can be clearly observed to be highly
fractured and disturbed.
Approximately 30 monuments were installed on the surface above the mining

horizon and monitored by the USBM between 1962 and 1969. Surveying
techniques used over the monitoring period included chaining, levelling and
triangulation. The maximum elevation change as of 1969 was reported to be 35 m.
Goodman (1970) noted that uplift generally occurred in pins around the periphery
of the orebody outline, while acceleration of subsidence was greatest directly over
recently developed panels. Surface uplift outside the undercut footprint was also
monitored at the Lakeshore Mine, in Arizona (Panek, 1984).

Figure 162. Photos showing the evolution of subsidence from 1963 to 1978 at the
Grace Mine.
13.5.2 Visual Observation of the Limit of Large-Scale Surface Cracking
Figure 163 illustrates the results of a field survey conducted during June 2004 to
identify the extent of large-scale surface cracking surrounding the subsidence
trough. The furthest observable surface cracks from the orebody outline have been
mapped and used to generate a contour line that represents the limit of the large-
scale surface-cracking zone (fractured zone).
In the southwest section of the subsidence trough, a shear-failure mechanism

appears to be predominant. Large scarps with offsets of approximately 3 m can be
observed. In the northeast section of the subsidence trough, both large tension
cracks and shear failure scarps can be observed. In the western section of the
subsidence trough, a continuous trough-like basin has formed over the shallower,
thinner section of the orebody. A sinkhole that developed after reclamation of the
waste dumps can be observed within the limits of the mine workings.

Figure 163. Limit of large-scale surface cracking observed at the Grace Mine during
2005 (after Sainsbury and Lorig, 2005).
13.5.3 Conceptual Model of Subsidence Formation at the Grace Mine
Figure 164 illustrates a conceptual model of the caving and subsidence formation
at the Grace Mine. In the years after initial breakthrough, the size of the actual
crater did not increase significantly, indicating that bulking-controlled caving had
prevented the progression of the caved zone to the ground surface. The
subsequent subsidence was observed to manifest as a subsidence trough following
the direction of mining.

Figure 164. Conceptual model of subsidence formation at the abandoned Grace Mine
(after Sainsbury and Lorig, 2005).
13.5.4 Modelling Methodology
In order to simulate the mass based production schedule at the Grace Mine,
recovered from actual hoist records (Table 24), production simulation within the
numerical caving model has completed based on the methods described in Section
8.
Bi-linear Mohr-Coulomb material properties were derived from a least-squares fit

to the Hoek-Brown failure envelope for each rock type using the GSI, UCS and mi
values presented in Table 22.
Table 24. Grace Mine production (after Eben, 2004 ).

1958 1961 1962 1965 1967 1969 1972 1975
- - - - - - - -
1960 1962 1963 1964 1966 1968 1971 1974 1977
MTonnes 1.18 2.89 2.35 2.23 4.04 3.65 5.74 5.74 4.82
Figure 165(a) illustrates the regional model geometry, while Figure 165(b)
illustrates the reconstructed undercut geometry used to represent the historical
production schedule.

Figure 165. Development of a numerical model of the Grace Mine a) regional extents of
model b) simulated undercut footprint.
13.5.5 Predicted Evolution of Cave Mobilised and Yield Zones
The evolution of the cave mobilised and yield zones above the undercut footprint
is illustrated in Figure 166. The predicted cave behaviour provides a close match
with the reported evolution of cave breakthrough and subsidence trough
formation. The yield zone is predicted to intersect the ground surface during 1962
(Figure 166b); while the mobilised zone is predicted to intersect the ground
surface towards the end of 1963 (Figure 166c). This coincides with the first
indication of subsidence reported on Dec. 10, 1962 and the breakthrough of the
caved (mobilised zone) on Dec. 16, 1963.
As production progresses to the thinner and deeper extents of the orebody (to the
east), the predicted mobilised zone does not reach the ground surface (Figure
166f). This coincides with the formation of a subsidence trough, rather than an
extension of the subsidence crater over the northeast region of the orebody.

Figure 166. Predicted evolution of cave mobilised and yield zones (looking south) during
simulation of production from the Grace Mine.

13.5.6 Model Validation
The predicted vertical displacement was monitored within the model at the same
locations as the USBM subsidence monitoring monuments that were documented
by Goodman (1970). The measured versus predicted vertical displacement at three
survey monuments surrounding the subsidence trough are illustrated in Figure
167. The predicted surface displacements provide a close match to the monitoring
results and provide good confidence in the predicted surface displacements
beyond 1969.
Figure 167. Measured verses predicted vertical displacements from numerical model of
Grace Mine.
The introduction of the tonnes-based production schedule (rather than traditional

height of draw) has highlighted the mechanism for the glory hole at the surface
that was created during 1963. A typical PCBC HOD schedule usually under-
estimates the tonnes withdrawn as a result of the uniform bulking factor applied.
The result of assuming a HOD production schedule with a typical Bulking Factor of
0.2 is provided in Figure 168.

Figure 168. Comparison of cave propagation results (a) HOD schedule - uniform
bulking factor of 0.2 (b) mass balance (tonnes-based) schedule.
13.6 Summary
The numerical model used for the simulation of historical conditions at the
abandoned Grace Mine are able to provide a propagation rate and cave shape that
fits with the interpreted cave behaviour.
This suggests that:
(a) The use of a bi-linear Mohr-Coulomb fit to the Hoek-Brown Curve

assessments of the rock mass strength and softening behaviour at the
abandoned Grace Mine.

the rock mass dilational and softening constitutive behaviour.

(c) The criteria for assessing the cave behavioural regions based on the
caving criteria outlined in Section 1.4 are valid at the abandoned Grace
Mine site.

14 CASE STUDY VALIDATION : CAVING INDUCED

SUBSIDENCE AT THE KIIRUNAVAARA LAKE
OREBODY SLC
14.1 Introduction
The Kiirunavaara mine, owned and operated by LKAB, is located in northern

Sweden near the township of Kiruna, approximately 180 km north of the Arctic
Circle. The orebody is 4 km long, 80-160 m thick and the mineralisation reaches a
depth of at least 2 km. It dips 50° to 60° to the east and plunges to the north-
northeast as the orebody lens tapers out. The orebody has been divided into two
main regions of mining; the Main Orebody and the Lake Orebody. The Lake
Orebody defines the northern extent of the mine and is located immediately west
of the township of Kiirunavaara (Figure 169).
Figure 169. Location and extent of the Lake and Main Orebody at the Kiirunavaara
Mine.
14.2 Historical Mining Record
Kiirunavaara is a transverse, sub-level caving mine that commenced underground

operation during the early 1960s after initially being mined as an open pit since
the start of the 20th century. Production from the non-daylighting Lake Orebody
commenced in 2003 when workings in the Main Orebody were on the 792 Level
(550 m below ground surface). During the years 2003-2010, the lowest level of
14 – Kiirunavaara Lake Orebody Case Study

production from the Lake Orebody was the 792 Level, and the 935 m Level (693 m
below ground surface) in the Main Orebody (Figure 169).
14.3 Evolution of Surface Subsidence
Since sub-level caving commenced at the Kiirunavaara Mine, the hangingwall has
experienced surface displacements ranging from millimetres up to several metres
in magnitude. The development of caving induced subsidence around the Lake
Orebody has been monitored through routine ground surveys since production
commenced in 2003. The following section provides a summary of the
observations and measurements. The observations are made in reference to cave
subsidence zones that are described in Section 1.4. A view of the existing surface
conditions are presented in Figure 170.
Figure 170. Subsidence regions at Kiirunavaara.
Approximately three years after mining commenced in the Lake Orebody, a crater
developed on the northern extents of the existing open pit - as illustrated in Figure
171. Initially developed as an isolated subsidence feature during 2006, additional
production during 2007 and 2008 caused the enlargement of the crater towards

the south. The development of this isolated crater can best be described as a
chimney or plug cave. Lupo (1997) completed a detailed review of the chimney
subsidence features that occur east of the Main Orebody, and suggested that they
are formed when the flow channel of a sub-level ring reaches the ground surface.
Figure 171. Photos showing the development of a crater at northern extent of Lake
Orebody during 2006 and its subsequent enlargement.
14.3.1 Limits of Large-Scale Fracturing / Yield Zone
The progression of the large-scale fracture limits at the ground surface between
1997 and 2006 is presented in Figure 172. An angle of break for the fractured
zone of approximately 60o has been reported by Villegas et al. (2011), Lupo (1996)
and Stephansson et al. (1978). Surface disturbances in this zone have previously
been documented by Lupo (1997) and consist largely of surface cracks, and shear
displacements.
Figure 172. Fracture mapping at Kiirunavaara (a) plan above Lake Orebody (b)
section through Main Orebody (modified after Villegas et al., 2011).

14.3.2 Limits of Continuous Deformation
GPS data shows that an area of continuous deformation extends approximately

150-200 m beyond the limits of large-scale fracturing (Villegas et al., 2011). The
measured surface displacements from 2002 – 2010 above the Lake Orebody are
provided in Figure 173. Surface displacements in the order of 0-250 mm were
measured prior to the commencement of mining of the Lake Orebody in 2003.
Figure 173. Evolution of measured surface total displacement profile around the Lake
Orebody.
14.4 Numerical Simulation of Caving Induced Subsidence
A three-dimensional model of the Kiirunavaara Mine and its surroundings has

been developed to assess the impact of the production schedule on the
development of the surface displacement profile that is evident today. The model
extents are provided in Figure 174.
Figure 174. Regional extents of Kiirunavaara numerical mesh.

