stability analysis of ore pass systems at brunswick mine

KAMRAN ESMAIELI

STABILITY ANALYSIS OF ORE PASS SYSTEMS AT BRUNSWICK MINE

Thèse présentée à la Faculté des études supérieures de l'Université Laval

dans le cadre du programme de doctorat en Génie des Mines pour l'obtention du grade de Philosophiae Doctor (Ph.D.)

DEPARTEMENT DE GENIE DES MINES, DE LA METALLURGIE ET DES MATÉRIAUX

FACULTÉ DES SCIENCES ET DE GÉNIE

UNIVERSITÉ LAVAL QUÉBEC

2010

Kamran Esmaieli, 2010

Résumé Des cheminées à minerai sont utilisées dans des mines pour transférer par gravité le minerai des niveaux supérieurs de la mine aux niveaux inférieurs. L'intégrité des cheminées à minerai peut être compromise par des changements des contraintes et par les structures géologiques en place dans le massif rocheux. Par ailleurs, le passage du minerai participe à l'usure des murs des cheminées, causée par les charges d'impact et par l'abrasion aux parois. La revue de la littérature pertinente sur la dégradation des cheminées à minerai, suggère que tous ces mécanismes peuvent potentiellement agir simultanément bien qu'un d'entre eux soit habituellement le plus dominant. La majorité des études abordant ce sujet ont considéré l'influence d'un seul mécanisme de rupture sur l'intégrité des cheminées à minerai ou n'ont pas pris en considération, de manière réaliste, l'interaction des différents mécanismes de rupture.

L'objectif de cette thèse était d'analyser la stabilité des cheminées à minerai à la Mine Brunswick. Une campagne de collecte de données a été réalisée à la mine. Une analyse qualitative et quantitative des données rassemblées a permis l'identification des pratiques actuelles et passées à la mine. Cette analyse s'est concentrée sur l'influence de la géométrie et de la configuration des cheminées à minerai, du massif rocheux, du régime des contraintes et des pratiques sur les problèmes opérationnels observés, reliés à l'utilisation des cheminées à minerai. Des cheminées à minerai avec des signes de dégradation majeure, ont fait l'objet d'une investigation plus détaillée. En particulier, l'évolution de la dégradation a été documentée et les mécanismes critiques de dégradation ont été identifiés.

Une approche de modélisation numérique étapiste a été utilisée pour étudier la stabilité des cheminées à minerai. L'approche utilisée considère l'interaction de l'influence des contraintes, de la fracturation inhérente au massif rocheux et de l'usure due au passage du matériel comme étant des mécanismes de dégradation des cheminées à minerai. La première étape de cette approche était l'utilisation des modèles d'analyse des contraintes 3D utilisant le logiciel Map3D pour calculer les contraintes globales. La seconde étape était la construction d'un modèle synthétique du massif rocheux qui simule plus adéquatement le comportement d'un massif rocheux fracturé. Le logiciel Fracture-SG a été utilisé pour représenter la géométrie du réseau de discontinuités existant, tel qu'observé sur des sites choisis à la Mine Brunswick. Le système de fracturation modélisé est ensuite couplé avec un modèle tridimensionnel d'éléments distincts, le Particle Flow Code (PFC). PFC simule la roche intacte comme un ensemble de plusieurs particules sphériques, aux dimensions uniformément distribuées, liées entre elles aux points de contact. Un modèle synthétique du massif rocheux peut être soumis à différents niveaux de contraintes tridimensionnelles. Selon le niveau des contraintes imposé, le massif rocheux synthétique peut céder par la rupture des portions intactes de roche ou par le glissement des discontinuités. Ceci est une innovation importante dans la simulation d'un massif rocheux fracturé.

Cette façon de modélisée un massif rocheux synthétique a été utilisée pour faire la rétro-analyse d'une cheminée déjà dégradée à la Mine Brunswick. Un modèle de fracturation du

massif rocheux a été généré en utilisant des données quantitatives obtenues pour un sulfure massif à la mine. Un modèle synthétique du sulfure massif a été construit en intégrant la géométrie du réseau de discontinuités modélisé dans un modèle PFC. Le modèle PFC a été ensuite calibré suivant les propriétés mécaniques des échantillons de roche intacte de sulfure massif testés en laboratoire. Ensuite, les contraintes aux frontières du modèle synthétique ont été appliquées et les volumes des cheminées à minerai ont été extraits du modèle. Ceci a permis la quantification de la grandeur des zones d'effondrement crées par l'interaction des contraintes et des structures en place autour de la cheminée à minerai. On a observé différents mécanismes d'effondrement tels que la rupture de portions de roche intacte entre les discontinuités préexistantes suivi par la chute de blocs rocheux générés par la propagation et l'intersection des discontinuités. L'influence de l'impact associé à l'écoulement des particules a alors été intégrée en projetant un fragment de roche (simulé par une particule sphérique rigide) sur les murs de la cheminée à minerai modélisés.

Cette thèse présente un cadre d'analyse portant sur l'étude de l'interaction et de l'influence des différents mécanismes de rupture sur la dégradation des cheminées à minerai. Utilisant des données de terrain amassées à la Mine Brunswick, il a été démontré que cette approche a des ramifications importantes sur la conception des cheminées à minerai. Elle peut potentiellement être employée pour la conception des cheminées à minerai dans une gamme de régimes structuraux et de contraintes. Cette approche peut faciliter le choix des configurations des cheminées qui peuvent atténuer la dimension des zones d'effondrement autour d'elles. Une contribution importante de cette thèse est l'intégration de l'influence de l'impact de matériel. Ceci peut aussi faciliter le choix d'une configuration de cheminées à minerai qui atténuerait l'apparition de dommages aux parois en raison de l'écoulement du matériel dans la cheminée. Ces techniques de conception améliorées peuvent potentiellement augmenter la longévité du système de cheminées à minerai et réduire les besoins en réhabilitation. Ceci aura des ramifications importantes sur l'investissement et les coûts opératoires.

Abstract

Ore pass systems are used in mines to transfer broken material from one mine level to a lower level using gravity. The integrity of ore pass systems can be compromised by changes in stress and the presence of fractures in the rock mass. Ore pass walls can also be damaged by the impact loads of rock fragments passing through an ore pass. Blast-induced damage for removal of material hang-ups in an ore pass can also contribute to ore pass wall degradation. A review of the pertinent literature on ore pass degradation suggests that all these failure mechanisms can potentially act simultaneously although one mechanism is usually the most dominant. The majorities of studies, addressing ore pass degradation, have either investigated the influence of only one failure mechanism or have not taken into account the full interaction of the different failure mechanisms.

The objective of this thesis was to analyze the stability of ore pass systems at Brunswick Mine. A comprehensive ore pass data collection campaign was undertaken at the mine site. A qualitative and quantitative analysis of the collected data allowed the identification of current and past ore pass practice at the mine. This analysis focused on the influence of ore pass geometry and configuration, rock mass and stress regime and practice on the observed ore pass operational problems. Ore pass systems that reported significant signs of degradation were selected for further study. In particular the evolution of degradation was documented and the critical degradation mechanisms identified.

A multi stage numerical approach was used to investigate stability of ore pass systems. The applied approach addressed the interaction of stress, rock structural defects and material flow impact as ore pass degradation mechanisms. The first element of this approach was the use of 3D stress analysis models using the Map3d stress analysis software package to calculate the global stresses. The second stage was the construction of a synthetic rock mass model to simulate the behavior of a fractured rock mass. Fracture-SG software was used to generate fracture systems similar to those observed at selected sites at Brunswick Mine. The fracture system models were then embedded into a three dimensional distinct element model, Particle Flow Code (PFC). The PFC simulates intact rock material as a dense pack of uniformly size distributed spherical bonded particles. A synthetic rock mass model can be subjected to different levels of stresses in three dimensions which depending on the stress level imposed, the synthetic rock mass can fail by fracturing through the intact rock matrix or by sliding along the pre-existent fractures. This is an important innovation in fractured rock mass simulation.

The synthetic rock mass approach was employed for back analysis of an ore pass failure at Brunswick Mine. Using quantitative field data collected for a massive sulphide rock mass at the mine, a fracture system was generated using Fracture-SG. The incorporation of the fracture network geometry into a Particle Flow Code (PFC) resulted in a synthetic rock mass. The PFC model was calibrated to the mechanical properties of the massive sulphide

m

intact rock samples measured in laboratory. In the next step, boundary stresses were introduced in the synthetic rock mass and the ore pass volume was excavated within the model. This allowed for the quantification of the failure zone extent inflicted by the interaction of stress and structure in the vicinity of the ore pass. Different failure mechanisms were observed in the form of fracturing of intact rock bridges between preexisting fractures followed by large displacements of rock slabs created by propagation and intersection of fractures. The influence of particle flow impact was then integrated by projecting a discrete rock fragment (simulated by a rigid spherical particle) against the ore pass walls represented by the synthetic rock mass.

This thesis has provided the framework for investigating the interaction and influence of different failure mechanisms on ore pass degradation. Using field data from Brunswick Mine it was demonstrated that this approach has important ramifications on ore pass design. It can potentially be used for the design of ore pass systems in a subjected range of stress and structural regimes. This approach can result in selecting ore pass configurations that can mitigate the size of failure zones around an ore pass. A major contribution of this thesis is the integration of the influence of material impact. This can further facilitate the choice of an ore pass configuration that will mitigate wall damage as a result of material flow in an ore pass. These improved design techniques can potentially extent the useful operating life of ore pass systems and reduce the need for ore pass rehabilitation. This will have important ramifications in both capital and operating expenditures.

Acknowledgements This project was financed by the Natural Sciences and Engineering Research Council of Canada and Xstrata Zinc. Access to the latest version of the PFC code was made possible by a grant from Itasca.

I would like to express my deep gratitude to all those who helped me to complete this thesis. First and foremost, I would like to thank my advisor, Professor John Hadjigeorgiou for all the support, friendship, enthusiasm and insight he has provided for me. His patient guidance and valuable suggestions have been a constant source of motivation for me in all the time of research.

My sincere thank goes to Dr. Martin Grenon (my co-supervisor) for his constructive comments and suggestions, especially on the fracture system modeling.

I am grateful to the Brunswick Mine's personnel: Richard Harrisson, Eric Coté, Terry McDonald and Patrick Mercier who facilitated data collection in the mine.

Special thanks to Matt Pierce and David DeGagne from Itasca, Minneapolis, for the useful suggestions and support they provided on modeling with the Particle Flow Code.

I would like to express my gratitude to Dr. Jean-Francois Lessard who gave me an introduction into the ore pass stability subject and for sharing his knowledge. Thanks to Geneviève Bruneau who has always been helpful to me.

Finally, I owe my loving thanks to my wife Shadi and my son Arvin. Without their encouragement and understanding it would have been impossible for me to finish this work.

November 2009

Quebec City, Quebec, Canada

This thesis is dedicated to my wife and my son.