14.4.1 In situ Stress
The pre-mining in situ stresses at the Kiirunavaara Mine have previously

documented by Sandström (2003). The major principal stress is estimated to be
aligned perpendicular to the orebody and is approximately 1.28 times the vertical
stress. Equations for deriving the principal stress components in MPa are
provided below.
    W  0.37 ym  3.7  h   NS  0.28 ym  2.8  v  0.029 ym  2.9

; ;
Where ym represents the depth below -100m RL and the stresses are expressed in
terms of MPa.
14.4.2 Rock Mass Properties
Historically, material properties for the Lake Orebody have been developed based
on a calibrated response of drive scale displacements and failure mechanisms in
the Main Orebody. Previous analyses conducted by Perman et al., (2011) have
derived a lower bound property set for the hangingwall domain in the Main
Orebody that is defined by a UCS 130 MPa, GSI 58, mi 16 and Erm 15.8 GPa. This
GSI value has been confirmed for the Lake Orebody by scanline mapping
conducted during 2010.
A bi-linear, Mohr-Coulomb, strain-softening constitutive model has been used to

simulate the complex process of the progressive failure and disintegration of the
rock mass from an intact, jointed material to a bulked state during the caving
process in the numerical model. The low magnitude in situ stresses in relation to
the strength estimates suggest a gravity driven caving mechanism is dominant at
Kiirunavaara.
14.4.3 Production Schedule
In order to ensure an accurate induced stress state in the model prior to the
simulation of mining from the Lake Orebody, simulation of the extents of open-cut
mining was conducted during the development of the initial model state.
Production from the Main Orebody has been simulated based on an elevation and

tonnes basis. Production from the Lake Orebody has been scheduled on a
drawpoint and tonnes basis consistent with the technique described in Section 8.
14.4.4 Simulation Results
Based on the displacement and strain criteria outlined in Section 3, the

development of historical caving induced subsidence from the Lake Orebody has
been assessed. The simulated crater and limit of large-scale fracturing is
Figure 175. Simulated evolution of crater and limits of large-scale fracturing within the
numerical model of Kiirunavaara.
Mobilisation of the ground surface above the Lake Orebody during 2006 is
observed in the numerical model. The location of this initial break-through is on
the northern extents of the existing open pit and is consistent with the in situ
observations. As the mining simulation continues beyond 2006, the crater is
observed to advance towards the south and joins with the main orebody crater
during the production years 2007 – 2008.
The development and enlargement of this crater can be attributed to the draw
schedule at the mine, since it occurred immediately above an area where draw was
occurring on multiple sub-levels that overlapped each other (vertically) at the
same time (Figure 176). The crater has formed when the flow channels of the sub-

level rings have combined. The rapid propagation to the surface suggests that the
secondary and tertiary flow channels have also contributed to the formation of the
isolated crater.
Figure 176. Production rate versus simulated mobilised zone within the numerical model
of Kiirunavaara.
Simulated displacements on the ground surface immediately above the Lake

Orebody from 2002 – 2010 are presented in Figure 177. The initial pre-mining
displacements along with the development of the crater location and shape are
consistent with the measured displacements. Since the monitoring points used in
each of the contour plots (as shown in Figure 177) are not consistent between the
observation years, significant interpretation between them has been completed. In
addition, the methodology used to complete the observation measurements does
not provide a great level of accuracy (+/- 0.3 m) and only general trends can be
interpreted.

Figure 177. Simulated evolution of total surface displacements within the numerical
model of Kiirunavaara compared to observations onsite.
The 1 m displacement contour has been used to define the crater limits within the
numerical model at the end of 2010 (Figure 178). The simulated limits compare
well with the in situ observations based on their comparison to the visual
observations presented in Figure 171.
Figure 178. Plan view of simulated displacement and strain-based subsidence criteria
and subsidence zone of influence at the end of 2010 as simulated in the
numerical model of Kiirunavaara.

A total strain criterion of 0.5% was used to confirm the limits of large-scale
fracturing that have an angle of draw consistent with approximately 60o. The
simulated extent and shape also compares well to the fracture limits defined by
Stöckel et al. (2012) presented in Figure 179.
The limits of continuous subsidence at the end of 2010 have been derived by
generating a contour line that encompasses all the areas of horizontal strain >0.2%
and angular distortion >0.3%. It extends approximately 200 m beyond the limits
of large-scale fracturing which is consistent with previous observations
documented by Villegas et al. (2011).
Figure 179. Plan view of simulated subsidence limits at the end of 2010 compared to
observations at Kiirunavaara.

14.5 Summary
The observed limits of caving induced subsidence at the Kiirunavaara Lake

Orebody have been accurately assessed by a numerical simulation of sub-level
caving from 2003-2010. Established displacement and strain-based criteria have
been used to successfully back-analyse the evolution of caving induced subsidence.
This suggests that:
(a) The use of a bi-linear Mohr-Coulomb fit to the Hoek-Brown Curve

assessments of the hangingwall rock mass response at the Lake
Orebody.

implementation of the new SLC production scheduling techniques as
well as the rock mass dilational and softening constitutive behaviour.
(c) The criteria for assessing the cave behavioural regions based on the
caving and subsidence criteria outlined in Section 1.4 are valid at the
Kiirunavaara Mine site.

15 CONCLUSIONS AND RECOMENDATIONS
Since the development of numerical methods in the 1970’s, the numerical

assessment of cave propagation has been continuously researched and improved
to help minimise the geotechnical risks associated with cave mining methods.
Since this time, numerical methodologies have evolved from being able to
accurately assess the primary risk of whether a cave will stall and develop an air-
gap, to now being able to assess detailed cave behaviour. Contrary to the opinion
of Laubscher (2000) who believes that “…modelling is not capable of coping with …
four dimensions”. Today, cave modelling methodologies are limited only by the
quality of data that can be collected in situ.
The numerical model of cave propagation and subsidence assessment that has
been developed provides an assessment of the evolving cave propagation
behaviour and subsidence zones of influence in response to the actual production
draw at the undercut/extraction level. The evolving cave shapes, propagation
rates, abutment stresses and subsidence limits can be readily assessed within the
one numerical model.
Comparison of the model predictions with the results of the four large-scale back-
analyses demonstrates that the model and cave and subsidence assessment
criteria are robust.
15 - Recommendations
15.1 Summary of Original Contributions
A model for cave propagation and subsidence assessment in jointed rock masses
has been developed that relies on the fundamental behaviour of rock masses for
caving analyses. In the process, original contributions were made to the numerical
simulation of rock mass behaviour and numerical methods for cave propagation
and subsidence assessment. A summary of the contributions are provided below.
15.1.1 Rock Mass Behaviour
15.1.1.1 Development of the Ubiquitous Joint Rock Mass (UJRM) Model
Synthetic Rock Mass Modelling (SRM) is generally accepted as the existing state-of-
the-art in anisotropic rock mass behaviour analysis. A methodology has been
developed that can be used to derive material input properties for the FLAC3D
Subiquitous (Strain-Softening Ubiquitous Joint) constitutive model so that it
exhibits strength and deformation behaviours similar to what may be derived from
SRM testing. The successful implementation of these strengths in a large-scale
caving back-analysis at the Palabora Mine provides validation for the technique.
15.1.1.2 Consideration of the Volumetric Changes that Accompany Cave

Propagation
Due to computational constraints at the present time, the numerical model of cave
propagation must be implemented using a small-strain calculation mode. As a
result of this, the manual modification of the density and bulking/dilational
behaviour of the rock mass during volumetric expansion is required. A non-linear
deformation modulus softening and dilation relation has been implemented within
the numerical model of cave propagation to provide a more rigorous assessment of
rock mass bulking. The methodology has been validated based on four large-scale
back-analyses of cave propagation behaviour detailed herein.
15.1.1.3 Impact of Large-Scale Discontinuities on Cave Propagation and

Subsidence Behaviour
The impact of large-scale structures on cave propagation and subsidence

behaviour is well known. Two methodologies used to simulate large-scale fault
structures in numerical meshes have been investigated. Based on the numerical
results and their comparison to case study observations, an explicit approach to
fault simulation has been validated and is recommended for the simulation of
large-scale discontinuities in caving analyses. The approach has been validated
through a large-scale back-analysis of caving behaviour at the Henderson Mine.
15.1.2 Production Simulation
15.1.2.1 Development of a Mass-Based Production Schedule
The existing methods for the simulation of production draw in a numerical model
have been expanded to include the simulation of production draw through mass-
balance calculations - rather than HOD estimates. In addition, a methodology has
been outlined that provides a production scheduling technique based on
drawpoint tonnes to allow for the accurate consideration of the bulking/dilational
behaviour evolving within the cave. The methodology has been validated based on
four large-scale back-analysis of cave propagation behaviour detailed herein.
15.1.2.2 Development of a Sub-Level Caving Algorithm
blasting. The disintegration is brought about by natural processes that include the
in situ fracturing of the rock mass, stress redistribution, the limited strength of the
rock mass and gravitational forces. Sub-level caving requires the transformation of
in situ ore into a mobile state by conventional drilling and blasting. The existing
numerical techniques for cave behaviour analysis were unable to represent a SLC
mining method. A technique has been developed that allows a sub-level caving
scheduled to be accurately reflected in the numerical model of cave propagation. It
has been validated through a large-scale back-analysis of caving behaviour at the
Kiirunavaara Lake Orebody.
15.1.2.3 Development of an Algorithm to Consider Evolving Surface Profile
Since the numerical simulations are conducted in small-strain (i.e., the mesh
gridpoint locations are not updated as a result of displacement) toppling failure
cannot be simulated around the evolving crater trough. An algorithm has been
developed that updates the ground surface profile as ore is withdrawn. This
allows additional instability around the crater trough to be predicted and a better
assessment of large-scale fracturing to be predicted for infrastructure stability
assessment.
15.2 Validation
Validation of the numerical model for cave propagation and subsidence

assessment has been completed at four large-scale case study applications that
include the Palabora Mine, the Abandoned Grace Mine, Henderson Mine and
Kiirunavaara Mine.
15.3 Recommendations for Further Work
The following research is recommended to assist in the further development of the

model presented in this thesis.
15.3.1 Rock Mass Behaviour
15.3.1.1 Ubiquitous Joint Rock Mass
Calibration of the UJRM assumes that the SRM testing is an accurate representation
of the rock mass strength and deformation behaviour in the tested loading
directions and sample scales. As changes are made to the SRM technique, a review
of the UJRM methodology is required to ensure that it still provides calibrated
results.
Although the existing UJRM technique has provided good calibrated results for the
case studies undertaken the methodology could be further validated through
additional application.
The tensile strength of the rock mass may currently being over predicted since
only one ubiquitous joint orientation can be specified for each zone. The ability to
define up to three joint orientations for each zone would reduce this mesh
dependency on the tensile strength results.
15.3.1.2 Time-Dependent Processes
Successful cave mining greatly depends on knowledge and understanding of rock

mass behaviour in different stress environments. Time-dependence is one of the
aspects of rock mass behaviour which is not understood completely, but plays an
important role in rock performance. The most important time-dependent
processes that may impact caving behaviour include; ground water and stress-
corrosion. At the present time, these time-dependent processes are not included in
the numerical model of cave propagation and subsidence.
15.3.2 Numerical Techniques
15.3.2.1 Small-Strain Calculation Model
The small-strain calculation mode does not allow the simulation of movement of
material through the cave or along slopes in a state of collapse. The
implementation of some large-strain marker logic would assist in interpreting
these conditions within the model.
15.3.2.2 Interactive Draw
The current algorithm assumes overlapping interactive draw at a minimal height

above the drawpoint. As a result of this, the impacts of isolated draw are not well-
handled. Coupling the large-scale modelling result with a flow program such as
REBOP is recommended to be able to simulate the response of isolated draw.
15.3.2.3 Hardware
Simulation run times are still prohibitive. The development of technology to enable
the model to run on high performance clusters or supercomputers would be
beneficial.
15.3.3 Validation
The application of this numerical technique to additional back-analyses is critical

in establishing confidence in the model predictions and further limitations that
need to be addressed.
16 REFERENCES
Abel, J.F. and Lee, F.T. (1980) Subsidence Potential in Shale and Crystalline Rocks, USGS,
OFR 80-1072.
Adler, L. (1970) Double elasticity in drill cores under flexure. Int. J. Rock Mech. Min. Sci.
Geomech. Abstr. 7, 357-370.
Agarwal, R., Eben, C. and Taylor, C. (1973) Rock mechanics program at Grace Mine,
Technical report No. 3, Henry Krumb School of Mines, Columbia University, New
York, USA, 1973.
Alejano, L. and Alonso, E. (2005) Considerations of the dilatancy angle in rocks and rock
masses. International Journal of Rock Mechanics and Mining Sciences, Vol. 42, pp.
481-507.
Andrieux P. & Hadjigeorgiou, J. (2008). The Destressability Index Methodology for the
Assessment of the Probability of Success of a Large-Scale Confined Destress Blast in
an underground Mine Pillar. International Journal of Rock Mechanics & Mining
Sciences 45 (2008) 407–421.
Anon (1947) Report of the Special Committee on Mining Subsidence, Institution of

Municipal Engineers, London.
Atlas Copco (2011). www.atlascopco.com.au. Accessed 22 November 2011.
Balia, R., Manca P.P. and Massacci G., (1990) Progressive Hangingwall Caving and
Subsidence Prediction at the San Giovanni Mine, Italy, Proc. 9th Int. Conf. On Ground
Control in Mining, ed. S. Peng., pp 303-311.
Bandis, S., Lumsden, A. and Barton, N. (1983) Fundamentals of rock joint deformation.
International Journal of Rock Mechanics and Mining Sciences . 1983, Vol. 20, 6.
Barla, G., Boshkov, S. and Pariseau, W. (1980) Numerical modeling of block caving at the
Grace Mine. Geomechanics applications in underground hardrock mining, Turin,
Italy. pp. 241-256.
Barton, N. and Bandis, S. (1982) Effects of block size on the shear behaviour of jointed
rocks. 23rd US Symposium on Rock Mechanics, Vol. 10, Rotterdam: Balkema,
pp.739-760.
Barton, N. and Pandey, S.K. (2011). Numerical modelling of two stoping methods in two
Indian mines using degradation of c and mobilisation of friction based on Q-
parameters. International Journal of Rock Mechanics and Mining Sciences, 48, pp.
1095-1112.
Basu, D. (1974) Genesis of the Grace Mine magnetite deposit, Morgantown, Berks County,
Southeastern Pennsylvania, PhD thesis, Lehigh University, Bethlehem, PA, USA,
1974.
Beck, D. and Pfitzner, M. (2012) Interaction between deep block caves and existing,
overlying caves or large open pits. Beck Engineering Pty Ltd, accessed 19th March,
212, < http://www.beckengineering.com.au/web/be_downloads.html>.
References
Beck, D., Reusch, F. and Arndt, S. (2007) Estimating the probability of mining-induced
seismic events using mine-scale, inelastic numerical models. Int. Symposium Deep
and High Stress Mining, 2007. The Australian Centre for Geomechanics, Perth,
Australia.
Beck, D., Reusch, F., Arndt, S., Thin, I., Stone, C., Heap, M. and Tyler, D. (2006) Numerical
Modelling of Seismogenic Development during Cave Initiation, Propagation and
Breakthrough. Deep and High Stress Mining 2006.
Beck, D., Sharrock, G. and Capes, G. (2011) A coupled DFE-Newtonian Cellular Automata
scheme for simulation of cave initiation, propagation and induced seismicity. 45th
US Rock Mechanics/Geomechanics Symposium held in San Francisco, CA, June 26-
29.
Besuelle, P. (2001) Evolution of strain localisation with stress in a sandstone : brittle and
semi-brittle regimes. Phys. Chem. Earth (A) Vol 26, No 1-2, pp. 101-106.
Bétournay, M. C., Mitri, H. S. and Hassani, F., (1994) Chimneying disintegration failure
mechanisms of hard rock mines. Rock Mechanics Models and Measurements -
Challenges from Industry, Proceedings 1st North American Rock Mechanics
Symposium, Austin, (Eds: P Nelson and S Laubach), 987-996. Balkema: Rotterdam.
Blodgett, S. (2002) Subsidence Impacts at the Molycorp Molybdenum Mine Questa, New
Mexico, Centre for Science in Public Participation.
Blodgett, S. and Kuipers, J. (2002) Underground Hard-Rock Mining: Subsidence and

Hydrologic Environmental Impacts, Centre for Science in Public Participation.
Board, M. and Pierce, M. (2009). A review of recent experience in modelling of caving.

International Workshop on Numerical Modelling for Underground Mine Excavation
Design, June 28, 2009. Asheville, North Carolina in conjunction with the 43rd US
Rock Mechanics Symposium.
Boyum, B. H. (1961) Subsidence case histories in Michigan mines. Proceedings 4th

Symposium on Rock Mechanics, Bulletin, Mineral Industries Experiment Station,
Pennsylvania State University, No 76: 19-57.
Brady, B. H. G. and Brown, E. T., (1993) Rock Mechanics for Underground Mining, 2nd
edition, 571 p. Chapman and Hall: London.
Brady, B.H.G. and Brown, E.T. (2006) Rock Mechanics For underground mining. Springer;
4th edition. Dordrecht; London: Kluwer Academic Publishers 2004, xviii, 626 p.
Brauner, G. (1973) Subsidence Due to Underground Mining, U.S. Bureau of Mines, IC 8571.
Bray, J. W. (1983) Rankine Lecture.
Brown, E. T. and Ferguson, G. A., (1979) Progressive hangingwall caving at Gath’s mine,
Rhodesia. Trans Instn Min Metall, Sect A: Min Industry, 88: A92-105.
Brown, E.T. (2003) Block Caving Geomechanics, JKMRC Monograph Series in Mining and
Mineral Processing 3. Julius Kruttschnitt Mineral Research Centre, the University of
Queensland: Brisbane. ISBN 1-74112-000-4.
Brown, E.T. and Trollope, D.H. (1970) Strength of a model of jointed rock. Journal of Soil
Mechanics; Foundations Division ASCE. 1970, 96.
References
Brumleve C.B. (1987) Rock Reinforcement of a Caving Block in Variable Ground