vi

Table of Contents

RESUME I

ABSTRACT Ill

ACKNOWLEDGEMENTS V

TABLE OF CONTENTS VII

LIST OF FIGURES XIV

LIST OF TABLES XXVI

LIST OF PRINCIPAL SYMBOLS AND ABBREVIATIONS XXIX

1 INTRODUCTION 1

1.1 BACKGROUND 1

1.2 OBJECTIVES 4

1.3 METHODOLOGY 5

1.4 THESIS STRUCTURE 5

2 ORE PASS DEGRADATION 8

2.1 INTRODUCTION 8

2.2 PROBLEM DEFINITION 8

2.3 MECHANISMS OF ORE PASS DEGRADATION 12

2.4 FACTORS INFLUENCING DEGRADATION OF ORE PASS SYSTEMS 14

2.4.1 Influence of ore pass configuration 14

2.4.1.1 Ore pass location 15

2.4.1.2 Ore pass section length 15

2.4.1.3 Shape and cross section size 16

2.4.1.4 Inclination of ore pass section 16

2.4.1.5 Finger raise 16

2.4.2 Influence of rock mass behavior 17

2.4.3 Influence of high stress 19

2.4.4 Influence of material flow 20

vii

2.4.5 Influence of other factors 21

2.4.5.1 Excavation methods 21

2.4.5.2 Reinforcement 22

2.4.5.3 Control of material flow 23

2.5 MEASURE OF ORE PASS DEGRADATION 23

2.6 ORE PASS DEGRADATION ANALYSIS 24

2.6.1 Consideration of stress influence in ore pass degradation 25

2.6.2 Consideration of fractures in ore pass degradation analysis 26

2.6.3 Consideration of the influence of material flow in ore pass degradation 26

2.7 REHABILITATION OF DEGRADED ORE PASS 27

2.8 ORE PASS LONGEVITY 28

2.9 SUMMARY 29

3 ORE PASS PERFORMANCE AT BRUNSWICK MINE 30

3.1 INTRODUCTION 30

3.2 BRUNSWICK MINE DESCRIPTION 31

3.2.1 Mine location 31

3.2.2 Brunswick mine geology 32

3.2.3 Mechanical properties of rock units 35

3.2.4 In-situ stresses at Brunswick Mine 37

3.2.5 Mine operation 37

3.3 CONFIGURATION OF ORE PASS SYSTEMS AT BRUNSWICK MINE 37

3.4 CASE STUDIES OF ORE PASS DEGRADATION AT BRUNSWICK MINE 45

3.4.1 The 1000SFR ore pass 47

3.4.2 The #15 ore pass 50

3.4.3 The ore pass complex 18-21 56

3.5 LESSONS AND RECOMMENDATIONS 59

viii

3.6 CONCLUSIONS 61

4 FRACTURE SYSTEM MODELING 63

4.1 INTRODUCTION 63

4.2 FRACTURE MAPPING 64

4.3 STATISTICAL ANALYSIS OF FIELD DATA COLLECTED AT BRUNSWICK MINE 69

4.3.1 Fracture orientation 69

4.3.2 Fracture spacing 74

4.3.3 Fracture trace length 77

4.4 FRACTURE SYSTEM MODELING 81

4.4.1 Model generation 82

4.4.2 Validation of the fracture system model 84

4.5 SUMMARY AND CONCLUSIONS 90

5 SYNTHETIC ROCK MASS MODEL 92

5.1 INTRODUCTION 92

5.2 NUMERICAL METHODS FOR FRACTURED ROCK MASS SIMULATION 92

5.3 THE PARTICLE FLOW CODE 96

5.3.1 Simulation of intact rock properties 98

5.3.2 Simulation of mechanical properties of fractures 99

5.4 GENERATION OF A SYNTHETIC ROCK MASS MODEL 102

5.5 SUMMARY AND CONCLUSIONS 105

6 STRUCTURAL AND MECHANICAL PROPERTIES OF A SYNTHETIC ROCK MASS; A CASE STUDY FROM THE BRUNSWICK MINE 107

6.1 INTRODUCTION 107

6.2 STRUCTURAL AND MECHANICAL BEHAVIOR OF ROCK MASSES 108

6.3 FRACTURE SYSTEM MODELING (FSM) 110

6.3.1 Sampling of the fracture system I l l

IX

6.3.2 Estimation of the rock mass structural properties 112

6.4 GENERATION OF A SYNTHETIC ROCK MASS MODEL 114

6.4.1 Intact rock simulation 114

6.4.2 Fracture properties 116

6.4.3 Sample generation 117

6.5 MECHANICAL PROPERTIES OF SYNTHETIC ROCK MASS SAMPLES 119

6.5.1 Uniaxial compressive strength of synthetic rock mass samples 120

6.5.2 Elastic modulus of the synthetic rock mass samples 122

6.6 ESTIMATION OF THE REV SIZE 123

6.7 CONCLUSIONS 131

7 STABILITY ANALYSIS OF THE #19A ORE PASS AT BRUNSWICK MINE 132


7.2 ORE PASS DEGRADATION IN HIGH STRESS STATES 133

7.3 DEFINITION OF THE #19A ORE PASS PROBLEM 135

7.4 STRESS ESTIMATION AROUND THE ORE PASS COMPLEX 18-21 137

7.5 STABILITY ANALYSIS OF THE # 19A ORE PASS 140

7.5.1 2D and 3D synthetic rock mass model 141

7.5.1.1 Simulation of intact rock properties 142

7.5.1.2 Simulation of fracture properties 144

7.5.1.3 Synthetic rock mass generation and in-situ stress installation 145

7.5.2 2D Stability analysis of the #19A ore pass 148

7.5.3 3D Stability analysis of the #19A ore pass 153

7.6 COMPARISON OF THE PFC NUMERICAL MODELS TO FIELD OBSERVATIONS FOR THE #19A ORE PASS 157

7.7 CONCLUSIONS 162

8 INVESTIGATION OF ORE PASS WALL DEGRADATION DUE TO MATERIAL IMPACT164


x

8.2 DEGRADATION OF ORE PASS STRUCTURAL INTEGRITY BY ROCK FRAGMENT IMPACT 165

8.3 FINGER RAISE CONFIGURATION 166

8.4 IMPACT LOAD SIMULATION 167

8.4.1 Simulation of rock fragments and ore pass wall 168

8.4.2 Simulation of ore pass and finger raise configuration 170

8.5 IMPACT LOAD ON ORE PASS WALL 172

8.5.1 Influence of finger raise inclination on particle velocity and kinetic energy 172

8.5.2 Effect of finger raise inclination on impact load of particles on ore pass wall 175

8.6 EFFECT OF PARTICLE IMPACT ON A DEFORMABLE ORE PASS WALL 182

8.6.1 Impact induced stresses on the ore pass wall 185

8.6.2 Impact induced damage on the ore pass wall 190

8.6.3 Balance and dispatching of energy in an ore pass 193

8.7 IMPACT INDUCED DAMAGE ON A FOLIATED ROCK MASS ALONG ORE PASS WALL 197

8.8 CONCLUSIONS 206

9 CONCLUSIONS AND FUTURE WORK 209


9.2 SUMMARY OF THE RESEARCH WORK 209

9.3 CONCLUSIONS 211

9.4 LIMITATION OF THE EMPLOYED METHODOLOGY 216

9.5 FUTURE WORK 216

APPENDIX A: ORE PASS CONFIGURATIONS AT BRUNSWICK MINE 219

APPENDIX B: STATISTICAL ANALYSIS OF FRACTURE DATA 224

B.l FRACTURE ORIENTATION 224

B.2 FRACTURE SPACING AND FREQUENCY 225

B.3 FRACTURE SIZE 226

B.4 SAMPLING ERRORS FROM BIASES 226

xi

APPENDIX C: BONDED PARTICLE MODEL 229

C.l INTRODUCTION 229

C.2 BEHAVIOR OF PARTICLES AND BONDS 230

C.3 BPM GENESIS PROCEDURE 231

APPENDIX D: STRESS-STRAIN CURVES FOR SYNTHETIC ROCK MASS SAMPLES 234

APPENDIX E: STABILITY ANALYSIS OF VERTICAL RAISES IN HARD ROCK BY INTEGRATING A FRACTURE SYSTEM INTO A PFC MODEL 242

E.l INTRODUCTION 242

E.2 STRUCTURALLY CONTROLLED INSTABILITY IN UNDERGROUND RAISES 242

E.2.1 Traditional wedge analysis 243

E.2.2 Fracture system models 243

E.3 A CASE STUDY OF A VERTICAL RAISE IN THE CANADIAN SHIELD 244

E.4 STRUCTURAL STABILITY OF A RAISE IN A FRACTURE SYSTEM 245

E.5 2D STRESS MODELING 247

E.5.1 Identification of longitudinal section planes 247

E.5.2 Rock mass simulation in PFC2D 248

E.5.2.1 Simulation of intact rock properties 248

E.5.2.2 Simulation of fracture properties 249

E.5.3 Particle assembly generation 250

E.5.4 Installation of in-situ stress 251

E.5.5 Linking a 2D PFC model to a fracture system 251

E.5.6 PFC2D stability analysis 252

E.6 3D SYNTHETIC ROCK MASS MODEL 255

E.6.1 Simulation of intact rock and fracture properties in PFC3D 255

E.6.2 Particle assembly generation and in-situ stress installation 256

E.6.3 Linking the 3D fracture system model to 3D bonded particle model 258

E.6.4 PFC3D stability analysis 260

xii

E.7 CONCLUSIONS 265

REFERENCES 266

LIST OF CONTRIBUTIONS 279

Xlll

List of Figures

Figure 1-1. Simplified underground mine layout and infrastructures; from Lessard (2004) and modified from

Hamrin & Kennedy (1982) 1

Figure 1-2. Typical ore pass configuration, after Lessard & Hadjigeorgiou (2006) 2

Figure 1-3. Ore pass hang-ups caused by: a) interlocking arches, b) cohesive arches, after Hadjigeorgiou et al.

(2005) 3

Figure 1-4. Degradation of an ore pass wall, after Miller & Jacob (1996) 3

Figure 2-1. Conceptual representation of ore pass wall degradation phenomenon, after Lessard (2004) 9

Figure 2-2. Dog-earing of an ore pass, after Brenchley & Spies (2006) 12

Figure 2-3. Probable failure mechanisms (database of Quebec mines, 153 sections), after Hadjigeorgiou et al.

(2005) 13

Figure 2-4. Ore pass failure mode grouped by controlling factors; a) high-strength, brittle rock conditions, b)

non-brittle rock conditions, and c) fractured rock conditions, after Sjoberg et al. (2003) 14

Figure 2-5. Relationship between raise diameter and the Qr values considering actual performance of the

raises, after Peck & Lee (2008) 17

Figure 2-6. Case studies of ore pass development with respect to bedding, after Hadjigeorgiou et al. (2005). 19

Figure 2-7. Flow pattern in a smooth and rough sided ore pass, after Rech et al. (1992) 21

Figure 2-8. Discretisation and close-up of damage after fragment A impacts the jointed shale, after Loughran

etal. (2003) 27

xiv

Figure 3-1. Brunswick Mine location 31

Figure 3-2. Mine site access road, (26 km from Bathurst to the Mine), after Google Map 32

Figure 3-3. Vertical section of the Brunswick Mine deposit (looking North) 33

Figure 3-4. Geological plan of the 1125 Sub4 34

Figure 3-5. Schematic layout of ore passes at Brunswick Mine (not to scale) 38

Figure 3-6. Representations of an ore pass with reference to its sections, after Mercier-Langevin &

Hadjigeorgiou (2004) 38

Figure 3-7. Lengths of ore pass sections at Brunswick Mine (98 sections) 41

Figure 3-8. Inclinations of ore pass sections at Brunswick Mine (98 sections) 42

Figure 3-9. Flow control infrastructure in ore pass sections (98 sections) 44

Figure 3-10. The Map3D model of the mine with the ore passes, (looking North-West) 46

Figure 3-11. The geological plan of the 725-5sub and the expanded dimensions of 1000SFR 47

Figure 3-12. Cross sections of 1000SFR ore pass in different levels of the upper section 48

Figure 3-13. Induced stress conditions around the 1000 SFR ore pass in 2007 49

Figure 3-14. Case studies of ore pass development with respect to rock bedding: a) favorable orientation; b)

poor orientation, after Hadjigeorgiou et al. (2005) 49

Figure 3-15. a) The upper and lower sections of the 1000 SFR with respect to the rock mass foliation, b)

Foliation in Quartz Augen Schist, (photo looking North) 50

xv

Figure 3-16. The #15 ore pass intersects metasediments on 575-5sub 52

Figure 3-17. Metasediment rock on 725-6sub, photo: a) looking North; b) looking East 52

Figure 3-18. Degradation of the #15 ore pass along its length 53

Figure 3-19. Induced principal stress (a,) around the upper section of the #15 ore pass (elevation 2116 m),

using Map3D elastic model 54

Figure 3-20. The extents of expansion zone around the upper sections of the #15 ore pass developed based on

CMS results taken in 2003 and 2004 55

Figure 3-21. Advancing directions of the mining zones 20 & 21 with respect to the 18-21 ore pass complex,

after Andrieux et al. (2006) 56

Figure 3-22. a) Original dimensions of the ore pass complex 18-21, b) Dimensions of the ore pass complex

18-21 after degradation 57

Figure 3-23. Stress measuring grid placed in elevation 1582 m for analysis of induced stresses around the 18-

21 ore pass complex 57

Figure 3-24. Induced stress states around the ore pass complex 18-21 58

Figure 3-25. Expansions of the ore pass # 21 due to interaction of stress and particle impact degradation

mechanisms 59

Figure 3-26. Undesirable orientation for ore pass (left) and preferred orientation (right), after Brummer

(1998) 59

Figure 4-1. Location of the scanline#l at 1125-5sub-level 65

Figure 4-2. Location of scanline #2 in 1125-4sub-level 65

xvi

Figure 4-3. Location of scanline #3 in 1000-2sub-level 66

Figure 4-4. Location of scanlines #4, #5 and #6 in 1125-2sub-level 66

Figure 4-5. Rock mass exposure along the Scanline #3 67

Figure 4-6. Rock mass exposure along Scanline #1 68

Figure 4-7. Orientation measurement along scanline #6 68

Figure 4-8. Stereonet constructed from scanlines data 70

Figure 4-9. Fracture sets in massive sulphide rock, Photo looking West 71

Figure 4-10. a) Stereonet shows great circles for all fracture sets; b) Stereonet shows poles of all fracture sets.

71

Figure 4-11. Number of fractures intersected by different scanlines 72

Figure 4-12. Distribution of the dip angle of fractures 73

Figure 4-13. A normal distribution fit over the histogram of the fracture sets orientation data for a) set #1, b)

set #2 and c) set #3. Orientation difference is the angle between the mean pole of a fracture set and a

given pole of a fracture belong to the same fracture set 74

Figure 4-14. A histogram for total fracture spacing from scanline mapping data 75

Figure 4-15. An exponential distribution fitted over the histogram of normal set spacing values, a) set #1, b)

set #2 and c) set #3 77

Figure 4-16. A log-normal distribution fitted over the histogram for trace length of fractures in a) set # 1, b)

set # 2 and c) set # 3 79

xvn

Figure 4-17. Methodology of the Fracture-SG, after Grenon & Hadjigeorgiou (2003a) 83

Figure 4-18. Visualization of the generated fracture system 84

Figure 4-19. Stereonet constructed from the six scanlines introduced in the fracture system model 85

Figure 4-20. Comparison of the distribution of the fracture sets orientation obtained from the field and the

simulated data for a) set #1, b) set #2, c) set #3. Orientation difference is the angle between the mean

pole of a fracture set and a given pole of a fracture belong to the same fracture set 86

Figure 4-21. Comparison of the distribution of the fracture sets trace length obtained from the field and

simulated data for a) set #1, b) set #2, c) set #3 87

Figure 4-22. A histogram for total fracture spacing obtained from the simulated model 88

Figure 4-23. Scanlines introduced normal to the orientation of each fracture set, a) set #1, b) set #2 and c) set

#3 88

Figure 4-24. The probability density function for the fracture set normal spacing obtained from the field and

simulated data; a) set #1, b) set #2 and c) set #3 89

Figure 5-1. Suitability of different numerical methods for analysis of an excavation in a rock mass, a)

continuum methods, b) either continuum or discrete methods, c) discrete methods, d) continuum method

with equivalent properties, after Brady (1987) 94

Figure 5-2. Calculation cycle in PFC, after Itasca (2008a) 97

Figure 5-3. a) Notation used to define joint and smooth-joint contact, b) Large shearing motion results in the

creation of smooth-joint contacts along the fracture plane, after Mas Ivars et al. (2008) 100

xvin

Figure 5-4. Fractured triaxial test specimen showing fracture plane, strain, and displacement orientations, after

Rosso (1976) 101

Figure 5-5. Triaxial test for a BPM sample having a joint inclined at 45° 102

Figure 5-6. a) Fracture traces on a 2D section cut through a 3D fracture system model, after Grenon &

Hadjigerogiou (2003a); b) the resulting 2D synthetic rock mass model with fracture traces 103

Figure 5-7. A 2D synthetic rock mass model for high slopes, after Cundall (2007) 103

Figure 5-8. An example of a 3D synthetic rock mass model 104

Figure 6-1. Transition from intact to heavily jointed rock mass with increasing sample size, after Hoek (2001).

109

Figure 6-2. Photo of massive sulphide rock mass looking North- West 110

Figure 6-3. Visualization of the generated fracture system for massive sulphide rock mass at Brunswick Mine.