Conditions, King Mine, Zimbabwe, Trans. Instn Min. Metall., sect A, April.
Brumleve C.B. and Maier M.M., (1981) Applied Investigations of Rock Mass Response to
Panel Caving, Henderson Mine, Colorado, in Design and Operation of Caving and
Sub-level Stoping Mines, ed R.D. Stewart, AIME, New York, 843 p.
Brummer, R., Li, H. and Moss, A. (2006) The transition from open pit to underground
mining : an unusual slope failure mechanism at Palabora. Proceedings International
Sympoisum on Stability of Rock Slope in Open Pit Mining and Civil Engieering. The
South African Institute of Mining and Metallurgy. pp. 411-420.
Brunton, I. (2009). Personal Communication.
Butcher, R. (2005) Subsidence effects associated with block and sub-level caving of
massive orebodies, Australian Centre for Geomechanics Newsletter, Volume No. 25
Butcher, R.J. (2005) Subsidence effects associated with block and sub-level caving of
massive orebodies, Australian Centre for Geomechanics Newsletter, Volume No. 25.
Cai, M. and Horii, H. (1993) A constitutive model and FEM analysis of jointed rock masses.
International Journal of Rock Mechanics and Mining Sciences. 1993, Vol. 30.
Cai, M. Kaiser, P.K., Tasaka, Y. and Minami, M. (2007) Determination of residual strength
parameters of jointed rock masses using the GSI system. International Journal of
Rock Mechanics and Mining Sciences 44 (2007) 247–265.
Cai. M. (2010) Practical estimates of tensile strength and Hoek-Brown strength parameter
mi of brittle rocks. Rock Mechanics and Rock Engineering, 2010, Vol. 43 pp.167-184.
Carlson, G. and Golden Jr. R. (2008). Initiation, growth, monitoring and management of the
7210 cave at Henderson Mine – A Case Study in Proceedings of the 5th International
Conference and Exhibition on Mass Mining, Lulea Sweden 9-11 June, 2008. Lulea
University of Technology Press, Lulea, Sweden.
Cavieres, P., S. Gaete, L. Lorig and P. Gómez. (2003) Three-Dimensional Analysis of

Fracturing Limits Induced by Large Scale Underground Mining at El Teniente Mine,
in Soil and Rock America 2003 (Proceedings of the 39th U.S. Rock Mechanics
Symposium, Cambridge, Massachusetts, June 2003), pp. 893-900. P. J. Culligan, H. H.
Einstein and A. J. Whittle, Eds. Essen: Verlag Glückauf.
Chappell, B. (1975) Component characteristics of jointed rock masses. International

Journal of Rock Mechanics, Mining Sciences and Geomechanics. 1975, Vol. 12.
Chen, R. and Stimpson, B. (1993) Interpretation of indirect tensile strength tests when
moduli of deformation in compression and in tension are different. Rock Mechanics
and Rock Engineering. Volume 26, Number 2, 183-189, DOI: 10.1007/BF01023622.
Clark, I.H. (2006) Simulation of Rock mass Strength Using Ubiquitous Joints. In: R. Hart and
P. Varona (eds), Numerical Modeling in Geomechanics — 2006; Proc. 4th
International FLAC Symposium, Madrid, May 2006. Paper No. 08-07, Minneapolis:
Itasca.
Clough, R.W. (1960) The finite element method in plane stress analysis. Proceedings of the
Second ASCE Conference on Electronic Computation, Pittsburgh, PA.
Collins, P.J. (1977) Measurement and Analysis of Residual Mining Subsidence Movements,
Large Ground Movements and Structures, in Proceedings of the Conference at
References
University of Wales, Institute of Science and Technology, Cardiff, Wales, Wiley and
Sons, New York.
Crane W.R. (1929) Subsidence and Ground Movement the Copper and Iron Mines of the
Upper Peninsula, Michigan. USBM Bulletin 285, 66 pp.
Crane W.R. (1931) Essential Factors Influencing Subsidence and Ground Movement, USBM
IC 6501 14 p.
Crook, T., Willson, S., Yu, J.G. and Owen, R. (2003) Computational modeling of the localized
defor-mation associated with borehole breakout in quasi-brittle materials. Journal
of Petroleum Science and Engineering, 38: 177-186.
Cundall, P. (1991) Shear band initiation and evolution in frictional materials. Mechanics
Computing in 1990's and Beyond. Proceedings of the Conference, Columbus, Ohio,
May 1991. Structural and Material Mechanics, 1991, Vol. 2.
Cundall, P. (2008) Recent advances in numerical modelling for large-scale mining projects,
ACG News., 30, 1-7 (June 2008).
Cundall, P., Gomez, P., Lorig, L. Pierce, M., Sainsbury, D., Young, P. and Napier, P. (2005)
Caving Mechanics Task 1 Scoping Study. Report to the International Caving Study,
ICG052292-1-8.
Dawson, E. and Cundall, P. (1995) Cosserat plasticity for modelling layered rock.
Fractured and Jointed Rock Masses, Myer, Cook, Goodman & Tsang (eds). 1995
Balkema, Rotterdam. ISBN 90 5410 591 7. pp. 267-274.
De Nicola Escobar, R. and Fishwick Tapia, M. (2000) An underground air blast – CODELCO
Chile – Division Salvador. Proceedings, MassMin 2000. The Australian Institute of
Mining and Metallurgy. Publication Series No. 7/2000, p. 279-288.
Deisman, N., Mas Ivars, D. and Pierce, M.E. (2008) PFC2D Smooth joint contact model
numerical experiments, In GeoEdmonton '08: A Heritage of Innovation, Proc. 61st
Canadian Geotechnical Conference & 9th Joint CGS/IAH-CNC Groundwater
Conference (Edmonton, September 21-24, 2008).
Detournay, C. (2009) Personal Communication.
Detournay, E. (1986) Elastoplastic model of a deep tunnel for a rock with variable
dilatancy. Rock Mechanics and Rock Engineering 19, 99-108.
Dickhout M.H., (1963) Ground Control at the Creighton Mine of the International Nickel
Company of Canada Limited, Proc. Rock Mechanics Symposium, McGill University,
Mines Branch, Ottawa.
Diederichs, M. (1999) Instability of Hard Rock masses : The Role of Tensile Damage and
Relaxation. PhD Thesis, University of Waterloo, Ontario, Canada.
Diederichs, M. S. (2002): Keynote: Stress induced accumulation and implications for hard
rock engineering. In: Hammah, R., Bawden, W., Curran, J., Telesnicki, M. (eds.) Proc.,
NARMS 2002, Toronto, University of Toronto Press, 3–14.
Diederichs, M.S. (2007) The 2003 Canadian Geotechnical Colloqium: Mechanistic

inpterpretaiton and practical application of damage and spalling prediction criteria
for deep tunnelling. Canadian Geotechnical Journal. 44, 1082-1116.
References
Diering, J.A.C. and Laubscher, D.H. (1987) Practical approach to the numerical stress
analysis of mass mining. Transactions of the Institution of Mining and Metallurgy,
Section A: Minerals Industry, 96: A179-A188.
Duncan-Fama M.E. (2003) Numerical modelling of yield zones in weak rock. In: Hudson J,
editor. Comprehensive rock engineering, vol. 2. Oxford: Pergamon Press; 1993. p.
49–75.
Duplancic, P. and Brady, B.H. (1999) Characterization of caving mechanisms by analysis of

seismicity and rock stress, Proceedings 9th International Congress on Rock
Mechanics (Paris) Vol. 2, pp. 1049–1053. Balkema, Rotterdam.
Eberhardt, E., Kaiser, P.K., Stead, D, (2002) Numerical analysis of progressive failure in
natural rock slopes. In: da Gama, Dinis, eSousa, Ribeiro (Eds.), EUROCK 2002 —
Proc. ISRM Int. Symp.on Rock Eng. for Mountainous Regions, Funchal, Madeira.
Sociedade Portuguesa de Geotecnia, Lisboa, pp. 145–153.
Eberthard, E. (2011) Impacts of Underground Mass Mining on Surface: Lessons Learned

from the Palabora Cave-Pit Interaction Study. CEMI Lecture Series, March 31, 2011.
Energy Information Administration (1995) Longwall Mining. DOE/EIA-TR-0688

Distribution Category UC-950. Office of Coal, Nuclear, Electric and Alternate Fuels,
U.S. Department of Energy, March 1995.
Ettema, R. and Urroz, G. E. (1989). On internal friction and cohesion in unconsolidated ice
rubble. Cold Regions Science and Technology, 16, pp. 237-247. Elsevier Science
Publishers B.V., Amsterdam.
Fairhurst, C. (1961) Laboratory measurement of some physical properties of rock.

Proceedings 4th Symposium of Rock Mechanics, 105-118.
Fletcher J.B. (1960) Ground Movement and Subsidence from Block Caving at Miami Mine,
AIME Transactions, v. 217, pp 413-421.
Flores, G and Karzulovic, A, (2004a) Surface Crown Pillar Guidelines. Prepared for
International Caving Study, Stage II, JKMRC: Brisbane.
Flores, G and Karzulovic, A, (2004b) Geotechnical Guidelines Subsidence. Prepared for

International Caving Study, Stage II, JKMRC: Brisbane.
Fossum, A. (1985) Effective elastic properties for a randomly jointed rock mass.
International Journal of Rock Mechanics and Mining Sciences. 1985, Vol. 22, 6.
GEMCOM (2012) PCBC Software. Gemcom Software International Inc. Vancouver, B.C.
Canada .
Gerrard, C.M. (1982) Elastic models of rock mass having one, two and three sets of joints.
International Journal of Rock Mechanics and Mineral Sciences. 1982, Vol. 19.
Gilbride, J. Free, K. S. and Kehrman, R (2005) Modeling Block Cave Subsidence at the
Molycorp, Inc., Questa Mine, Alaska Rocks 2005, The 40th U.S. Symposium on Rock
Mechanics (USRMS), June 25 - 29, 2005 , Anchorage, AK
Glazer, S. and Hepworth, N. (2004) Seismic Monitoring of Block Cave Crown Pillar.
Proceedings MassMin 2004. Santiago.
References
Glazer, S. (2006). Applications of Mines Seismology Methods in Block Cave Mining.