I l l

Figure 6-4. Rock mass samples drawn from the fracture system model, (not to scale) 112

Figure 6-5. Relationship between sample size and number of fractures, including variations 113

Figure 6-6. Fracture intensity (P32) of different sample size, P32(ave)=2.65 113

Figure 6-7. Bonded particle models of intact rock samples 118

Figure 6-8. Synthetic rock mass samples generated by linking the PFC3D and fracture system models 119

Figure 6-9. Uniaxial compressive test of synthetic samples: a) using wall for loading, b) using spherical grip

particles for loading 120

xix

Figure 6-10. Influence of specimen size on the strength of rock mass 121

Figure 6-11. The uniaxial compressive strength and failure pattern of two same size samples (0.05 m x 0.05 m

x 0.1 m); Yellow: tension cracks, Black: shear cracks 122

Figure 6-12. Influence of specimen size on the elastic modulus of rock mass 123

Figure 6-13. The REV concept; after Hudson & Harrison (1997) 124

Figure 6-14. Coefficient of variation vs. sample size 128

Figure 7-1. Schematic longitudinal section of the Brunswick Mine looking North-West, indicates the location

of Zones 20 & 21 at the bottom mining block of the mine and the ore pass complex 18-21 135

Figure 7-2. Advancing directions of the mining zones 20 & 21 with respect to the ore pass complex 18-21,

after Andrieux et al. (2006) 136

Figure 7-3. Ore pass complex 18-21, a) Original dimensions, b) Dimensions after degradation 136

Figure 7-4. The vertical grids placed perpendicular to each other in the Map3D model 139

Figure 7-5. Variation of stress states in different directions around the ore pass complex 18-21 during the

mining sequence 1-18 139

Figure 7-6. Fracture system model of the massive sulphide rock mass, (40 m x 40 m x 40 m) 141

Figure 7-7. Trace of fractures on the horizontal section 142

Figure 7-8. In-situ stress installation on the 2D bonded particle model 146

Figure 7-9. Construction of the 2D synthetic rock mass model by linking the 2D bonded particle model and

the trace of fractures on the horizontal plane 146

xx

Figure 7-10. Installation of in-situ stresses on the 3D bonded particle model 147

Figure 7-11. Representation of the 3D synthetic rock mass model 148

Figure 7-12. a) Simulation of a 3m diameter ore pass in the 2D synthetic rock mass model, b) The vectors of

velocity in the 2D ore pass model, some instance after the excavation, Y axis represents the North. ..149

Figure 7-13. Contact force distribution in the ore pass model (the thickness of the lines is proportional to force

magnitude) 150

Figure 7-14. Evolutions of damage zone around the 3 m of diameter #19A ore pass in the 2D model 151

Figure 7-15. Failure zone around the #19A ore pass after stress relaxation (zoomed view) 152

Figure 7-16. a) Displacement vectors in the 2D model after stress relaxation, b) The pattern of developed

micro-cracks around the # 19A ore pass 152

Figure 7-17. A plane view of the failure zone in the 3D model (superimposed) 153

Figure 7-18. A cross-section of the extents of damage zone around the #19A ore pass in the 3D model (Z = 2

m), (Blue: Tension crack, Red: Shear crack) 154

Figure 7-19. a) A longitudinal section of the #19A ore pass along the North-South direction a moment after

the ore pass excavation (particle or cluster of particles with brown, orange and yellow colors represent

rock wedges), b) the same section after stress induced damage, c) The vectors of particles displacement

along the North-South walls, d) The location of bond failures along the North-South walls, (Blue:

Tension crack, Red: Shear crack) 155

Figure 7-20. a) A longitudinal section of the #19A ore pass along the East-West direction a moment after the

ore pass excavation (particle or cluster of particles with brown, orange and yellow colors represent rock

wedges), b) the same section after stress induced damage, c) The vectors of particles displacement

xxi

along the East-West walls, d) The location of bond failures along the East-West walls, (Blue: Tension

crack, Red: Shear crack) 156

Figure 7-21. A cross section of the CMS results forthe#19A ore pass in 2003 and 2006 157

Figure 7-22. a) A cross section of the CMS results for the #19A ore pass indicating the dimensions of

expansion in different directions at 1563 elevation, b) A Longitudinal section of the CMS results for the

19Aore pass (looking West) 158

Figure 7-23. The results of CMS for the #19, #19A and #21 ore passes compared with the results of CMS for

an exhaust raise 159

Figure 7-24. A projectile particle impacting the South wall of the #19A ore pass 160

Figure 7-25. a) Impact of rock fragment with the South wall side, b) Damage due to the rock fragment impact

on the South wall side, c) Impact of rock fragment with the East wall side, d) Damage due to the rock

fragment impact on the East wall side 161

Figure 8-1. Damage and wear zones in an ore pass, Hadjigeorgiou et al. (2005) 165

Figure 8-2. A typical finger raise-ore pass configuration 167

Figure 8-3. Ore pass and finger raise configurations used in the numerical modeling 171

Figure 8-4. Temporal responses of two particles velocity, flow through a finger raise inclined at 60° 173

Figure 8-5. The influences of finger raise inclination on the particle impact velocity; for 70°, 80°, and 90° ore

pass inclination 174

Figure 8-6. Influence of the finger raise inclination on the kinetic energy of particles 175

Figure 8-7. Influence of the finger raise inclination on the impact duration 176

xxii

Figure 8-8. Normal and shear impact forces acting on inclined ore pass walls with the finger raise inclination

of 60°; a) Ore pass inclination 90°, b) Ore pass inclination 80° and c) Ore pass inclination 70° 177

Figure 8-9. The influence of finger raise inclination on the average normal and shear impact forces; a)

Vertical ore pass, b) Ore pass inclination 80°, c) Ore pass inclination 70° 179

Figure 8-10. Influence of finger raise inclination on the peak impact load of particles on the ore pass wall

inclined at a) 90° b) 80° and c) 70° 182

Figure 8-11. A 5m x 5 m bonded particle model represents a block of rock along an ore pass wall 184

Figure 8-12. A projectile particle impacting the 80° inclined ore pass 185

Figure 8-13. Impact of projectile particle with ore pass wall inclined at a) 90° b) 80° and c) 70°. The

measuring circles are identified in black on each ore pass wall 186

Figure 8-14. a) Contact force distribution in the ore pass wall inclined at 80° due to the impact of a rock

fragment, b) A close-up view of the model 187

Figure 8-15. The relationship between impact angles and the horizontal and vertical stresses induced on the

ore pass walls inclined at a) 90°, b) 80° and c) 70° 188

Figure 8-16. Measuring circles of different radii used to compute the impact induced stresses around the

impact region on the ore pass wall 189

Figure 8-17. Impact induced stresses at different distances from the impact point 189

Figure 8-18. Impact induced damage on the ore pass wall inclined at a) 90°, b) 80° and c) 70°, with a close-up

view of the damage zones 191

xxin

Figure 8-19. Influence of impact angle on the number of micro-cracks initiate on the ore pass wall inclined at

a) 90°, b) 80° and c) 70° 192

Figure 8-20. Impact of a single particle 193

Figure 8-21. Impact of multiple particles in an ore pass system 195

Figure 8-22. Case studies of ore pass development with respect to bedding, after Hadjigeorgiou et al. (2005).

198

Figure 8-23. The upper and lower sections of the 1000 SFR ore pass with respect to the rock mass foliation.

199

Figure 8-24. Foliated rock mass models with foliation angles of a) 60°, b) 90° and c) 120° 200

Figure 8-25. Impact induced damage on the ore pass walls inclined at 80° and comprising, a) a solid rock

mass, b) a 60° foliated rock mass, c) a 90° foliated rock mass and d) a 120° foliated rock mass 202

Figure 8-26. The patterns of impact induced damage on the ore pass wall inclined at a) 90° b) 80° and c) 70°

which contains foliation planes of 60° 203

Figure 8-27. Evolution of an impact induced damage zone in an ore pass wall with an inclination of 80° and

containing foliation planes of a) 60°, b) 90° and c) 120° 204

Figure 8-28. A zoom view of the evolution of a damage zone along the ore pass wall with an inclination of

80° and containing foliation planes of a) 60°, b) 90° and c) 120° 205

Figure 8-29. a) Impact induced damage on the ore pass wall with foliation angle of 60°, b) The same model

after sliding of a rock wedge into the ore pass 206

XXIV

Figure 9-1. Extent and pattern of failure zone around the #19A ore pass using the 2D synthetic rock mass

model 212

Figure 9-2. Extent of damage zone along the North and South wall sides of the #19A ore pass using the 3D

synthetic rock mass model 213

Figure 9-3. Structural failure (wedge falling) along the East and West wall sides of the #19A ore pass using

the 3D synthetic rock mass model 213

Figure 9-4. Interaction of stress, structure and particle impact on degradation of the # 21 ore pass 214

XXV

List of Tables

Table 2-1. Level of enlargement caused by wall degradation in Quebec mines (153 Ore pass Sections), after

Hadjigeorgiou et al. (2005) 10

Table 2-2. Status of ore passes in South African database (in percent and by length), after Joughin & Stacey

(2005) 11

Table 2-3. Distribution of ranges of dog ear spans in South African database, after Joughin & Stacey (2005).

11

Table 2-4. Wall degradation monitoring systems in Canadian ore passes, after Hadjigeorgiou et al. (2005)... 24

Table 3-1. Stratigraphie sequence of rock formations near Brunswick Mine, after Goodfellow (1975) 33

Table 3-2. Mechanical properties of various rock types at Brunswick Mine 36

Table 3-3. Investigated ore passes 39

Table 3-4. Ore pass excavation methods and section length 40

Table 3-5. Rock units distribution along the #15 ore pass 51

Table 4-1. The specification of the scanline mappings was done at Brunswick Mine 64

Table 4-2. Orientation characteristics of fracture sets 70

Table 4-3. Total spacing values for scanline mapping data 75

Table 4-4. Spacing characteristics of the fracture sets 76

Table 4-5. The mean trace lengths of fracture sets 78

xxvi

Table 4-6. Results of Kolmogorov-Smirnov goodness of fit test for distribution of sampled fracture sets

characteristics 80

Table 4-7. Summary of fracture set characteristics measured in the field 81

Table 4-8. Results of Kolmogorov-Smirnov goodness of fit test to validate the simulated fracture network.. 90

Table 6-1. Dimensions of samples collected from the Master fracture system model 111

Table 6-2. Micro-properties of the PFC3D models 115

Table 6-3. Mechanical properties of intact rock and bonded particle models 115

Table 6-4. Number of particles in different sample size 117

Table 6-5. The results of T-test for P30, P32. UCS and elastic modulus of synthetic rock samples, (5 samples

per each sample size) 125

Table 6-6. The results of F-test for P30, P32, UCS and elastic modulus of synthetic rock samples, (5 samples

per each sample size) 126

Table 6-7. Results of structural rock mass characterization 129

Table 6-8. Results of mechanical rock mass characterization 130

Table 6-9. Determination of REV size of rock mass based on coefficient of variation 130

Table 7-1. Mining Sequences at Zones 20-21 between June 2002 and Feb 2007 138

Table 7-2. Fracture sets characteristic for Massive Sulphide rock mass 140

Table 7-3. Micro-mechanical properties of 2D BPM model 143

xxvn

Table 7-4. Micro-mechanical properties of 3D BPM model 143

Table 7-5. Mechanical properties of intact rock in laboratory and 2D and 3D BPM 143

Table 8-1. Material properties used in the PFC models 168

Table 8-2. Ore pass and finger raise configurations and resulting angles of intersection 171

Table 8-3. Micro-mechanical properties employed for generation of the 2D bonded particle model 183

Table 8-4. Macro-mechanical properties of the 2D bonded particle model 183

Table 8-5. Impact velocities employed for different numerical experiments 185

Table 8-6. The results of energy balance in different ore pass configurations 196

Table 8-7. Recommendations for the design of different ore pass and finger raise configurations 207

xxvm

List of principal symbols and abbreviations

AE: Absorbed energy.

BEM: Boundary Element Method.

BPM: Bonded Particle Model.

BSE: Bond Strain Energy.

CDF: Cumulative Density Function.

CMS: Cavity Monitoring System.

COR: Coefficient of Restitution.

CV: Coefficient of Variation.

D/d: Ratio of the dimension of an ore pass (D) to the maximum size of rock fragment to be handled (d).

DE: Dissipated Energy.

DEM: Distinct Element Method.

DFN: Discrete Fracture Network.

E: Elastic modulus.

Ec, Ëc: Particle and parallel bond elastic modulus.

F: Sum of all externally applied forces acting on a particle.

FE: Friction energy.

FEM: Finite Element Method.

FSM: Fracture System Model.

g: The gravitational acceleration.

hj: Initial high of particle.

xxix

hr: High of bounce after impact.

I;: Inertia tensor of particle i.

ISRM: International Society for Rock Mechanics.

K-S test: Kolmogorov-Smirnov statistical test.

Kn, Kn'. Particle and parallel bond normal stiffness.

Ks, K$: Particle and parallel bond shear stiffness.

KE: Kinetic energy.

LRF: Ore Pass Longevity Reduction Factor.

LEF: Ore Pass Longevity Extension Factor.

m: Particle mass.

Nb: Number of particles.

P30: Number of fractures per rock volume.

P32: Cumulative fracture area per rock volume.

PDF: Probability Density Function.

PE: Potential energy.

PFC: Particle Flow Code.

Q value: Q index from the Barton et al. (1974) rock mass classification.

Qr: Modified Q value from McCracken & Stacey (1989) for stability analysis of raise bored excavations.

RMR89: Rock Mass Rating index from Bieniawski (1989) rock mass classification.

RSR: Raise Stability Ratio.

Rmin: Minimum particle radius.

REV: Representative Elemental Volume.

SE: Strain energy.

xxx

SJM: Smooth joint model.

SRM: Synthetic Rock Mass.

V: Particle incoming velocity.

Vr: Particle rebounding velocity.

Wext: Work done by external force.

WfriC: Friction work,

a: Ore pass inclination.

P: Finger raise inclination.

ON-S, GEW, GV: Principal stress ingredients along North-South, East-West and Vertical directions.

G] or amax: Maximum induced stress.

Oxx, <7yy and Gzz: Stress magnitudes along x, y and z directions.

UCS or Gc: Uniaxial compressive strength.

Gt: Tensile strength.

02'. Intermediate principal stress.

Gy. Minor principal stress.

Gn: Normal stress acting on a fracture plane.

Gb, Tb: Parallel bond tensile and shear strength.

T: Shear stress acting on a fracture plane.

u: Poisson's ratio.

y: Angle of intersection between finger raise and ore pass.

At: Time-step.

co;: Rotational velocity of particle i.

xxxi

p: Friction coefficient.

xxxii

Chapter 1 : Introduction

1 Introduction

1.1 Background

Ore passes are networks of vertical or steeply inclined openings in underground mines, and are commonly used to transfer ore or waste material to the lower level of a mine through the use of gravity flow. They can also serve as ore storage in the mines. Figure 1-

1 illustrates the layout of a typical metal mine, including ore and waste passes and other elements of the mine infrastructure. In the present dissertation, the term of "ore pass" is used to indicate all raises which serve to transport ore or waste.

^m/^M

g ^ ^ P - * ^S ; j J

I shaft

WW Level 1

,y'I- ••']f-^^M

^ y p ^ % Level 2

Ore Body ̂ k ^ M

Underground T Crusher T

Level 3

Ore Bin W i SWP

1 Waste puss 1 ■ Ore pass

■ I ^ I J I ^ ^ ^ M

Figure 1-1. Simplified underground mine layout and infrastructures; from Lessard (2004) and modified from Hamrin & Kennedy (1982).

The profitability of a mining operation is strongly influenced by the performance of its material handling systems. Ore passes are an integral part of a material handling systems.

1


Therefore, the performance of a material handling systems relies on the smooth and efficient operation of its ore passes. Figure 1-2 illustrates the components of a typical ore pass system.

Finger Raise

Ore Past Sections

Figure 1-2. Typical ore pass configuration, after Lessard & Hadjigeorgiou (2006).

Ore passes are commonly used because they provide a low cost means to move and store ore or waste rock materials. However, reports from mining operations indicate that ore pass systems frequently encounter two types of problems: degradation of the structural integrity of the ore pass, and hang-up of materials in the ore pass. Comprehensive reviews of ore pass related problems have been undertaken by Ferguson (1991), Stacey & Swart (1997), Hagan & Acheampong (1999) and Lessard & Hadjigeorgiou (2006).

Problems associated with the disruption of material flow in an ore pass are reported in most mines that operate an ore pass system. Hadjigeorgiou et al. (2005) classified the material flow problems as hang-ups or as blockages, based on the location of the material obstruction in the ore pass system. The blockages occur in the vicinity of the chute, while hang-ups are found in the ore pass. Hambley (1987) suggests that the transfer of coarse material can result in hang-ups due to interlocking arches, while the transfer of fine material results in hang-ups due to cohesive arches, Figure 1-3. Subsequent efforts to restore flow are often hindered by the lack of suitable access. Furthermore, traditional hang-up release methods result in production slowdowns and can cause safety problems for workers.


a) b)

Figure 1-3. Ore pass hang-ups caused by: a) interlocking arches, b) cohesive arches, after Hadjigeorgiou et al. (2005).

Ore pass integrity can be seriously compromised as a result of stress and structural failures, as well as by repeated blasting that is used to clear hang-ups, the impact from free falling material and abrasive wear caused by material flow. In addition, the frequency of ore pass blockage increases considerably in ore passes in which structural degradation is underway. Figure 1-4 illustrates the degradation of an ore pass system using a laser survey. At a certain level of degradation, the ore pass becomes unusable; it is because the ore pass blockage occurs very frequently or the enlarging of the volume of the ore pass is returned at a point which in the stability of the surrounding openings threatens. This condition may necessitate the rehabilitation or, in more severe cases, the abandonment of the ore pass.

Dimensions at the time of land surveying with laser

Original Dimension Original diameter = 3 m

Figure 1-4. Degradation of an ore pass wall, after Miller & Jacob (1996).

Chapter 1: Introduction

The problems associated with ore pass systems have a significant economic impact on the operation of a mine. This results in the need for rehabilitation or replacement of an ore pass, and in disruption of ore production and internal dilution. The ore pass problems may result in increase of production costs which can have significant repercussions, particularly when bulk mining methods are employed, in which case the profitability is dependent on low production cost.

Many accidents related to the operation and maintenance of ore passes occur every year. In Quebec, between 1987 and 1999, 60 accidents, including 9 fatalities, were listed relative to the operation of ore passes in underground mines, (Lessard, 2004). In Ontario, for the period of 1989 to 1999, 33 accidents were reported, of which 4 were fatal. In the United States, accident statistics reported by the U.S. Mine Safety and Health Administration (MSHA), indicate that, between 1987 and 2004, there were 7 fatalities and several hundred non-fatal ore pass related accidents, (Musto, 2004). In South Africa, according to the Department of Minerals and Energy, for the period of 1999 to 2004, there was an average of 6 accidents and 3 deaths per year in connection with ore passes, (Stacey & Erasmus, 2005). It can be seen from the above records that ore pass accidents remain significant in mining operations around the world.

Thus, proper functioning of the ore passes is critical to the operation of the mine as a whole. The ore pass must be designed so as to ensure the flow of material and the integrity of its structure over its life span. The current thesis will address issues related to the degradation of the structural integrity of ore pass systems.

1.2 Objectives

The primary objective of this thesis is to improve our understanding of ore pass degradation problems. The secondary objectives of this thesis are:

• To investigate ore pass stability problems at Brunswick Mine.

• To use a numerical approach that allows for quantification of the interaction of different ore pass failure mechanisms, including the influence of stress regime, rock structural defects present in rock mass and impact of material flow.

• To investigate the responses of ore pass structural integrity to material flow impact.


1.3 Methodology

The first step within the framework of the current thesis was the collection of ore pass design data at Brunswick Mine. The collected data was subjected to a statistical analysis to evaluate the relationship between the ore pass configurations and their performance at the mine. A detailed investigation was conducted on three ore pass case studies, where the original dimensions of these ore passes were significantly expanded. The important degradation mechanisms were identified for these ore pass systems. The #19A ore pass was then selected as a case study for development of the new ore pass stability approach.

A program of fracture mapping was carried out to collect fracture data on the massive sulphide rock mass exposures (the rock mass in which the #19A ore pass was excavated). A fracture system model was then derived from the in-situ mapping of fractures. The generated fracture system was linked to the Particle Flow Code (PFC), in order to construct a synthetic rock mass model. The intact rock properties of the synthetic rock mass were represented by an assemblage of bonded particles calibrated to the massive sulphide mechanical properties measured in the laboratory. Once the synthetic rock mass was built, in-situ stresses were then introduced on the boundaries of the model to account for interaction of stress and structure. Dimensions of the #19A ore pass were subsequently introduced into the synthetic rock mass to assess the stability of the ore pass. Finally, the effect of a rock fragment impact on degradation of the #19A ore pass was investigated by projection of a discrete particle against the ore pass walls, which were represented by the synthetic rock mass.

Response of ore pass structural integrity to material flow impact was further investigated using the Particle Flow Code. Several ore pass and finger raise configurations were investigated to find an optimum configuration that can minimize the impact induced damage of rock fragments on the ore pass wall.

1.4 Thesis structure

This thesis consists of 9 chapters and five appendices. The present chapter serves as an introduction to the entire dissertation, while chapter 9 provides a summary and conclusion of the research, together with recommendations for future research. A brief outline of each of the remaining chapters is as follows:

• Chapter 2 illustrates the ore pass degradation problem. A comprehensive literature review was conducted to identify ore pass failure mechanisms and factors which can influence the ore pass degradation rate. The influence of ore pass design parameters, the rock mass quality and the effect of stress and material flow on degradation of ore pass systems were discussed. Finally, the methods of stability analysis for ore pass systems were reviewed.


• Chapter 3 addresses the observation of ore pass performance at Brunswick Mine. A database of ore pass design was developed for the mine and the influences of different design parameters on the ore pass degradation were discussed. The ore pass systems with significant degradation problems were further investigated to determine failure mechanisms that have contributed to the expansion of their original dimension. The ore pass systems at Brunswick Mine are not lined. The influence of liners was not addressed in this thesis.

• Chapter 4 presents the fracture data collection for a massive sulphide rock mass at Brunswick Mine. This was followed by a statistical analysis of the collected data for the site of interest close to the #19A ore pass (the ore pass which was used as the main case study for the developing numerical approach). A fracture system was generated and validated based on the information on orientation, trace length and spacing which were calculated for each fracture set data at the mine. The validated fracture system represents the source of 2D trace sections and 3D fracture planes which were subsequently imported into the distinct element method of Particle Flow Code (PFC) as part of the study to construct a synthetic rock mass model.

• Chapter 5 presents the synthetic rock mass (SRM) approach, which attempts to simulate a fractured rock mass by linking a fracture system with a bonded particle model. This approach produces a fractured rock mass of any scale by incorporating intact rock mechanisms with mechanisms that include discontinuities. In this chapter, the principles of the bonded particle model and the methods of intact rock and rock fracture simulation using the bonded particle model were discussed. The procedure of a synthetic rock mass construction in 2D and 3D was then described.

• Chapter 6 illustrates the development of a synthetic rock mass model for the massive sulphide rock mass at Brunswick Mine. This was followed by characterization of the synthetic rock mass behaviors. A computational analysis was carried out to investigate the effect of sample size on the overall strength and deformation of the synthetic rock mass samples containing different fracture geometries. This computational analysis allowed for determination of a representative elemental volume (REV) size of the synthetic rock mass accounting for structural and mechanical properties.

• Chapter 7 presents the geotechnical analysis of the #19A ore pass which is located in a high stress area. A global stress analysis was used to estimate the variation of stress states around the #19A ore pass related to mining sequences. The stress analysis provided necessary boundary conditions for the 2D and 3D synthetic rock mass models in which the dimensions of the #19A ore pass was introduced. The extent of failure zones around the #19A ore pass, inflicted by interaction of stress and structure, was then quantified. Finally, a discrete rock fragment was thrown to the ore pass walls to integrate the influence of rock fragment impact in the ore pass wall degradation.

Chapter 1: Introduction

Chapter 8 illustrates the influence of particle impact on the degradation of ore pass walls. First, several ore pass and finger raise configurations were simulated with the Particle Flow Code. The worst configurations, which resulted in high impact load on the ore pass wall, were identified. This was followed by simulation of a block of foliated rock mass along an ore pass wall. A projectile particle was thrown against the rock mass block. Finally, impact induced damage inflicted by the projectile particle on the ore pass wall was quantified.

Chapter 2: Ore Pass Degradation

2 Ore Pass Degradation

2.1 Introduction

This chapter addresses issues concerning ore pass degradation. An overview of the literature on the observations of ore pass degradation in underground mines is presented. This includes the observation of ore pass degradation in Canadian and South African underground mines. A discussion follows on the mechanisms of ore pass degradation according to the different experiences regarding to the ore pass degradation analysis. The factors which can influence the degradation rate of ore pass systems are described. These comprises of the influence of ore pass configuration, rock mass behavior, stress states, material flow and other factors. Finally, the methods of ore pass degradation measurement are introduced and available approaches for ore pass degradation analysis are briefly explained.

2.2 Problem definition

Ore pass systems are an important element of materials handling in underground mines. These systems play a critical role in the maintenance of a constant level of production in mines. Any interruption in the operation of ore pass systems can reduce mine production and can endanger mine workers either during operation or during the implementation of reparative measures.

The successful operation of an ore pass is dependent on several factors. Beus et al. (1998a) grouped these factors into design parameters and operational considerations. The design parameters have interrelated structural and functional components. The structural components are associated mainly with the stability of ore pass systems, while the functional design is concerned with the material hang-ups in the ore pass systems. This thesis only dealt with structural components.

Ore pass systems are subjected to stress, and their stability is influenced by the presence of structural defects in the rock mass. Furthermore, ore pass systems are subjected to impact loads and wall wear as a result of the transfer of material through the ore pass. These mechanisms can lead to a gradual failure or a sudden instability in the structural integrity of an ore pass system, a condition defined as ore pass degradation. Figure 2-1, illustrates the phenomenon of degradation of the ore pass walls and the associated mechanisms.


Problems of ore pass Integrity

Stress Effect

Fracture Effect

Effect of Material Flow

Figure 2-1. Conceptual representation of ore pass wall degradation phenomenon, after Lessard (2004).

The degradation of ore pass wall results in the expansion of the original dimensions of an ore pass. This can influence the stability of adjacent openings (e.g. shafts) and the reliability of the material handling system. In some cases, the diameter of ore pass may expand to close to 20 times that of the original ore pass diameter, Hadjigeorgiou et al. (2005). The extent of the expansion zone around the ore pass is often irregular along the entire length of ore pass. The extent of this zone increases significantly when the ore pass traverses weak rock masses, Kazakidis & Morrison (1994). The size of the rocks that peel away from the ore pass wall depends on the failure mechanisms. In some cases, the rock wedges that dislodge from the ore pass walls are large enough to cause blockage or hangups in the ore pass, and they must be removed by blasting or other hang-up release methods. This can result in the delay of material transportation and the disruption of mine production. Furthermore, clearing the ore pass can put the miners in danger. If the ore pass wall continues to degrade, the ore pass may become unusable for transportation purposes, in which case it must be rehabilitated or replaced by an alternative ore pass, a process which poses significant cost implications.

Gardner & Fernandes (2006) summarized the signs of ore pass degradation as follows:

Significant increase in ore pass cross section


• Irregular ore flow through the ore pass

• Blockage of the control chutes at the bottom of the ore pass by large waste rocks fallen from the ore pass wall

• Mixing of ore and waste when neighboring ore and waste passes are merged

The degradation of ore pass systems has been observed in many underground mines. However, different criteria have been considered to describe the degree of ore pass degradation. Hadjigeorgiou et al. (2005) developed a database of ore pass systems for Quebec underground mines, including 153 ore pass sections, defined the degree of ore pass degradation based on the volume of ore pass sections in the observation time per their original volume. According to this definition, their database indicates that a majority of ore passes experience wall degradation, Table 2-1. Only 59 sections out of 153 sections (38%), did not show clear signs of degradation. Waste passes experienced less operational problems than ore passes. This is attributed to the transfer of material of lower mass weight.

Table 2-1. Level of enlargement caused by wall degradation in Quebec mines (153 Ore pass Sections), after Hadjigeorgiou et al. (2005).

Degree of degradation Ore pass sections

Negligible 38.6%

> 2 x original volume 9.8%

2 x original volume 7.8%



Not defined 26.1%

Joughin & Stacey (2005) investigated the ore pass degradation in South African mines. Of the 200 ore pass sections studied by them, more than 50% had experienced degradation problems and 16% had been abandoned, Table 2-2. They also introduced the term "dog ear" to describe the degradation degree of the ore pass walls that they observed. Dog-ear is a type of ore pass degradation that is the result of a stress induced failure initiation on the ore pass wall boundary, Figure 2-2. The span of the dog ear, the distance from the tip of one dog ear to the tip of the other, has been used to clarify degree of ore pass degradation, Table 2-3.

10


Table 2-2. Status of ore passes in South African database (in percent and by length), after Joughin & Stacey (2005).

Ore pass status Ore pass sections (length)

Good condition 48% (6350 m)

Poor condition 33% (4595 m)

Rehabilitated 3% (392 m)

Abandoned 16% (3074 m)

Table 2-3. Distribution of ranges of dog ear spans in South African database, after Joughin & Stacey (2005).

Dog ear span (m) Ore pass sections

< 4 m 43%

4 m - 10 m 19%

10 m - 2 0 m 31%

>20m 7%

11


Major horizontal field stress

j ^ .

; j Major horizontal field stress

É. JP^ Mr -tâàWWWW

i ^

Figure 2-2. Dog-earing of an ore pass, after Brenchley & Spies (2006).

The results of South African and Canadian ore pass performances illustrate the severity of ore pass degradation in underground mines. In the following sections, the mechanisms of ore pass degradation and the parameters which can influence the rate of ore pass degradation are discussed.

2.3 Mechanisms of ore pass degradation

The analysis of the phenomenon of ore pass degradation is complex, as it is often the result of the simultaneous interaction of various mechanisms. In different cases one can be predominant. These mechanisms are mainly: the stress regime near the excavation, the inherent fracture system of rock mass and wear due to the movement of the ore.