Proceedings 1st International Symposium on Block and Sub-Level Caving. The
South African Institute of Mining and Metallurgy. pp.281-302.
Goodman, R.E. & Bray, J.W. 1976, Toppling of rock slopes. Rock engineering for
foundations and slopes; proceedings of a specialty conference, Vol. 2 p201-233, Am.
Soc. Civ. Eng.. New York.
Goodman, T. (1970) Internal USBM monthly subsidence monitoring report.
Grard, C. (1969) Mining Subsidence and the Means Permitting the Limiting of their Effects
on the Surface (in French), Revue de l’Industrie Mineral, 51, (January).
Gray, R.E., Bruhn, R.W. and Turka, R.J. (1977) Study and Analysis of Surface Subsidence
Over Mined Pittsburgh Coalbed, U.S. Bureau of Mines, J0366047.
Guest, A. (2009) Personal Communication.
Haimson, B. C. and Tharp, T., Real stresses around boreholes, Soc. Petrol. Engr. J., 14, 145–
151, 1974.
Hajiabdolmajid, V. and Kaiser, P. K. (2003) Brittleness of rock and stability assessment in

hard rock tunnelling. Tunnelling and Underground Space Technology, 18 (1): 35-48.
Hajiabdolmajid, V., Kaiser, P. K., and Martin, C. D. (2002) Modelling brittle failure of rock.
International Journal of Rock Mechanics and Mining Sciences, 39 (6): 731-741
Hatheway, A.W., (1966) Engineering geology of subsidence at San Manuel mine, Pinal
County, Arizona: M.S. thesis, University of Arizona, Tucson, 110 pp.
Hebblewhite, B. (1995) Ground Control Newsletter. The Department of Mining

Engineering, The University of New South Wales, Sydney, Australia.
Hellewell, F.G. (1988) The Influence of Faulting on Ground Movement due to Coal Mining:
The U.K. and European Experience. Mining Engineer, 147, pp 334-337.
Herdocia, A, (1991) Hanging wall stability of sub-level caving mines in Sweden. PhD
thesis (unpublished), Luleå University of Technology, Luleå, Sweden.
Heslop T.G. (1974) Failure by Overturning in Ground Adjacent to Cave Mining at Havelock
Mine, 3rd. ISRM Congress, Denver.
Heslop T.G. (1984) The Application of Interactive Draw Theory to Draw Control Practice in
Large Chrysotile Asbestos Mines, Chamber of Mines Journal (S.A.) June 1984, pp 23-
41.
Hill, R. (1950) The mathematical theory of plasticity. Oxford:ClarendonPress;1950.
Hoek E. and Martin, D. (2010) Synthetic Rock Mass Review. Confidential Report
Submitted to Dr Gideon Chitombo, MMT2 Technical Director, The Sustainable
Minerals Institute. The University of Queensland, Brisbane, Australia.
Hoek E., Kaiser P.K. and Bawden W.F. (1995) Support of underground excavations in hard
rock. p. 215. Rotterdam, Balkema.
Hoek, E. (1974) Progressive caving induced by mining an inclined orebody. Trans Inst.
Min Metall., Sect A: Min Industry, 83: A133-139.
References
Hoek, E. and Bray, J.W. (1981) Rock Slope Engineering. 2nd Edition. Institute of Mining
and Metallurgy, London, 358 pp.
Hoek, E. and Brown, E T. (1997) Practical estimates of rock mass strength. International
Journal of Rock Mechanics and Mineral Sciences. 1997, Vol. 34, 8.
Hoek, E. and Brown, E.T. (1980) Underground excavations in rock. London : Institution of
Mining and Metallurgy, 1980.
Hoek, E. and Diederichs, M.S. (2006) Empirical estimation of rock mass modulus,
International Journal of Rock Mechanics and Mining Sciences, Vol 43, pp. 203-215.
Hoek, E. and Karzulovic, A. (2000) Rock-Mass properties for surface mines. In Slope
Stability in Surface Mining (Edited by W. A. Hustralid, M.K. McCarter and D.J.A. van
Zyl), Littleton, CO: Society for Mining, Metallurgical and Exploration (SME), pages
59-70.
Holla, L. and Buizen, M., (1990) Strata Movement Due to Shallow Longwall Mining and the
Effect on Ground Permeability. Proc. Aus. IMM, No. 1, pp 11-18.
Hood, M., Ewy, R.T. and Riddle, L.R. (1981) Empirical Methods of Subsidence Prediction —
A Case Study, Workshop on Surface Subsidence Due to Underground Mining, West
Virginia University, Morgantown.
Hudson, J.A., Crouch, S.L and Fairhurst, C. (1971) Soft, stiff and servo-controlled testing
machines: a review with reference to rock failure. Engineering Geology Vol 6, Issue
3, October 1972, pp. 155-189.
Hughes, H.W. (1917) A text-Book of Coal Mining. 6th Edition. London: Charles Griffen &
Co, Ltd. P. 200.
Hutchinson, D.J. and Diederichs, M.S. (1996) Cablebolting in Underground Mines. Bitech
Publishers Ltd., Vancouver. 416p.
Itasca Consulting Group, Inc. (1995) PFC – Particle Flow Code in 2 Dimensions, Ver. 1.1,
User’s Manual. Minneapolis: Itasca.
Itasca Consulting Group, Inc. (2000) FLAC – Fast Lagrangian Analysis of Continua, Ver. 4.0,
Itasca Consulting Group, Inc. (2005) FLAC – Fast Lagrangian Analysis of Continua, Ver. 5.0,
Itasca Consulting Group, Inc. (2007) PFC3D – Particle Flow Code in 3 Dimensions, Ver. 4.0.
User's Manual. Minneapolis: Itasca.
Itasca Consulting Group, Inc. (2009) FLAC3D – Fast Lagrangian Analysis of Continua in 3
Dimensions, Ver. 4.0. User's Manual. Minneapolis: Itasca.
Itasca Consulting Group, Inc. (2012) REBOP – Rapid Emulator Based on the Particle Flow
Code. User's Manual. Minneapolis: Itasca.
Janelid, I. (1972), Study of the gravity flow process in sub-level caving. International Sub-
level Caving Symposium, Stockholm, September.
Johnsone G.H. and Soule J.H., (1963) Measurements of Surface Subsidence . San Manuel
Mine, Pinal County, Arizona, USBM R.I. 6204.
References
Ju, J.W. and Chen, T.M. (1994) Micromechanics and effective moduli of elastic composites
containing randomly dispersed ellipsoidal inhomogeneities. Acta Mechanica, 103,
103-144.
Kakavas, P.A. and Anifantis, N.K. (2003) Effective moduli of hyperelastic porous media at
large deformation. Acta Mechanica 160, 127–147 (2003) DOI 10.1007/s00707-002-
0982-1
Kantner, W.H., (1934) Surface Subsidence Over Porphyry Caving Blocks, Phelps Dodge
Corporation, Copper Queen Branch, AIME Transactions V. 109, pp 181-194.
Karekal, S., Das, R., Losse, L. and Cleary, P.W (2011). Application of a mesh-free continuum
method for simulation of rock caving processes. International Journal of Rock
Mechanics and Mining Sciences, 48(5): 703-711.
Karzulovic, A., (1990) Evaluation of angle of break to define the subsidence crater of Rio
Blanco Mine’s Panel III (in Spanish). Technical Report, Andina Division, CODELCO-
Chile.
Karzulovic, A., Cavieres, P. and Pardo, C., (1999) Caving subsidence at El Teniente Mine (in
Spanish). Pro-ceedings Symposium on Mining Engineering, SIMIN 99, Santiago.
Department of Mining Engineering, University of Santiago: Santiago.
Karzulovic, A. and Flores, H. (2003) Geotechnical guidelines for a transition from open pit
to under-ground mining. Report to International Caving Study, July, 2003.
Khan, A.S. and Yuan, S. (1988) A three-dimensional finite element program for brittle
bimodular rock-like material. Int. J. Numer. Anal. Methods Geomech. 12, 599-609.
Kratzsch, H., (1983) Mining Subsidence Engineering, 543 p. Springer-Verlag: Berlin.
Kubrix (Simulation Works, 2012)
Kvapil R., (1982) The Mechanics and Design of Sub Level Caving Systems, Underground
Mining Methods Handbook, Am. Inst. of Min, Metall. and Pet. Eng.,pp 880-897.
Kvapil, R., Ceccarelli, B. and Lonergan, J., (1989) Quantitative Analysis of Subsidence at El
Teniente Mine. Technical Report, El Teniente Division, CODELCO-Chile.
Laubscher, D. (1990) A geomechanical classification system for the rating of rock mass in
mine de-sign. Journal of the South African Institute of Mining and Metallurgy. Vol.
90, No. 10, pp. 257-273.
Laubscher, D.H. (1975). Class distinction in rock masses. Coal, Gold, Base Metals, South
Africa. Vol. 23, August.
Laubscher, D.H. (1994) Cave mining – the state of the art. Journal of the South African
Institute of Mining and Metallurgy, October 1994, pp. 279-293.
Laubscher, D.H. (2000) Block Caving Manual. Prepared for International Caving Study.
JKMRC and Itasca Consulting Group, Brisbane.
Laubscher, D.H., (1990) A geomechanics classification system for the rating of rock mass in
mine design. South African Institute for Mining and Metallurgy 90 (10), 257–273.
Li, A.J., Merifield, R.S. and Lyamin, A.V. (2008) Stability charts for rock slopes base don the
Hoek-Brown failure criterion. International Journal of Rock Mechanics & Mining
Sciences 45 (2008) 689-700.
References
Liang, Z.Z., Tang, C.A., Li, H.X., Xu, T. and Zhang, Y.B. (2004) Numerical simulation of 3-d
failure process in hetrogeneous rocks. International Journal of Rock Mechanics and
Mining Sciences. SINOROCK2004, 2004, Vol. 41, 3.
Liu, Y.A. (1981) Novel High Gradient Magnetic Separation (HGMS) Processes for
Desulfurization of Dry Pulverized Coal, Chapter 9, in Recent Developments in
Separation Science, Volume VI, pp. 149-171, Normal N. Li, editor, CRC Press, Boca
Ratan, FL.
Lorig, L., Board, M., Potyondy, D. and Coetzee, M. (1995) Numerical modelling of caving
using continuum and micro-mechanical models. CAMI'95:3rd Canadian Conference
on Computer Applications in the Mineral Industry. Montreal, Quebec, Canada,
October 22-25, 1995, pp. 416-425.
Lorig, L. (2000) The Roel of Numerical Modelling in Assessing Caveabilty, Itasca