Based on the observation of ore pass degradation at the Sudbury mining camp, Morrison & Kazakidis (1995) proposed that five principal mechanisms contribute, to varying degrees, to the phenomenon of degradation. They did not, however, specify the frequency of relative importance of these factors, which include structural failures caused by material flow, wear due to impact loading caused by material flow, wear due to abrasion, blast damage caused by the hang up clearing methods and scaling of walls due to high stresses.

Morrison & Kazakidis (1995) introduced blast damage due to the hang-up clearing as another source of ore pass wall degradation. They indicated that in low stress environments ore pass failure is a combined result of the first four factors listed above.

12


However, in the case of deep mines (which they did not clarify how deep) where high stresses exist, stress can be an additional factor of ore pass expansion due to failure.

An example of the various degradation mechanisms which have been observed in Quebec underground mines was presented by Hadjigeorgiou et al. (2005), Figure 2-3. It should be noted that the same sections can present various mechanisms. Also it is recognized that the volume of material transferred has a significant impact on the rate of abrasion and wear.

Structural Failure 43%

Undetermined 6%

Blasting >1%

Figure 2-3. Probable failure mechanisms (database of Quebec mines, 153 sections), after Hadjigeorgiou et al. (2005).

Sjoberg et al. (2003) introduced a causative model in which failure mechanisms and failure development were described based on the observations collected in the Kiirunavaara mine in Sweden, Figure 2-4. The model was derived from a compilation of ore pass and inspection data for identification of failure modes. The effect of wear and rock fragment impact was expressed as tonnage that had been transported through the ore pass. However, this is a conceptual model rather than a quantifying model.

13


a) High-strength brittle rock

Low stress - ^

Low tonnage No failures

High tonnage

Groove

0 o I Spalling + groove

b)

Low tonnage

High tonnage

High- and low-strength (non-brittle) rook

C) Fractured rock

Low and High stress conditions

Low & High tonnage

I Dominating N-S

joints

1

I Blocky rock

Figure 2-4. Ore pass failure mode grouped by controlling factors; a) high-strength, brittle rock conditions, b) non-brittle rock conditions, and c) fractured rock conditions, after Sjoberg et al.

(2003).

Based on the reported ore pass failure mechanisms in a number of case studies, one can conclude that ore pass degradation can be a result of interaction of several failure mechanisms acting simultaneously. However, there are cases where only one failure mechanism is important.

2.4 Factors influencing degradation of ore pass systems

2.4.1 Influence of ore pass configuration

A review of ore pass design demonstrated that various direct rules suggested for the proper design of ore passes based on databases of previous experiences. These databases

14


come mostly from experiences at South African mines; Stacy & Swart (1997), Hagan & Acheampong (1999), Stacey & Erasmus (2005) or experiences at North American mines; Ferguson (1991), Hambley et al. (1983), Hadjigeorgiou et al. (2005) and Lessard & Hadjigeorgiou (2006). The ore pass design considerations which influence the structural integrity of the ore pass system are presented here.

2.4.1.1 Ore pass location

The choice of a location for ore pass systems in underground mines aims to achieve a balance between geotechnical constraints and operational considerations. The ore pass location depends on the choice and location of crushing systems, such as jaw crushers and rock breakers. It is often dictated by proximity to the shaft or to the stopes. Locating the ore handling system near the shaft reduces hauling distance between the crusher and the loading pocket. However, in this configuration, any failure of the ore pass can compromise the integrity of the shaft. Ore passes located close to extraction stopes are exposed to increases in induced stress levels or to unfavorable changes in stress. This configuration, however, results in a longer hauling distance between the crusher and the mine shaft while reducing the handling distance from the stopes to the ore pass.

Another consideration, which is extremely important from a geotechnical point of view, is the quality of the rock mass in which the ore pass is located. The best possible rock mass quality must be selected for ore pass construction. Moreover, the choice of orientation must be considered with regards to geological structures and stress state orientation. In high-stress conditions, this includes an ore pass orientation sub-parallel to the maximum principal stress or to locate it in a stress shadow area, and under low stress conditions, to locate it in an area that provides well confined conditions, Stacey & Swart (1997).

2.4.1.2 Ore pass section length

The length of an individual ore pass section depends on the vertical distance between mine levels and/or sublevels. The length of ore pass sections is generally dictated by the ore pass excavation method. Typically, drop raising or conventional raising are used for excavation of short ore pass sections, while sections driven by Alimak or raise bored sections are long.

Long ore pass sections have a greater probability of intersecting zones of poor ground, which results in a higher potential for operational problems and is harder to bypass. When the ore pass is kept empty, the longer sections can result in higher material flow velocity and, consequently, substantial impacts and abrasion on the ore pass wall. Moreover, the use of longer sections requires the development of finger raises in several mining levels or sub levels to funnel the broken material to the ore pass. This will result

15


in more dump points along the ore pass system, which can adversely influence the structural integrity of the ore pass system.

2.4.1.3 Shape and cross section size

The minimum dimension of an ore pass (D) compared to the maximum size of rock fragment to be handled (d), is an important factor with regard to the possibility of hangups and blockages (3<D/d<5), Stacey & Swart (1997). Any increase in the number of hang-ups or blockages in an ore pass system results in the more frequent application of ore pass clearance methods which include drilling and blasting and which can lead to additional degradation of the ore pass system.

The dimensions of an ore pass have to be addressed in connection with its shape. Circular ore pass sections are usually associated with raise boring methods, while rectangular sections are excavated using Alimak methods or drop raise methods. Circular shapes are generally used for higher stress regimes, as the distribution of stress around the circular excavations is uniform. However, Hadjigeorgiou et al. (2005) reported that under high stress condition, and with material flowing in an ore pass, an unlined circular ore pass does not maintain its shape for long.

2.4.1.4 Inclination of ore pass section

The inclination of an ore pass section is dictated by the need to simultaneously facilitate and decelerate material flow. The ore pass inclination has an effect on the flow characteristics. In steeper ore pass sections, material travels at a higher velocity down the ore pass which can cause more damage and wear, preferentially on the footwall side of the ore pass. High speed impact of rock fragments on the broken material in controlled ore passes (ore passes which are kept full) causes compaction, which promotes hang-ups. On the other hand, shallow ore pass sections may restrict flow, especially if a large proportion of the flowing materials are fine materials. In addition, the shallower inclination necessitates an ore pass of a greater length.

2.4.1.5 Finger raise

It has been noted that the use of long sections requires the development of finger raises, Lessard & Hadjigeorgiou (2006). Whenever an ore pass intersects two or more levels, finger raises are used to funnel material into the pass. Review of cavity monitoring surveys and site investigations in Canadian underground mines by Hadjigeorgiou et al. (2005) suggests that fall of rock fragments from a finger raise often results in a localized impact zone on the ore pass wall, which can subsequently result in an enlargement in the vicinity of the intersection of the finger raise and the ore pass.

16


2.4.2 Influence of rock mass behavior

The quality of rock mass in which an ore pass is constructed may have the most significant effect on the performance of the ore pass. The influence of ground conditions on the stability of raise bored excavations has been investigated by McCracken & Stacey (1989). They suggest that the Q system, developed by Barton et al. (1974) can be modified to provide design guidelines for raise bored shafts. The modifications were introduced to account for wall stability, excavation orientation relative to structural features and rock weathering and alteration. They related the modified Q value (called Qr) to the maximum stable diameter of the raise using the following equation:

Design Diameter (m) - 2 x RSR x (Qr) 0.4 ( 2 - 1 )

Where Qr is the modified Q value and the RSR is Raise Stability Ratio which depends on the type and function of the raise. For temporary mine excavations the RSR value is bigger than the long-term civil openings.

Recently, Peck & Lee (2008) who documented 50 cases of raisebored vertical excavations in Australian and Papua New Guinean mines, Figure 2-5, concluded that the McCracken & Stacey method cannot provide a basis for estimating 'how quickly an unstable raise might deteriorate, become unserviceable and perhaps collapse'. They furthermore concluded that for the raises constructed in the rock masses with Qr value less than 0.1 it is more probable that the raise would be collapsed or significantly over-broken, irrespective of the raise diameter. On the other hand, there is a high chance to have a stable raise if it is constructed in rock mass with Qr > 1.0.

01

X Statu» B Stable w*h Support

0001 0100 Lower Bound Qr Values

1000 10 000 20 000

Figure 2-5. Relationship between raise diameter and the Qr values considering actual performance of the raises, after Peck & Lee (2008).

17


The method by McCracken & Stacey cannot be used for conventional and Alimak driven raises. Brummer (1998) suggested that although the McCracken & Stacey method accounts for stress through the Stress Reduction Factor (SRF), but it is more applicable for assessing the stability of raises under lower stress conditions. In addition, the method requires detailed geomechanical data which in most mining cases is not available.

Hadjigeorgiou & Lessard (2003) investigated the relationship between rock mass quality and ore pass operational failure in 10 underground mines in Quebec. They have reported that there is no incidence of uncontrolled ore pass failure in any ore pass section that had a Q value greater than 5. In another case study of the investigation of ore pass systems at East Boulder Mine (Stillwater Mining Co.) in Montana, USA, Smith et al. (2006) reported three ore pass systems constructed in rock masses having a Q value of less than 3. The ore passes failed after less than 2 years of operation, mostly due to the intersection of the ore passes with fault zones or highly fractured zones, which resulted in kinematic wedge failures in the investigated ore passes. Of the three passes, two were abandoned and one was rehabilitated.

It has been widely accepted that the more homogeneous the outer walls of an ore pass (no structural defects present), the more stable the ore pass will be. The presence of structural defects in the rock, such as bedding planes, faults, dykes and shear zones or geological horizons having of varying strength (such as shale) on the surface of ore pass wall, tend to result in inconsistent degradation of the ore pass. Softer strata degrade more rapidly, which compromises the original design shape of ore pass and may interrupt the flow of ore. It should be expected that as soon as the weaker parts of the rock mass yield, failure will shift to stronger parts.

Several case studies of ore pass failure in which the principal mechanism of failure is believed to be the intersection of the ore pass with rock structural defects or low strength strata have been documented, including, Buffelsfontein Gold Mine by Hilner (1966), Kloof Gold Mine Shaft # 1 by Gibbon (1976), Doornfontein Gold Mine by Gibbs (1980), South Division of Vaal Reefs Mine by Schoombee & Van Wyk (1982), St-Helena Gold Mine by Stringer & Lategan (1982), Creighton Mine by Morrison & Kazakidis (1995), Impala Platinum Mine by Gardner & Fernandes (2006) and East Boulder Mine by Smith et al. (2006). Based on these case studies, the thickness of a poor rock mass zone intersected by an ore pass and the dip of the strata in relation to the ore pass are critical to the durability of the ore pass. Low angles of intersection between the ore pass and the strata layers create wedges, which break under impact or wear, resulting in uneven ore pass walls and blockage of the ore pass.

Stacey & Swart (1997) suggested the driving of the ore passes against the foliation dip. For bedded rock masses they recommended that ore passes should be designed to intersect the strata as close to perpendicular as possible. Hadjigeorgiou et al. (2005) pointed out a similar problem at Quebec underground mines, where they compared two ore passes at two mines in the same structural regime characterized by steep bedding, Figure 2-6. The first of the two ore passes excavated at a favorable orientation with

18


respect to the bedding was stable, Figure 2-6a, while the second ore pass excavated along the bedding had experienced severe degradation, Figure 2-6b.


2.4.3 Influence of high stress

Stress-induced instability in ore passes is often associated with the development of dog-earing or scaling and spalling. This behavior is the predominant failure mode of deep ore pass sections or ore pass sections located in the abutments of mining stopes. In addition, proximity of an ore pass to major structural features can result in a stress concentration zone around the ore pass. The failure occurs in an anisotropic stress field as a result of stress concentrations on the ore pass wall situated perpendicular to the axis of major in-situ stress. The extent of the scaling depends on the strength of the rock mass in abutment of the ore pass and the ratio of the horizontal stresses. The extent of spalling or dog-earing decreases if the ratio of the horizontal stresses is closer to unity.

Several case studies of stress induced ore pass failure have been reported, including, Tau Lekoa Mine by Dukes et al. (2006), Moab Khotsong Mine, Brenchley & Spies (2006), East Boulder Mine by Smith et al. (2006) and Kloof Gold Mine Shaft # 3 by Hart (2006). In most of the reviewed case studies the flow of ore material aggravates ore pass scaling. Moreover, the interaction of stress induced failure and geological structures leads to the creation of unstable rock wedges around the ore pass wall and to frequent ore pass blockage.

19


2.4.4 Influence of material flow

What distinguishes ore passes from other openings is that they are used to transport waste/ore rock. Flow of material within an ore pass can adversely influence the structural integrity and life expectancy of the ore pass. However, flow dynamics are often given little consideration, principally because the concepts involved are hard to quantify.

The inclination of the ore pass has an influence on the flow characteristics, the possibility of hang up and wear of the pass surfaces. The velocity of rock flow down an ore pass is controlled by the inclination of the excavation. It is therefore necessary to arrive at a compromise between the flow of material and the velocity of impact. In a very steep ore pass, when empty, blocks of rock travel down the pass at high speed, bouncing against the walls and possibly causing damage. In inclined passes, as the material tends to slide down the footwall surface, most wear occurs on this surface, Stacey & Swart (1997).

Wear of an ore pass wall can be caused by either impact or frictional sliding (abrasion). Impact wear caused by a stream of falling rock can induce degradation on the ore pass wall. Goodwill et al. (1999) noted that as the diameter of an ore pass increases, the likelihood of the direct impact of rocks on the ore pass walls declines. However, wall damage attributed to impact loading is most often localized at the intersection of finger raises to the ore pass. In order to quantify the impact load on the ore pass wall, Hambley et al. (1983) suggested that the average force of particle-wall impact is directly proportional to particle velocity and particle weight, and inversely proportional to impact duration.

The wear caused by rock and ore flow within an ore pass involves several parameters: grading, size of larger blocks, specific gravity of rock, hardness of ore, angle of chute and type of circulation within the ore pass (flow through or controlled flow), Van Heerden et al. (2005). The flow of rock fragments having great unit weight can result in more damage to the ore pass wall. This issue has been reported by Hadjigeorgiou et al. (2005), who have compared the degradation of ore and waste passes in Quebec underground mines. The waste passes exhibited less degradation as they are used for transportation of lighter rock fragments.

Frictional sliding or abrasive wear occurs when the solid bulk particle is harder than the wearing resistance of ore pass wall surface. Abrasive wear depends on the abrasiveness of rock fragments and the resistance of the ore pass wall to abrasion. In ore passes, it is the interaction of the material transported through the pass and the excavation wall that are of interest. Hadjigeorgiou & Lessard (2003) suggested that there is an important distinction to be made between the abrasiveness of transiting material in an ore pass and the resistance to abrasion of the rock surrounding the excavation. Quantification of these distinct properties has been hindered by the absence of standardized tests.

The abrasiveness of rock fragments is dictated by the type and quantity of the various mineral constituents of the rock and their bond strength. The abrasiveness of a material can be assessed by petrological as well as mechanical methods. Petrological methods

20


attempt to link the mineralogical composition of a rock to the hardness of its constitutive minerals. It is generally accepted that highly siliceous ore, when blasted, produce a strong, abrasive, brittle rock, while calcareous rock results in non abrasive material.

Quantification of the resistance of the ore pass walls to abrasion is complicated by the fact that the tests available to quantify resistance to abrasion are designed for aggregates or intact rock. It can be argued that neither of these adequately represent a jointed rock mass.

Hadjigeorgiou & Lessard (2003) suggested that abrasion wear, by itself, is not likely to be the critical factor in ore pass failure. Nevertheless, if the rock mass is not able to provide adequate resistance to abrasion, consideration should be given to the use of liners.

2.4.5 Influence of other factors

2.4.5.1 Excavation methods

Ore passes are excavated using either mechanical (raise borer) or drill and blast techniques (Alimak, conventional raising and drop raising). Mechanical methods that employ raise boring machines result in less ground disturbance during excavation. Moreover, raise bored ore passes have a smooth wall surface which can promote flow and reduce the possibility of hang-ups. Conversely, this smoothness allows the material to flow at a faster rate, which can lead to greater compaction and wear problems. Rech et al. (1992) suggest that an ore pass with smooth walls (lined or bored ore passes) will lead to laminar flow. Ore passes with rough walls, excavated with conventional drilling and blasting method, will lead to flow from the outside inward as depicted in Figure 2-7.

Haulage Haulage

a g j j ^ /#f- SS58555^

Figure 2-7. Flow pattern in a smooth and rough sided ore pass, after Rech et al. (1992).

Methods relying on the use of explosives, such as conventional raising or the use of an Alimak, require that men work inside the ore pass during development while allowing the

21


installation of support. This is not the case for drop raises; however, they do require more precise drilling.

2.4.5.2 Reinforcement

The purpose of ground support in ore pass construction is to ensure stability during the excavation process and to prevent structural failures during the useful life of the pass, Hadjigeorgiou et al. (2004). In the case of ore passes, it is convenient to treat separately the support provided by reinforcement units, such as bolts, and the reinforcement provided by liners.

Various support types have been used for ore pass stability, with varying success. Rock bolt reinforcement has been used frequently, but with little success. In blocky rock and scaling rock situations, wear of the ore pass causes the rock between the bolts to fall out. Conventional rigid rock bolts are usually inappropriate, as the impact of rock fragments causes the bolt to vibrate, thus destroying the bond between the bolt and the rock. This vibration tends to crack and break up the grout annulus surrounding the bolt, which then is easily dislodged by falling rock, Stacey & Swart (1997) and Brenchley & Spies (2006). Fiberglass bolts and resin-grouted cables are more effective, as they are less prone to vibrate under the impact of rock travelling down the pass. Rock reinforcement should be installed in upward-inclined holes so that any impact from material flowing down the pass does not contact the support at an acute angle.

In general, liners may be used where it is expected to encounter stability problems or for design configuration requirements. Unlined ore passes ending to a crusher station can unravel to impair the integrity of the station. The liners can be also employed in response to various scenarios that dictate the use of rehabilitation. It is generally accepted that a lined ore pass will result in a lower angle of friction at the broken ore-ore pass wall interface than an unlined ore pass driven with conventional blasting or boring methods. Hadjigeorgiou & Lessard (2003) summarized the successful functions of liners in ore pass systems as follow:

• To improve the structural stability of ore pass excavations

• To resist uncontrolled enlargement of an ore pass volume due to the wear, impact of rock fragments, etc

• To improve material flow

Liners can also provide protection to the primary installed support systems to protect them against impact or frictional wear. When wear is a problem, special abrasion resistant linings are used such as fiber, corundum and andésite lava based shotcretes or cast concretes, Van Heerden et al. (2005). Concrete, however, does not accommodate the deformations that occur with stress changes, and has therefore not found much

22


application in deep level mining. The application of spray-on liners in high impact and abrasion areas has also been reported at Stobie Mines, Inco, Ltd., Sudbury (O'Shaughnessy,2001).

2.4.5.3 Control of material flow

Ore passes can be operated as flow-through (free flow) or they can be kept full. Different infrastructures including control chains, chute with control chains and etc. can be installed in the discharge point of an ore pass to restrict the flow of material in the ore pass. The flow-through ore passes are more susceptible to wall damage. Several ore pass operation experiences suggest that a full ore pass prevents wear and damage due to impact of rock blocks hitting the walls at high velocity. A further advantage of keeping the ore pass full is that it provides confinement to the sides of the ore pass, thus reducing structural failure or scaling due to high stress, Stacey & Swart (1997) and Morrison & Kazakidis (1995). Keeping an ore pass full, however, increases the risk of hang-ups, particularly if there is a high level of fine material. When material flow is interrupted for a long period of time, fine materials tend to segregate towards the bottom of the ore pass and/or material has time to oxidize and harden. This process can be mitigated by drawing material out of the ore pass at regular intervals.

The flow of large rocks through an ore pass can contribute to wall degradation and lead to blockages or interlocking hang-ups. In order to control the size distribution, size control infrastructures are installed to restrict the entrance of oversize material. This can be achieved by installing flow control infrastructures such as grizzlies, scalpers and mantles. Iverson et al. (2003) used numerical models to demonstrate that installing a grizzly could reduce the impact load up to 20 times.

2.5 Measure of ore pass degradation

Ore pass management necessitates the use of reliable monitoring techniques. These systems can be either direct or indirect and range from very simple (visual inspections) to fully automated systems (laser). The techniques available to monitor the integrity of an ore pass are listed in Table 2-4.

One of the principal problems for the comprehension of the degradation phenomenon is that it is seldom possible to visually inspect the zones of degradation, because of the dangerous conditions of inside the passes. The indirect observation of the condition of the walls through the use of a camera has spread considerably in recent years. However, it is difficult to obtain quality images because of the conditions that exist in the ore passes. In most cases, there is a large degree of interpretation. Land surveying through the use of laser such as Cavity Monitoring Systems (CMS) are beginning to be used regularly in mining operations in order to evaluate the localization and extent of degradation zones.

23


These tools have the advantage of greatly facilitating the interpretation of the zones of degradation. Another indirect method which is widely used to evaluate the volume of ore pass degradation is by comparing the current tonnage of material which a pass can contain to the tonnage that it could contain initially.

Table 2-4. Wall degradation monitoring systems in Canadian ore passes, after Hadjigeorgiou et al. (2005).

Type System Comments

Direct

Visual Evaluation of wall degradation from the upper access. Inaccurate. Pass must be emptied.

Direct

Drilling Drilling into the pass from nearby drifts to determine an

approximate profile. Pass can be kept full. Provides only localized data. Necessitates access.

Direct Camera A camera is lowered into the pass. Necessitates the use of a

buggy. Pass must be emptied.

Direct

CMS A cavity monitoring system provides a full 3D profile of the pass Accurate (occasional blind spots). High capital cost.

Necessitates access. Pass must be emptied.

Indirect OreAVaste Tonnage

Reconciliation Subtracting pass output to pass input (scoop buckets). Pass

can be kept full. Highly inaccurate.

2.6 Ore pass degradation analysis

Empirical, analytical and numerical methods are the techniques that have been frequently used in ore pass degradation analysis. Empirical approaches are based on previous experience and databases of ore pass behavior dealing with particular geological structures, in situ and induced stress conditions and local technical experiments.

In analytical approaches, various theoretical and analytical methods are used to evaluate the effect of stress condition, rock mass strength and geological structural features on the stability condition of ore passes.

Numerical models are computer programs that attempt to represent the mechanical response of a rock mass subjected to a set of initial conditions (e.g., in situ stresses), boundary conditions, and induced changes (e.g., ore pass dimension). Numerical methods allow the analysis of complex factors such as inclusion of complex discontinuity pattern, dynamic loading, inhomogeneity, anisotropy, material softening and time dependent behavior.

24


A review of the literature available on the subject shows that different methods have been used to evaluate ore passes integrity analysis. While most analyses considered the effect of only one failure mechanism, some took into account the interactions of several mechanisms.

2.6.1 Consideration of stress influence in ore pass degradation

The ore pass systems which are located in high stress areas are frequently encountered with the problem of degradation. The geotechnical investigation methods developed for the ore pass systems in high stress environments are based on the estimation of induced stresses around the opening and the choice of a proper failure criterion for the rock mass. By these means, analytical and numerical methods can be used for the evaluation of induced stresses around the ore pass wall. Closed form solution and computer solutions using 2D or 3D boundary element methods are the means of determining induced stress. To assess the likelihood as well as the extent of failure of rock around an ore pass, it is necessary to use failure criteria.

Kirsten & Klokow (1979) applied an elastic solution for the distribution of stress around ore passes to predict failure in ore passes. Compressive strength of intact rock (the ratio of the maximum tangential stress to the uniaxial compressive strength of rock), shearing strength of weak planes and extensive strain fracturing were used as failure criteria. Based on case studies in South Africa, they demonstrated that predicted failures were consistent with observed failures.

Joughin & Stacey (2005) used the Kirsch closed-form method to estimate the maximum tangential stress acting on the ore pass wall as well as the Rockwall Condition Factor, the ratio of the maximum tangential stress to the uniaxial compressive strength of rock, which is used as a strength criterion for the evaluation of ore pass stability. The same approach has been used by Vieira & Durrheim (2005) for ore pass design in ultra deep levels. However, they used the advantages of boundary element numerical method (Map3D) to estimate the induced stress around the ore pass. They then took into account the effect of other openings on the stress concentration around the ore pass boundary. The same failure criteria (Rockwall Condition Factor) was then used to evaluate the stability of ore passes. Martin et al. (1999) suggest that the initiation of failure in brittle rock occurs when the ratio of the maximum tangential boundary stress to the laboratory unconfined compressive strength (Rockwall Condition Factor) exceeds « 0.4.

Kazakidis & Morrison (1994) attempted to combine global and local numerical modeling of the ore pass #3900 in INCO's Creighton mine and compared the results with actual stability problems encountered in the ore pass. The effect of mining sequence on the ore pass was investigated through three dimensional elastic numerical modeling, using EXAMINE3D software. In addition, the Ubiquitous joint analysis was used to determine the importance of the orientation of a shear zone with respect to the mining induced

25


stress. Finally, to investigate the effect of rock mass blocks, a Distinct Element analysis (UDEC) was conducted.

2.6.2 Consideration of fractures in ore pass degradation analysis

An example of how to use fracture data to investigate the instability of ore passes is presented by Stacey et al. (2005). Through the use of typical statistical fracture distributions for quartzite rocks, fracture traces were generated in orthogonal planes. They subsequently superimposed different ore pass geometries on the fracture traces, thus identifying unstable wedges. In this evaluation, the influence of stress has been taken into account by determining the increased size of the passes at the different depths, as a result of stress induced fracturing of the rock using dog-earing concept. However, the analysis has not taken into account the effects of changing stress conditions and the corresponding changes in rock fracturing that occur due to scaling (stress induced fracturing).

Hadgigeorgiou & Grenon (2005) used distribution of fracture sets orientation, trace length and spacing to generate a fracture system model in order to determine the stability of ventilation raises in fractured hard rocks. Once the fracture system was generated, it was possible to introduce the actual configuration of the ventilation raise into the model. With this model, the authors were able to estimate the number, size and probability of occurrence of wedges at the sides of the raise. Limit equilibrium analysis was used to determine the stability of all individual wedges that act on each raise wall. The effect of stress has been ignored in this analysis.

2.6.3 Consideration of the influence of material flow in ore pass degradation

Material transport in ore pass systems has been investigated using numerical models, and in particular the distinct element models. Flow of granular rock fragments requires large displacements of discrete particles. This behavior is well represented by distinct element methods. The distinct element method was employed by Iverson et al. (2003) and Nazeri & Rozgonyi (2003) to quantify the impact of rock fragments on various ore pass components.

Iverson et al. (2003) used Particle Flow Code (PFC) to simulate the influence of different parameters on impacts of rock fragments on chute gates of an ore pass. They evaluated the effect of particle properties (normal and shear stiffness, friction coefficient) and the angle of the ore pass inclination on the impact force of particles on the gate. The results indicate that the inclination of the ore pass can significantly decrease the impact force of particles on the gate. In addition, higher coefficient of friction of particles and the ore pass walls has results in lower impact force on the ore pass gate.

A two dimensional model was developed by Loughran et al. (2003), to simulate the interaction of stress, structure and impact on degradation of an ore pass wall. They used the continuum to discontinuum fracture finite element approach to demonstrate the

26


damage of an ore pass wall caused by the dynamic interaction of rock fragments against it. The ore pass wall was discretized by finite elements and boundary condition was used to represent in-situ stress state. Two perpendicular rock fractures were taken into account in the model. Six disk fragments of different diameters were fired so that they would impact the wall at a 30° angle. The wall damage induced by rock impact is depicted in Figure 2-8. Although this model took into consideration most of the failure mechanisms, it did not quantify the influence of these mechanisms on ore pass wall degradation.

Figure 2-8. Discretisation and close-up of damage after fragment A impacts the jointed shale, after Loughran et al. (2003).

2.7 Rehabilitation of degraded ore pass

The repair or the replacement of a severely damaged ore pass is a critical decision. In order to make an informed choice the replacement cost plus any costs incurred until re-commissioning must be compared with the cost of repair and its associated delays. Gardner & Fernandes (2006) reported three commonly employed strategies for ore pass rehabilitation.

• Reinforcement of the existing excavation is typically employed in cases where degradation is detected early, or where sufficient pillar exist between the ore pass and adjacent excavations to provide stability. The ore pass is reinforced to accommodate its enlarged size and shape, rather than restored to its original dimensions. The application of this method in several case studies has been documented, including: Gibbon (1976), Schoombee & Van Wyk (1982), Osae & Lange (2003).

• Installation of tube and backfill is generally employed in cases where degradation is severe or ongoing, or where the ore pass has merged with another excavation. The use of a tube means that the ore pass is returned to a size close to, or even smaller than, its original dimensions, usually in the same position. The application of this method at Impala Platinum Mine has been reported by Gardner & Fernandes (2006).

27


• Fill and redevelopment through fill is a variation of complete replacement. This method entails filling the ore pass with cemented waste rock, followed by redevelopment through the fill, usually by raise boring. The application of this method at Doomfontein Gold Mine reported by Gibbs (1979).