Consulting Group Inc., Report to the International Caving Study, ICG00-099-3-16,
October 2000.
Luo, Y. and Peng, S.S. (2000) Long-Term Subsidence Associated with Longwall Mining —
Its Causes, Development and Magnitude, Mining Engineering, 52(10), (October).
Lupo, J.F. (1996) Evaluation of Deformations Resulting from Mass Mining of an Inclined
Orebody. Ph.D. thesis (unpublished), Colorado School of Mines, Golden, Colorado.
Lupo, J.F. (1997). Progressive Failure of Hangingwall and Footwall at the Kiirunavaara
Mine, Sweden. Int. J. Rock Mech. and Min. Sci. 34:3-4, paper No. 184.
Lupo, J.F. (1998) Large-scale surface disturbances resulting from underground mass
mining, Int. J. Rock Mech. Min. Sci., 35(4-5), Paper No. 25.
Mahtab M., (1976) Influence of Rock Fractures on Caving, Trans. AIME, Vol 260.
Marachi, N., Chan, C. and Seed, H. (1972) Evaluation of properties of rockfill materials.
Journal Soil Mechanics, Div ASCE. 1972, 98.
Mas Ivars, D., Deisman, N., Pierce, M. and Fairhurst, C. (2007) The Synthetic Rock Mass
approach - A step forward in the characterization of jointed rock masses. 11th
Congress of the International Society for Rock Mechanics. 2007, 485-490.
Mas Ivars, D., Pierce, M., Darcel, C., Reyes-Montes, J., Potyondy, D., Young, P. and Cundall, P.
(2011) The synthetic rock mass approach for jointed rock mass modelling.
International Journal of Rock Mechanics & Mining Sciences 48 (2011) 219–244.
Mas Ivars, D., Pierce, M., DeGagné, D. and Darcel, C. (2008) Anisotropy and scale
dependency in jointed rock mass strength – A synthetic rock mass study. In R. Hart,
C. Detournay and P. Cundall (eds), Proceedings of the First International FLAC/DEM
Symposium on Numerical Modeling, August 25-27, 2008, Minneapolis, MN USA.
Minneapolis: Itasca.
Mathews, K.E, Hoek, E., Wyllie, D.C and Stewart, S. (1981) Prediction of stable excavation
spans for mining at depths below 1000m in hard rock. Report No. DSS Serial
No.OSQ80-00081, DSS File No. 17SQ.233440-0-9020, 43 p.
Medhurst, T.P. (1996) Estimation of the in situ strength and deformation of coal for
engineering design. PhD Thesis. University of Queensland.
References
Milne, D., Hadjigeorgiou, J. and Pakalnis, R. (1998) Rock mass characterization for
underground hard rock mines. Tunnelling and Underground Space Technology
13(4):383-391.
National Coal Board, (1975) Subsidence Engineers Handbook, 2nd edition. National Coal
Board Mining Dept, London.
Nelson W.I. and Fahrni K.C., (1950) Caving and Subsidence at the Copper Mountain Mine,
CIM Transactions Bulletin, v. 43, pp 2-10.
Nicieza, C.G., A´lvarez Fernańdez, M.I., Menendez Dıáz, A., A´lvarez Vigil, A.E. (2006)
Modification of rock failure criteria considering the RMR caused by joints.
Computers and Geotechnics 33 (2006) 419–431.
Nielsen, L. (1990) Strength and stiffness of porous materials. Journal American Ceramics
Society. Vol 73, No. 9, pp. 2684-2689.
Obert L. and Long A.E., (1962) Underground Borate Mining, Kern Country, California
,USBM R.I.
Obert, L. and Duvall, W. I., (1967) Rock Mechanics and the Design of Structures in Rock,
650 p. J Wiley & Sons: New York.
Orchard, R.J. and Allen, W.S. (1974) Time-dependence in Mining Subsidence, Proceedings
International Symposium on Minerals and the Environment, Institution of Mining
and Metallurgy, London, pp. 643-659.
Palma, R. and Agarwal, R. (1973) A Study of the Caveability of Primary Ore at the El
Teniente Mine, Technical Report from Colombia University, NY.
Panek, L.A. (1984) Subsidence in Undercut-Cave Operations, Subsidence Resulting from

Limited Ex-traction of Two Neighboring-Cave Operations, Department of Mining
Engineering, Michigan Technological University, Houghton, Michigan.
Pappas, D. and Mark, C. (1993) Behaviour of Simulated Longwall Gob Material. Report of
Investigations 9458. United States Department of the Interior, Bureau of Mines.
Parker J., (1978) What Can Be Learned From Surface Subsidence, E&MJ, July 1978.
Passaris, E.K.S. (1977) The effect of directional anisotropy on the mechanical behaviour of
rocksalt. Proceedings, Conference Rock Engineering, British Geotechnical Society,
Department of Mining Engineering. Newcastle upon Tyne, U.K., 11-31.
Peele, R. (1918), Mining Engineers’ Handbook. John Wiley & Sons, New York.
Peng, S.S. (1992) Surface Subsidence Engineering, Society For Mining, Metallurgy and
Exploration Inc. , Littleton, Colorado. 161 pp.
Perman, F. and Sjoberg, J. (2011) Numerisk analys av brytningssekvenser i block 19.

LKAB Utredning 11-776, 2011-05-20 (in Swedish).
Phillips, K.A.S. and Hellewell, F.G. (1994) Three-dimensional ground movements in the
vicinity of a mining activated geological fault. Quarterly Journal Engineering
Geology, 27, 7-14.
Pierce, M., Cundall, P., Potyondy, D. and Mas Ivars, D. (2007) A synthetic rock mass model
for jointed rock. Rock Mechanics: Meeting Society's Challenges and Demands –
References
Eberhart, Stead and Morrison (eds) 2007 Taylor and Francis Group, London, ISBN
978-0-415-44401-9, 341-349.
Pierce, M. (1999) Stress Changes resulting From Plug Caving of Lift 1 at Northparkes Mine.
Itasca report to Northparkes Mine, ICG99-1043-14F (Confidential)
Pierce, M. (2010) Personal Communication.
Pierce, M. and Lorig, L. (1998) FLAC3D Analysis of Caveability of the Northparkes E26 Lift 2
Orebody, Itasca Consulting Group, Inc. Report to Northparkes Mines, Parkes, NSW,
Australia # 1017, November, 1998.
Pierce, M., Cundall, P. Mas Ivars, D., Darcel, C., Young, R.P., Reyes-Montes, J. and Pettitt, W.
(2006) Six Monthly Technical Report, Caving Mechanics, Sub-Project No. 4.2:
Research and Methodology Improvement, and Sub-Project 4.3, Case Study
Application, Itasca Consulting Group Inc, Report to Mass Mining Technology Project,
2004-2007, ICG06-2292-1-Tasks 2-3-14, March.
Pierce, M., Cundall, P., Potyondy, D. and Mas Ivars, D. (2007) A synthetic rock mass model
for jointed rock, In Rock Mechanics: Meeting Society's Challenges and Demands,
Proc. 1st Canada-U.S. Rock Mechanics Symposium Vol. 1, pp. 341-349
Pierce, M., Gaida, M. and Degagne, D. (2009) Estimation of rock block strength.
ROCKENG09. Proceedings 3rd CANUS Rock Mechanics Symposium, Toronto, May,
2009. Paper No. 4360. M. Diederichs and G. Grasselli, eds.
Pierce, M., Mas Ivars, D. and Sainsbury, B. (2009). Use of Synthetic Rock Masses (SRM) to
investigate jointed rock mass strength and deformation behaviour in Proceedings of
the International Conference on Rock Joints and Jointed Rock Masses in Tucson,
Arizona. May 14 and 15, 2009.
Pine, R.J. Owen, D.R.J., Coggan, J.S. and Rance, J.M. (2007) A new discrete fracture
modelling approach for rock masses. Geótechnique 57, No. 9, 757–766 [doi:
10.1680/geot.2007.57.9.757]
Portu, G.D. and Vincenzini, P. (1981) XIII Siliconf. Vol. 1 (1981) Budapest, Hungary, p. 265.
Pratt, H.R., Black, A.D., Brown, W.S., Brace, W.F. (1972) The effect of specimen size on the
mechanical properties of unjointed diorite. Int. J. Rock Mech. Min. Sci. 9, 513–529.
Rech, W.D, 2001. Underground Mining Methods. Engineering Fundamentals and