2.8 Ore pass longevity

The longevity of an ore pass is measured by the tonnage of material that can safely be throughput the ore pass without necessitating rehabilitation or major reparation. The greater the volumes of material planned to pass through an ore pass, the greater the attention that must be given to the ore pass to keep it functioning during its design life. In addition, the probability of significant degradation is also enhanced. In order to improve the longevity of an ore pass, blockage and degradation rate of the ore pass must be minimized. For this purpose, design considerations, ore pass operating practices and rock conditions, including rock mass strength, structural defects, and stress regime, must be well managed from the conception stage.

Brummer (1998) developed a design chart to predict the life time of an ore pass located in highly stressed ground. This method was developed based on selected case studies, primarily from Sudbury, Canada. He discovered a relation between stress condition (ratio of maximum stress developed in ore pass walls to uniaxial compressive strength of intact rock (cm a x /a c)) and total rock fragments passed in ore passes. This chart, although conceptually valid, does not take into consideration the potential of structurally controlled instability.

More recently, a comprehensive tool to predict the longevity of ore passes has been developed by Hadjigeorgiou & Mercier-Langevin (2008). Based on an iterative analysis, they assigned a rating to all the critical factors that result in a reduction or a prolongation of the operational life of an ore pass. Adverse ground condition (stress regime, rock mass quality, presence of rock structural defects), wall impact factors (material size, fingers and knuckles), and ore pass operation factors (blasting for hang-up release, cushion guidelines) were considered to be the factors which can result in reduction of the operational life time of an ore pass. The longevity extension factors are considered to be ground support and lining of the ore pass. Finally, the following expression has been proposed for the estimation of ore pass longevity:

Ore Pass Longivity (million tons) = 20 (LRF)(LEF) (2 - 2)

Where: LRF: Longevity Reduction Factor includes: adverse ground condition, wall impact factors and ore pass orientation factors and LEF: Longevity Extension Factor includes: ground support and lining.

Hadjigeorgiou & Mercier Langevin (2008) used 13 case studies to calibrate the predictive level of their model. Although they did not explain in detail how they selected different factors for the investigated case studies, they have demonstrated that the Ore Pass

28


Longevity Factor can made acceptable predictions about the actual tonnage of rock that has been passed before failure for the 13 case studies.

2.9 Summary

This chapter has reviewed the ore pass degradation mechanisms and the parameters which can influence the degradation of the structural integrity of an ore pass. Configuration of ore pass systems, rock mass behavior, flow of rock fragments and other factors such as excavation method, reinforcement and flow control are the factors which can influence the degradation of an ore pass. Moreover, when hang-up or blockage occurs in an ore pass, most of the techniques which are used to restore the flow of material can jeopardize the structural integrity of the ore pass.

Review of ore pass failure experiments has shown that the ore pass degradation is a result of the interaction of different failure mechanisms. The effects of the degradation mechanisms can be evaluated by addressing each mechanism individually (i.e. stress induced damage using numerical methods or structural failure using UNWEDGE). Furthermore, the interaction of different mechanisms has been addressed using numerical methods (i.e., stress and fracture modeling by UDEC or Ubiquitous joint modeling using EXAMINE3D). Although these stress models can accommodate the influence of rock structure, there are still important limitations on how to best introduce and interpret fracture behavior in stress models. Moreover, these models do not take into consideration material flow.

Such conventional numerical models rely on a user specified constitutive model assigned to both the intact rock and the fractures. An explicitly defined numerical model for ore pass degradation analysis is one which considers the interaction of all mechanisms and in which behavior of the rock mass is not prescribed a priori by the user.

The main objective of this research is the introduction of a new approach to the modeling of ore pass degradation, with incorporation of stress, structure and material flow. This modeling is achieved via utilization of a distinct element method in conjunction with a fracture system model. Rock mass fracturing is represented by the fracture system model, and the resulting model is subsequently linked to the Particle Flow Code (PFC). The numerical model which is generated by this link is called a synthetic rock mass model. Once force is applied to a synthetic rock mass model, all pre-existing intersecting or isolated fractures in the model can influence the mechanical response of the rock mass. The Particle Flow Code can also simulate the granular flow of rock fragments in an ore pass. This capacity is part of the development of an approach to investigate the dynamic response of rock mass to impact forces from rock fragments.

In order to develop the new methodology based on field data, an ore pass observation and data collection campaign was performed at Brunswick Mine. Several case studies of ore pass degradation were identified and the mechanisms of ore pass failure for each case were investigated. This is discussed in detail in the next chapter.

29

Chapter 3: Ore Pass Performance at Brunswick Mine

3 Ore Pass Performance at Brunswick Mine

3.1 Introduction

The current chapter reports on the Brunswick Mine practices relating to ore pass configuration and performance. All pertinent data related to the design parameters of the ore pass systems at Brunswick Mine were quantified. The data was used to assess the performance of the ore pass systems. This was followed by three case studies of ore pass systems at the mine where the original dimensions of the ore passes were significantly expanded. Detailed investigation of these case studies allowed the identification of important mechanisms which resulted in degradation of these ore pass systems.

A similar approach developed by Hadjigeorgiou et al. (2005) was employed for ore pass data collection at Brunswick Mine. Hadjigeorgiou et al. (2005) and Joughin & Stacey (2005) developed the database of ore pass systems for several Canadian and South African underground mines. One of the objectives of the developed databases was to statistically analyze the ore pass design parameters in the selected mine sites and to find a relationship between the design parameters and the ore pass performance. Consequently, by the assessment of the ore pass performance in the mines, guidelines were developed to improve design and maintenance of the ore pass systems.

This thesis investigated ore pass performance at the Brunswick Mine site. The Brunswick Mine was selected because of the large number of ore pass systems which have been excavated in the mine since it began operation. Moreover, several ore pass related problems, including ore pass instabilities and hang-ups, have been reported by ground control engineers at the mine. A comprehensive database of ore pass engineering and performance was developed for the Brunswick Mine. This was followed by identification and investigation of specific ore pass systems. These ore passes exhibited significant degradation. Preliminary results of this work were presented by Hadjigeorgiou et al. (2008).

In order to collect the data from the mine site, two data collection campaigns were conducted in 2007 and 2008. The author spent 6 months (4 months in 2007 and 2 months in 2008) at the mine site to collect the ore pass design data and other pertinent information including rock properties, joint mapping, etc. Three questionnaires developed by Lessard & Hadjigeorgiou (2006) were used to collect the ore pass data relating to their design, engineering, problems encountered during their operation, and the success of different interventions. All the documentation about the ore pass systems in the mine was also collected. The data collection was complemented by several interviews which were performed with key mine personnel from production, safety and engineering departments.

30


3.2 Brunswick Mine description

3.2.1 Mine location

The Brunswick Mine, operated by Xstrata Zinc, is located 26 km South-West of Bathurst, New Brunswick, in Atlantic Canada, Figure 3-1. It is an underground lead-zinc-copper-silver mine, which is known as the world's largest underground zinc mine. The lead, copper, and silver are by-products.

The mine has been in operation since 1964 and produces almost 10,000 tonnes of ore per day. With the current life of mine plan, the mine will be depleted near the end of 2010. Almost 800 employees work in the mine. The production and maintenance personnel work two shifts per day. The mine operates seven days a week.

Figure 3-1. Brunswick Mine location.

From Bathurst, the mine is accessed via Route 430, Figure 3-2. Bathurst itself is connected to the rest of country by road, rail, and air (Bathurst Regional Airport). To the southeast, Highway 8 connects it to Moncton, Halifax, and the United States. Highway 11 leads northwest to the Quebec border and links to major highways leading to Montreal, Toronto and the rest of Canada.

31


"V/

i ■ hofl

Middle River R l

° Gr»™»e Chamberlain

SitMnMM

Gloucester Junction

Pabmuu Indian Reserve 11

Pabmeau 11

M

S I

Brunswick Mines

Lecresle, * » £ « < " ^ E T T

Mhursl

Bruce

i"î«n

Figure 3-2. Mine site access road, (26 km from Bathurst to the Mine), after Google Map.

3.2.2 Brunswick mine geology

The Brunswick deposit is a volcanic massive sulphide deposit type hosted by metamorphosed volcaniclastic sediments and tuffs which overlie felsic volcanic rocks (Luff, 1977). The ore body consists of close to ten sub-parallel massive sulphide lenses striking North-South and dipping 75° West. The overall strike length of the ore body is 1200 m with a width of up to 200 m. The ore body extends from the surface to a depth of close to 1200 m. The stratigraphie sequence of rocks in the area of the Brunswick Mine deposit has been described by Goodfellow (1975). The sequence consists of meta-

sedimentary rocks, augen schist, siliceous chloritic sericitic schist with associated massive sulphides and overlying iron formation and volcanic rocks, Table 3-1.

32


A longitudinal section and a plan view of geological sequence of Brunswick deposit have been presented in Figures 3-3 and 3-4 respectively.

Table 3-1. Stratigraphie sequence of rock formations near Brunswick Mine, after Coodfellow (1975).

Epoch Group Lithology

Ordovician Quartz-feldespar porphyry dyke

Cambro-ordovician Tetagouche

Volcanic Rocks Siliceous chloritic sericitic schist and iron formation

Augen schist Meta-sedimentary rocks

Section 5S

Looking North across Shaft #2

Mine Dyke

Basalts

HW Sediments

Iron Formation

Massive Sulphide (Ore)

Massive Sulphide (Waste)

Altère d F W S e diments

FW Sediments

QFAS-QAS-CT

Old Sediments

Figure 3-3. Vertical section of the Brunswick Mine deposit (looking North).

33


Sulphide

Poiphyiy Dyke

F on two 11 Sediments

Alteieit Footwoll Sediments

Massive sulphide High Grade

f-N

Figure 3-4. Geological plan of the 1125 Sub4.

McCutcheon et al. (1997) provided a synopsis of the geological stratigraphie and structural relationships of rock units at Brunswick Mine. The rock units were described as follows:

Old sediment rocks; this unit consists of finely-bedded, highly contorted grey to green phyllitic schist, fine to medium grained greywacke, graphitic schist, and siliceous argillite. It is generally barren of economic occurrences of mineralization.

Augen schist, in which the majority of the principal infrastructures (such as shafts and garages) are located. This unit, consists of quartz augen schist, quartz-feldspar augen schist, and crystal tuff. The quartz-feldspar augen schist, which is massive and relatively homogeneous, is composed of plagioclase, potash feldspar and quartz, set in a fine ground mass of quartz, chlorite and sericite.

Footwall Metasediments; this unit consists of finely banded, highly contorted, grey-green chlorite sericite schist, with stringers and disseminations of pyrite, pyrrhotite and calcite. There are minor amounts of chalcopyrite, sphalerite, and galena present.

Massive sulphides; this unit contains not only the ore but also the waste rocks depend on the grade of metals in the rock (less than 7.5% combined Pb-Zn considered as waste). The main massive sulphide minerals are pyrite, sphalerite, galena, pyrrhotite, chalcopyrite, tetrahedrite, and bornite.

34


Carbonate-oxide-silicate iron formation; this unit is located directly above and laterally North and South of the ore deposit, except in areas that have been complicated structurally by tectonic events. The iron formation is characterized by a mineralogical assemblage of siderite, magnetite, cryptocrystalline quartz, and magnesium iron silicates.

Hangingwall Metasediments; this unit consists of light to dark grey, fine grained sedimentary rocks and inter-bedded felsic hyaloclastite with minor massive rhyolite and associated breccias.

Volcanic rocks; this unit overlies the hanging wall felsic rocks and consists of massive to pillowed alkali basalt, pillow breccias and hyaloclastite with minor inter-bedded sedimentary rocks that include dark grey siltstone and red or green in places magnetic, Fe-Mn-rich slate and chert.

The porphyritic dyke; a composite mafic and felsic, quartz-feldspar porphyry dyke cuts the ore body and directed North-South. At surface, the dyke occurs predominantly in the hanging wall rock, but at the deeper levels (below 1125 m level) it occurs in footwall sedimentary rocks. It cuts massive sulphides in the intermediate levels (between 575 m and 1000 m).

Five phases of structural deformation are responsible for the ore body's present form. The phases have led to the presence of folding on the mine wide. Although small scales faulting with local shears are spaced a few tens of meters apart in the deposit, no large scale faulting occurs within the deposit. A larger geological structure that has planar extent in the hundreds of square meters is the mine dyke, (Luff, 1977). The sharp mechanical contrast between the massive sulphide ore body and the host rocks can behave as pseudo-faults under induced stress conditions, (Simser & Andrieux, 1999 and Simser et al., 2002).

3.2.3 Mechanical properties of rock units

The mechanical properties of rock units at Brunswick Mine are very diverse. The massive sulphides are competent, stiff and dense. The contact between the waste sulphides and the ore is generally the result of an economic cut-off limit rather than an actual geological discontinuity, even though the ore is usually more massive and homogeneous, and less structured, than the waste sulphides. On the other hand, meta-sedimentary units are weaker, softer and lighter. They also tend to be highly laminated (particularly those on the hanging wall side). The properties of the dyke also vary widely, from strong in massive horizons, to weak in narrower regions, (Andrieux & Simser, 2001).

Due to the brittleness of the massive sulphides and the dyke, seismicity at Brunswick Mine is generally located within these rock units. The metasediment sequences which are highly laminated usually unravel along the laminations. This unraveling is another mechanism of seismicity in the mine, (Simser & Andrieux, 1999).

35


In reviewing the literature available concerning the Brunswick Mine, one finds that different numbers have been reported for mechanical properties of rock units. Since the 1980s, several laboratory tests have been conducted to determine the mechanical properties of various rock units in the mine. The tests were done in different laboratories and on samples that varied in their composition because they were collected from various locations in the mine. The most recent intact rock properties reported by the Noranda Technology Center (1999) and the results of laboratory tests conducted at the Rock Mechanics Laboratory of Laval University (2007) were used to summarize the mechanical properties of rock units at Brunswick Mine. The mechanical properties of different rock units are listed in Table 3-2. For each rock unit, the average mechanical property and its variation have been reported.

Table 3-2. Mechanical properties of various rock types at Brunswick Mine.

Rock Type Nb.of Samples*

Uniaxial Compressive

Strength (MPa)

Elastic Modulus

(GPa)

Poisson Ratio

Density (t/m3)

Massive Sulphides (Ore) 9

205 (90-266)

104 (52-193)

0.29 (0.17-0.44)

4.27

Massive Sulphides (Waste)

6 252

(221-268) 156

(140-194) 0.23

(0.13-0.3) 4.27

Dyke 9 173

(125-208) 77

(73-80) 0.30

(0.27-0.32) 2.78

Hangingwall Sediments 7

69 (42-102)

83 (63-94)

0.26 (0.19-0.33)

2.91

Footwall Sediments 5

63 (24-110)

55 (42-62)

0.20 (0.14-0.27)

2.80

Chloritic Iron Formation Unknown1 70 60 0.26 2.80

Crystal Tuff 8 91

(32-138) - - 2.79

Quartz-Augen Schist Unknown1 60-75 - - 2.80

*The number of samples used to estimate the UCS of each rock type 1 The number of samples tested for estimation of the mechanical properties of the rock unit was not available.