International Case Studies, Chapter 48 - Henderson Mine. Hustrulid, W. and R.
Bullock (eds). Society for Mining and Exploration.Rech, W. and Lorig, L. (1992)
Predictive numerical stress analysis of panel caving at the Henderson Mine.
MASSMIN’92, Johannesburg, SAIMM, 1992, pp 55-62.
Rech, W.D. and Lorig, L. (1992) Predictive numerical stress analysis of panel caving at the
Henderson Mine. MASSMIN ’92. Pp 55-62, Johannesburg, SAIMM, 1992.
Ribacchi, R. (2000) Mechanical Tests on Pervasively Jointed Rock Material: Insight into
Rock Mass Behaviour. Rock Mech. Rock Engng. (2000) 33 (4), 243-266.
Rice, G. (1934) Ground movement from mining in Brier Hill mine, Norway, Michigan.
Mining and Metallurgy, v 15, n 325, p.12-14.
Rice, R. (1976) Extension of the Exponential Porosity Dependence of Strength and Elastic
Moduli. Journal of The American Ceramic Society-Discussions and Notes Vol. 59, No.
11- 1, pp 536-537.
References
Rockfield (2007) ELFEN. User Guide. Technium, Kings Road, Prince of Wales Dock,
Swansea, SA1 8PH, West Glamorgan, U.K.
Rocscience (2002) PHASE2. Finite element analysis and support design for excavations.
Toronto: Rocscience, Inc.
Rocscience (2012). RocLab Users Guide,

http://www.rocscience.com/products/14/RocLab. Accessed January 30, 2012.
Rogers, S., Elmo, D., Webb, G. and Catalan, A. (2010) Simulating the impacts of hydraulic
fracture preconditioning on cavability and fragmentation at the planned Cadia East
panel cave in Proceedings of the Second International Symposium on Block and Sub-
level Caving (Perth, April 2010). Y. Potvin, Ed. Australian Centre for Geomechanics,
pp. 663-676.
Rojas, E., Molina, R and Cavieres, P. (2001) Preundercut caving in El Teniente, Chile.
Underground Mining Methods: Engineering Fundamentals and International Case
Studies, SME, Colorado.
Ross, I. and van As, A. (2005) Northparkes Mines – Design, Sudden Failure, Air-Blast and
Hazard Management at the E26 Block Cave. Ninth Underground Operators
Conference, Perth, WA, 7-9 March, 2005.
Rudnicki, J.W. and Rice, J.R. (1975) Conditiosn for the locatlization of deformation in
pressure-senstivie dilatant materials. J. Mech. Phys. Solids, 1975, Vol 23, pp 371-
394.
Russo, G., Kalamaras, G.S. and Grasso, P.A. (1998) Discussion on the concepts of
geomechnical classes behaviour categoris and technical classes for an underground
project. Gallerie e Grandi Opere Sotterranee 1998; 54.
Sainsbury B., Mas Ivars, D. and Darcel C. (2009). Synthetic Rock Mass Simulations of the
1200, 1700 and 1800 Domains at the Mt Keith Open Pit. Itasca Consulting Group.
Technical report number 08025.
Sainsbury, B. (2010) Sensitivities in the numerical assessment of cave propagation in

Caving 2010: Second International Symposium on Block and Sub-level Caving. 20–
22 April 2010, Australian Centre for Geomechanics.
Sainsbury, B., Mas Ivars, D. and Darcel, C. (2008) Synthetic Rock Mass Simulations of the
1200, 1700 and 1800 Domains at the Mt Keith Open Pit. Itasca Australia Report
08025 to BHP Billiton Nickel West.
Sainsbury, B., Pierce, M. and Mas Ivars, D. (2008) Analysis of Caving Behaviour Using a
Synthetic Rock Mass —Ubiquitous Joint Rock Mass Modelling Technique, In
Proceedings of the 1st Southern Hemisphere International Rock Mechanics
Symposium (SHIRMS), Y. Potvin, J. Carter, A. Dyskin and R. Jeffrey (eds), 16–19
September 2009, Perth, Australia, Australian Centre for Geomechanics, Perth, Vol. 1
– Mining and Civil, pp. 243–254.
Sainsbury, D. and Lorig, L. (2005) Caving Induced Subsidence at the Abandoned Grace
Mine. ICG Report to Arcadia Land Company.
Sainsbury, D., Board, M. and Sainsbury, B. (2010) Analysis of Caving Behaviour and
Extraction Level Performance at the Cadia East Underground Project – 2010
Feasibility Study. Itasca Australia Pty Ltd Report No. 10001D to Newcrest Mining
Limited.
References
Sainsbury, D., Lorig, L. and Sainsbury, B. (2010) Investigation of caving induced subsidence
at the abandoned Grace Mine, in Caving 2010: Second International Symposium on
Block and Sub-level Caving. 20–22 April 2010, Australian Centre for Geomechanics.
Sainsbury, D. and Sainsbury, B. (2010) A Review of Caving Induced Surface Subsidence.

Itasca Australia Report # 10007 to CEMI.
Salamon, M. (1968) Elastic moduli of a stratified rock mass. International Journal of Rock
Mechanics and Mining Sciences. 1968, Vol. 5.
Samadhiya, N, Viladkar,m and Singh, B. (2004) Three-dimensional analysis of a

powerhouse cavern in an anisotropic rock mass. International Journal of Rock
Mechanics and Mining Sciences. SINOROCK2004, 2004, Vol. 41, 3.
Sandström, D. (2003) Analysis of the Virgin State of Stress at the Kiirunavaara Mine.
Licentiate thesis 2003:02, Luleå University of Technology, Luleå, Sweden.
Santarelli, F.J. (1989) Theoretical and experimental investigation of the stability of the
axisymmetric wellbore. PhD thesis, Imperial College, London.
Schmertmann, J.H. and Osterberg, J.O. (1960) An experimental study of the development
of cohesion and friction with axial strain in saturated cohesive soils. Proc. ASCE
Research Conference on the Shear Strength of Cohesive Soils, Boulder, Colorado,
19060, pp 643-694.
Serafirm, J.L. and Pereira, J.P. (1983) Considerations of the geomechanical classification of
Bieniawski. Proceedings, International Symposium on Engineering Geology and
Underground Construction, Lisbon, Vol 2, pp II-33 to II-42, LNEC, Lisbon.
Shadbolt C.H. (1978) Mining Subsidence - Historical Review and State of the Art, Conf. on
Large Ground Movements and Structures, Cardiff. Pantech Press, London, ed. J.D.
Geddes. pp 705-748.
Shadbolt C.H., (1987) A Study of The Effects of Geology On Mining Subsidence in the East
Pennine Coalfield. Ph.D. Theses, University of Nottingham.
Shadrin, A.G. and Zomotin, V.B. (1977) Calculating the Durations of Surface Movements (in
Russian), Fiziko-Tekhnicheskie Problemy Razrabotki Poleznykh Isopaemykh, No. 6.
Sharrock, G. (2010). Personal communication.
Sharrock, G., Vakili, A., Duplancic, P. and Hastings, N. (2011) Numerical analysis of
subsidence for Perserverence Deeps Block Cave in Continuum and Distinct Element
Numerical Modelling in Geomechanics, 2011. Sainsbury, Hart, Detournay and
Nelson (eds.), Paper 06-03, Itasca International Inc, Minneapolis, ISBN 978-0-
9767577-2-6.
Sims, S.J. (1968) The Grace Mine magnetite deposit, Berks County, Pennsylvania, in Ore
deposits of the United States, 1933–1967: the Graton-Sales volume (ed. J. D. Ridge),
Vol. 1, 108–124; 1968, New York, AIME.
Simulation Works (2012). Kubrix. Minneapolis, MN, USA.
Simulia (2011) ABAQUS, Dassault Systèmes, 2004, 2012 - All Rights Reserved.
Singh, B. (1973) Continuum characterization of jointed rock masses Part I - The

constitutive equations. International Journal of Rock Mechanics and Mineral
Sciences. 1973, Vol. 10.
References
Singh, M.M., (2003) Mine Subsidence. SME Mining Engineering Handbook (3rd ed.).
Littleton,CO: Society for Mining Metallurgy and Exploration, Inc.
Singh, U K, Stephansson, O J and Herdocia, A, (1993) Simulation of progressive failure in

hanging-wall and footwall for mining with sub-level caving. Trans Instn Min Metall,
Sect A: Min Industry, 102: A188-194.
Sitharam, T.G. and Latha, G.M. (2002) Simulation of excavations in jointed rock masses
using a practical equivalent continuum approach. International Journal of Rock
Mechanics and Mining Sciences, Vol. 39, No. 4, pp. 517-525.
Sluys, L.J. (1992) Wave Propagation, Localization and Dispersion in Softening Solids. Ph.D
Thesis, Delft University Technology, Netherlands.
Spinner, S., Knudsen, F.P. and Stone, L. (1963) J. Res. Natl. Bur. Stand. 67C, 39.
Stacey, T.R. and Swart, A.H. (2001) Practical rock engineering practice for practice for
shallow and opencast mines. SIMRAC The safety of mines research advisory
committee, 66pp.
Stafford, B. (2002) The Grace Mine story. Bee Line, 2002, 24, (3–4), 55.
Stephansson, O., Borg, T. and Bäckblom, G. (1978) Sprickbildning i norra Kiirunavaaras

hängvägg. Teknisk rapport 1978:51T, Avdelningen för Bergmekanik, Högskolan i
Luleå.
Sterpi, D. (1999) An analysis of geotechnical problems involving strain-softening effects.

Int. J. Numer. Anal. Meth. Geomech. 1999;23:1427–54.
Stewart, D., Rein, R. and Firewick, D. (1984) Surface Subsidence at the Henderson Mine, in
Geomechanics Applications in Underground Hardrock Mining, SME/AIME, New
York.
Stöckel, B-M., Sjoberg, J., Makitaavola, K. and Savilahti, T. (2012) Mining-induced ground
eformations in Kiirunavaaraand Malmberget, EUROCK 2012, 28-30 May, Stockholm,
Sweden. (in press).
Sundaram, R.N. and Corrales, J.M. (1980) Brazilian tensile strength of rocks with different
elastic properties in tension and compression. International Journal Rock Mechanics
Mining Geomechanics Abstracts, 17, 131-133.
Szwedzicki, T., Widijanto, E. and Sinaga, F. (2006) "Propagation of a Caving Zone, A Case
Study from PT Freeport, Indonesia." in Proud to Be Miners (MassMin 2004, Santiago,
August 2004), Santiago: Mineria Chilena.
Taylor, C. (1993) Personal communication
Thomas, L.A., (1971) Subsidence and related caving phenomena at the San Manuel mine,
San Manuel, Arizona: Magma Copper Company, San Manuel Division, unpublished
report.
Tiwari, R. and Rao, S. (2006). Post failure behaviour of a rock mass under the influence of
triaxial and true triaxial confinement. Engineering Geology 84 (2006) 112–129.
Trueman, R. and Mawdesley, C. (2003) Predicting cave initiation and propagation, CIM
Bulletin; May 2003; 96, 1071; ProQuest Science Journals, p. 54.
References
van As, A, Davison, J. and Moss, I. (2003) Subsidence Definitions for Block Caving Mines.
Rio Tinto Technical report. 59pp
van As, A. and Jeffrey, R. (2000) Hydraulic fracturing as a cave inducement technique at
Northparkes Mines. Proceedings, MassMin 2000. The Australian Institute of Mining
and Metallurgy. Publica-tions Series No. 7/2000, p. 165-172.
van As, A., Davison, J. and Moss, A. (2003) Subsidence Definitions for Block Caving Mines.
Rio Tinto Technical report. 59pp.
van der Merwe, N.J. (1999) Sub-surface Erosion — A Post Subsidence Phenomenon, in
Proceedings of the 9th ISRM Congress on Rock Mechanics, Paris, Balkema,
Rotterdam, Vol. 1.
Vanderwilt J.W. (1946) Ground Movement Adjacent to a Caving Block in the Climax
Molybdenum Mine, Technical Paper #200, Mining Technology.
Varas, F., Alonso, E., Alejano, L.R., Fdez-Manin G. (2005) Study of bifurcation in the
problem of unloading a circular excavation in a strain-softening material. Tunnel.
Undergr. Space Technol., 20 (2005), pp. 311–322
Vermeer, P.A. and de Borst. R. (1984) Non-associated plasticity for soils, concrte and rock.
Heron, Vol 29, 1984, no. 3.
Villegas, T. (2008) Numerical Analyses of the Hangingwall the the Kiirunavaara Mine.
Licentiate Thesis 2008:11. Division of Mining and Geotechnical Engineering,. Lulea
University of Technology, Sweden.
Villegas, T., Nordlund, E. and Dahner-Lindqvist, C. (2011) Hangingwall surface subsidence

at the Kiirunavaara Mine, Sweden. Eng. Geol., doi:10.1016/j.enggeo.2011.04.010
Vongpaisal S. (1974) Prediction of Subsidence Resulting From Mining Operations, PhD.

Thesis, McGill University, Montreal.
Vyazmensky, A., Elmo, D. and Stead, D. (2010) Role of Rock Mass Fabric and Faulting in the
Development of Block Caving Induced Surface Subsidence, Rock Mech. Rock Eng
43:533–556
Vyazmensky, A., Elmo, D., Stead, D. and Rance, J. (2007) Combined finite-discrete element
modeling of surface subsidence associated with block caving mining. Proceedings of
1st Canada-U.S. Rock Mechanics Symposium. Vancouver, Canada. 467-475.
White, D. (1984) Premining stress and its impact on block caving. AIME. Geomechanics
Applications in Underground Hardrock Mining . s.l. : AIME , 1984.
Whittaker, B N and Reddish, D J, (1989) Subsidence, Occurrence, Prediction and Control,

528 p. Elsevier: Oxford.
Wyllie, D. and Mah, C. (2007) Rock Slope Engineering, 4th Edition. Taylor and Francis,
Hampshire, United Kingdom. ISBN 041528001X
Yoshida, H and Horii, H. (2004) Micromechanics-based continuum model for a jointed rock
mass and excavation analyses of a large-scale cavern. International Journal of Rock
Mechanics and Mining Sciences. 2004, 41.
Zhang, Z. (2005) Finite Element Analysis of Ductile Fracture Behaviour of Pipe Sections
with SurfaceD Crack. PhD Thesis. Norwegian University of Science and
Technology.
References
Zhao, X. and Cai, M. (2010). A mobilised dilation angle model for rocks. Int J Rock Mech
Mining Sci (2010), doi:10.1016/j.ijrmms.2009.12.007
Zhao, Y.H., Tandon, G.P. and Weng, G.J. (1989) Elastic moduli for a class of porous
materials. Acta Mechanica, 76, 105-130.
Zhu, W.C. and Tang, C.A. (2003) Numerical Simulation on Fracture Characteristics of effect
in rock fracture process, Key Engineering Materials, 2003, 243-244: 339-344.
References

Cave Prop PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cave Prop PDF

Uploaded by

Copyright:

Available Formats

A MODEL FOR CAVE PROPAGATION AND SUBSIDENCE

ASSESSMENT IN JOINTED ROCK MASSES

B.E. (Geological Engineering) Royal Melbourne Institute of Technology

M.E. (Mining) The University of New South Wales

A Thesis submitted to The University of New South Wales in fulfilment of the

Dr. David Sainsbury, my long suffering husband, for encouraging me to complete

2.2.2.3 Mesh Dependency ............................................................................................................ 52

2.2.3 The Synthetic Rock Mass (SRM) Modelling Approach ............................................ 66

5.2.1.2 Discrete Fracture Network.........................................................................................109

5.2.6 Calibrated Laboratory Stress-Strain Curves .............................................................114

7.4.1.2 Zone of Large-Scale Fracturing.................................................................................155

9.3 Limitations of the Small-Strain Numerical Approach..................................................... 197

13.5.3 Conceptual Model of Subsidence Formation at the Grace Mine........................240

15.3.2.1 Small-Strain Calculation Model ..........................................................................261

Figure 1. Location of some historical and currently operating caving mines

Figure 18. Development of a discrete element model to study cave propagation

Figure 37. Example of simulated bi-linear, strain-softening response (after

Figure 60. Simulation of different boundary loading conditions on the response of

Table 1. Documented yield zone propagation rates from caving operations

BPM Bonded Particle Model

CPF Cave Propagation Factor

CWFS Cohesion Weakening, Friction Strengthening

DFN Discrete Fracture Network

DPE Dissipated Plastic Energy

FEM Finite Element Model

GSI Geological Strength Index

HOD Height of Draw

ICS International Caving Study

MMT Mass Mining Technology

MRMR Mining Rock Mass Rating

REV Representative Elemental Volume

SJM Smooth Joint Model

SLC Sub Level Cave/ing

SRM Synthetic Rock Mass

TDR Time Domain Reflectometry

UCS Unconfined Compressive Strength

UJRM Ubiquitous Joint Rock Mass

Direct Stress Positive stresses indicate tension. Negative stress indicates

Direct Strain Positive strain indicates tension. Negative strain indicates

Pressure A positive pressure will act normal to and in the direction

Vector Quantities The x, y and z components of vector quantities such as forces

1.1 Cave Mining Method

Cave mines require extensive infrastructure to be in place prior to any production.

Sub-level caving typically operates on numerous mining horizons simultaneously

1.2 Development of the Cave Mining Method

1.3 Caving Mechanics

Figure 4. Typical caving stress-paths representing stress and gravity caving

Figure 5. Conceptual stress-state development around a propagating cave (modified

(e) Rock Mass Discontinuities (faults, joints). The influence of geologic

(g) Other Excavations. Induced stresses may be impacted strongly by the

fundamental geomechanical principles is required to predict cave propagation

A conceptual model of a self-sustained propagating cave has previously been

Figure 6. Conceptual diagram of showing the main behavioural regions of a

The characteristics of each region are defined below:

 Seismogenic Zone - Microseismic (and sometimes seismic) activity is

The overall advance rate, thickness and spatial distribution of the

The damage threshold ( ) has been shown to correspond to the point

 Mobilised Zone – This zone gives an estimate of portion of the orebody

Cave propagation behaviour and subsidence are closely linked geomechanical

Figure 7. Conceptual schematic diagram showing the main behavioural regions of a

The characteristics of each region are described below.

 A crater is a common surface feature of many caving mines; it also is

several metres in magnitude. It develops as the mobilised zone influence