36


3.2.4 In-situ stresses at Brunswick Mine

The orientation of mine's far field principal stress is sub-horizontal East-West, perpendicular to the overall strike of the deposit. The far field intermediate and minor stresses are sub-horizontal (oriented North-South) and sub-vertical respectively. Based on the in-situ stress measurement tests conducted by Golder Associates in the mine, the gradients of principal in-situ stresses are as follows <xw_5 = 0.044 MPa/m, o~E_w = 0.055 MPa/m and av = 0.028 MPa/m.. As a result of extraction rates of more than 70%, and an evolution toward deeper mining zones, the mine has been managing high stress conditions and a high rate of seismic activities.

3.2.5 Mine operation

The mining method employed at Brunswick Mine is longhole open stoping with delayed backfill. Mine production occurs in six main levels including 425, 575, 725, 850, 1000 and 1125 levels. The mine has two access shafts. The shaft No.3 has been sunk to 1337 m deep to hoist ore, personnel and equipment. The shaft No.2 is 900 m deep and is used for hoisting personnel and supplies.

3.3 Configuration of ore pass systems at Brunswick Mine

Because Brunswick Mine covers a wide area, it has developed and operates a multiple ore pass systems. Historically, the mine has constructed close to 25 ore pass systems with a total length of 7200 m. Currently, only 10 systems are still in operation, with the remaining abandoned either due to depletion of material to be transferred or due to operational failures (critical expansion of the ore pass section). Figure 3-5 is a vertical view of the mine, illustrating the relative location of the ore passes.

An ore or waste pass is comprised of a number of sections of different length. A section is defined as a segment of an ore pass that links two levels and is excavated in one run, Figure 3-6. It is recognized that in a mature mine that has been in operation as long as Brunswick, it is sometimes difficult to provide a complete historical record. Nevertheless, a comprehensive database has been constructed based on mine records and interviews with mining personnel. The present situation of ore pass sections is listed in Table 3-3. For each ore pass system, information pertaining to design, degradation and hang-up problem were gathered. For the purpose of this thesis we are focusing on the data related to design and degradation.

37


OP #19A #18 #21 A, #20 #21 #19

I ml i r commission

Vbaodoocd

Total I < nm h about 7200m

11?5 Level

Crusher #4

Figure 3-5. Schematic layout of ore passes at Brunswick Mine (not to scale).

_Q_

r

Figure 3-6. Representations of an ore pass with reference to its sections, after Mercier-Langevin & Hadjigeorgiou (2004).

38


Table 3-3. Investigated ore passes.

Commissioning Ore Pass Abandoned Ore Pass

Ore Pass #

Length (m) Nb. of Sections Ore Pass

# Length (m) Nb. of Sections

24 257 2 19 260 2

23 181 2 21 136 3

19A 70+50 1+1 (abandoned) 12 274 2

18 261 2 8 506 7

15 408+89 3+1 (abandoned) 9 278 5

1000SFR 551 2 11 52 3

21A 29 1 6 873 20

12A 50 1 3 1895 27

25 143 2 2 511 8

20 47 1 1 293 2

Total 1997 17 Total 5217 81

Various excavation methods have been used to construct the ore pass systems at Brunswick Mine. Table 3-4 illustrates the frequency for each method both as a function of section length and number of sections. The section length associated with each method is also given in Table 3-4. Alimak raising is the most popular excavation method. It is also worth noting that both Alimak and raise boring have been used to develop long sections. The average section length driven by Alimak raise and raise boring methods are 93 m and 129 m respectively. Conversely, the ore pass sections driven by conventional and drop raising methods are short.

39


Table 3-4. Ore pass excavation methods and section length.

Excavation Method

Number of sections

Mean (m)

Std. (m)

Max. Cm)

Min. (m)

Total Length

(m)

Total Length

(%)

Alimak 49 93.2 47.7 207 13 4568 63.3

Conventional Raise 35 26.7 12.1 57 10 935 12.9

Drop Raise 1 29.0 0 - - 29 0.4

Raise Boring 13 129.3 86.6 325 34 1681 23.3

The excavation methods used for construction of the ore pass systems have influenced the cross-section shape of the ore passes. All the ore pass sections with circular cross-section are associated with raise boring while rectangular and square ore pass cross-sections have been excavated using drilling and blasting. At Brunswick Mine, 41 sections had a rectangular shape, 44 were square and 13 sections were circular, Appendix-A, Figure A-l.

The minimum dimension of an ore pass cross-section has an important influence on flow of material in the ore pass, Lessard & Hadjigeorgiou (2006), Stacey & Swart (1997), Hambley (1987). Quite often, however, it is the size of the Alimak platform that dictates the ore pass dimensions. The wider the ore pass dimension, the harder to ensure the ore pass wall stability. At Brunswick Mine the modal class of ore pass dimensions is ranging from 2 m to 2.49 m with mean dimension of 2.4 m and standard deviation of 0.8 m, Appendix-A, Figure A-2.

The selection of a typical length for an ore pass section is influenced by several factors, including the capitalized development practices of an operation, Lessard & Hadjigeorgiou (2006). Operations that aim at minimizing capitalized development will end up driving short ore pass sections, going from one level or sub-level to the next during the time the various levels are entering into production. Local experience is also a deciding factor on selecting a section length. Quite often, mines that experienced problems when driving and operating long sections will subsequently opt to drive shorter sections for new or replacement ore and waste passes. At Brunswick, the length of ore pass sections varied widely from under 25 m to over 300 m. The majority of excavated sections were less than 50 m long, with the mean length of all excavated sections being 73.6 m, Figure 3-7. This is explained by the presence in the database of 3 long sections longer than 200 m each.

40


<fi f f f fi fi fi # fi f fi f fi * * * # fi $ fi 4 fi 4 fi 0 Section-Length (m)

Figure 3-7. Lengths of ore pass sections at Brunswick Mine (98 sections).

The choice for a particular inclination is dictated by the need to facilitate material flow, and, at the same time, slow it down. Shallow sections may restrict flow, especially if a high proportion of fine material is present, while steeper excavations result in higher material velocities and compaction. At Brunswick Mine, ore pass inclination varied between 45° and 90°, Figure 3-8, with the mean inclination being 66.2°. It should be noted that flow is hindered at inclinations of less than 70°. Also the past experience elsewhere suggests that vertical sections result in compaction of the material in the ore pass, Lessard & Hadjigeorgiou (2006).

The majority of the inclined ore pass sections at Brunswick Mine have a trend ranging from 225° to 315°. This implies that most of the ore pass sections have been inclined toward the West direction (parallel with the dip direction of the massive sulphide lenses), Appendix-A, Figure A-3.

41


25

20 C o

a is

u -O E = 10

22 22

45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-90

Section Inclination (°)

Figure 3-8. Inclinations of ore pass sections at Brunswick Mine (98 sections).

The nature and trajectory of rock fragments flow through an ore pass can be significantly influenced by configuration of dump points. Finger raises are the typical excavations which are used to dump the materials into the ore pass. However, using finger raises can result in creation of concentrated impact zones on the ore pass walls which are detrimental to the ore pass stability. Increasing the number of finger raises along an ore pass section can enhance the rate of ore pass wall degradation. At Brunswick Mine 86% of the ore pass sections have at least one finger raise, Appendix-A, Figure A-4.

The orientation and angle of inclination of a finger raise can also influence the extent of impact induced damage zone on the ore pass wall. Mercier-Langevin & Hadjigeorgiou (2004) recommended that finger raises be oriented so as to create an impact zone on the sidewalls of the ore pass rather than on the footwall.

The trends and plunges of all finger raises at Brunswick Mine were measured using the available design plans for ore pass systems in the mine. In order to find whether the dumped materials hit footwall or sidewalls of the inclined ore pass sections, the trend of each finger raise was compared to the trend of the associated inclined ore pass section, Appendix-A, Figure A-5. The results indicate that 33% of the finger raises used at the Brunswick Mine have been oriented in directions which have caused footwall damage on the ore pass walls. The worst cases were along the ore pass #15, #25, #11, #3 and #2.

Plunge of a finger raise can affect the velocity of materials hitting the ore pass wall in the face of junction point and consequently the extent of damage zone on the ore pass wall.

42


Several authors including Ferguson (1991) recommended an inclination angle of 60° and greater for finger raises. However, this recommendation is only to ensure the free flow of rock fragments in the finger raise. The angle of intersection between the finger raise and the ore pass section is an important design parameter which can significantly influence the impact force inflicted by rock fragments on the ore pass wall. If the angle of intersection is defined as the inclination angle of the ore pass plus the inclination angle of the finger raise, at Brunswick Mine this angle varies from more than 110° to less than 170°, Appendix-A, Figure A-6. The majority of finger raises (47%) have an intersection angle of 140° -150°. Based on the analysis which is addressed in chapter-8, this is not an optimal angle.

Of interest is the type of infrastructure used to control flow at Brunswick Mine. As illustrated in Figure 3-9, the majority of ore pass sections at Brunswick Mine are not equipped with flow control infrastructures (54%). This is comparatively less than what is observed at Quebec underground mines, reported by Lessard & Hadjigeorgiou (2006), in which only 21 % of ore pass sections have no control over flow. Chutes with doors or fingers are the most popular flow control infrastructures at Brunswick Mine.

Grizzlies, scalpers and mantels are the three popular block size control infrastructures. At Brunswick Mine the scalpers of 0.91m x 1.4 m are the typical block size control infrastructures used at dump points of ore pass systems, Appendix-A, Figure A-7. High maintenance costs are often associated with scalpers, since the rock blocks can wedge between the bars of scalper, Hadjigeorgiou et al. (2005). Pushing the rock block with bucket can damage the scalper bars which can lead falling of the scalper bars into the ore pass. The intrusion of a scalper bar in the ore pass can result in blockage of material in the pass.

43


None Control chain Chute with chute with chute with ross feeder control chains doors or fingers control chains and arc gate

Figure 3-9. Flow control infrastructure in ore pass sections (98 sections).

Based on the Brunswick Mine ore pass configuration practice the following conclusions can be drawn:

• The results of ore pass section length distribution at Brunswick Mine presented in Figure 3-7 show that more than 50% of the ore pass sections are short in length (< 50 m). Nevertheless, 43% of the ore pass sections in the mine are longer than 70 m. This is against the advice by Hagan & Acheampong (1999) who suggested that the ore pass section lengths should be smaller than 70 m. In addition, the analysis of ore pass design data in time suggests that Brunswick Mine has moved away from designing the short ore pass sections. Most of the ore pass sections which have been designed since the end of 1980s and in 1990s have longer lengths than the mature ore passes designed in 1970s. This may be justified by using the raise boring and Alimak raise excavation methods which replaced the conventional raising at the time. The 1000SFR ore pass is an example of such ore pass systems which comprised of two long sections of 325 m and 226 m and excavated by raise boring.

• Drilling and blasting methods represent 87% of all ore pass sections from Brunswick Mine operation (5532 m length out of 7200 m total length). Using explosives for ore pass construction causes disturbance in the rock mass which can adversely influence the stability of the ore pass systems.

• At Brunswick Mine the presence of high stresses in hard brittle massive sulphide rock mass results in seismic activities. Circular shape ore pass sections are considered more stable in high stress conditions. This design fact, it seems, was

44


considered for the design of some ore pass sections in the mine, particularly for the #19, #19A and #21 ore passes which are located in high stress region.

• It is accepted that hang-ups caused by interlocking arches can be prevented if a minimum ratio of smaller ore pass dimension to maximum rock block size is maintained. Several authors including Hambley et al. (1983), Stacey & Swart (1997) and Lessard & Hadjigeorgiou (2003) recommended that a ratio of 4 would be sufficient for free flow in ore passes equipped by block size control infrastructures and a ratio less than 3 will increase the risk of hang-ups significantly. The mean size for smaller ore pass dimension at Brunswick Mine is 2.4 m while the minimum dimension of scalpers installed in dump points at the mine is 0.91 m. This has resulted in a ratio of 2.6 which is less than the acceptable ratio (3 < D/d <5).

• The majority of inclined ore pass sections at Brunswick Mine have a trend towards the West. The foliated hanging wall and footwall rock masses like Quartz Augen Schists or metasediments have been also oriented toward the West parallel with the massive sulphide lenses with a dip angle varying from 70° to 80°. This has resulted in the development of some ore pass sections semi-parallel to the rock mass foliation planes. Serious expansion was observed in the original dimensions of such ore pass sections.

• The design of a finger raise orientation should be considered with respect to the orientation of the associated ore pass section. It is recommended that the dumping materials hit the side walls of the ore pass instead of the ore pass footwall. This has been respected for most of the finger raises designed at Brunswick Mine. However, there are some critical cases in #15, #21, #11 and #25 ore passes.

• No support system was used for the ore passes at Brunswick Mine. Although the function of rigid support systems like conventional bolts in ore passes is questionable, fiberglass bolts and resin-grouted cables are effective support systems to maintain the ore pass integrity.

The experiences learned from the ore pass design at Brunswick Mine can be useful for the developments of future ore pass systems. Successful ore pass operation, however, not only depends on design parameters but also depend on the manner in which the ore pass is operated. These factors are critical to long term stability and performance of an ore pass, Mercier-Langevin & Hadjigeorgiou (2004).

3.4 Case studies of ore pass degradation at Brunswick Mine

In a mine that has been in operation as long as Brunswick, it is not unusual to note that there have been cases of ore pass degradation. Some of them have been significant and have resulted in intervention in the form of rehabilitation and replacement of ore passes.

45


Among the 98 ore pass sections investigated in this study, degradation status of 69 ore pass sections was impossible to determine. This is due to the lack of information on monitoring of degradation degree for these ore pass sections. The majority of these sections have been decommissioned many years ago. Significant degradation was observed in 13 ore pass sections among which 5 sections have been abandoned, Appendix-A, Figure A-8.

In this section, three case studies of ore pass degradation at Brunswick Mine were investigated in order to determine the possible mechanisms of degradation for each ore pass system. For each case the induced stresses around the ore pass were estimated using Map3D-V50, Mine Modeling (2008), which is a 3D boundary element numerical method. Brunswick Mine uses Map3D for stress analysis of mining sequences. For the purpose of this thesis the mine global Map3D model was updated and the dimensions of the commissioning ore pass systems were integrated into the mined out stopes model, Figure 3-10. Localized models were then constructed and run for the areas in the vicinity of each investigated ore pass system. The stress analysis, for each local model, was carried out based on the latest mined out stope in the area of interest. As input data for the models, the gradient of pre-mining stress states was selected based on the information presented in section 3.2.4. All the input data for the Map3D model is presented in Appendix-A, Figure A-9

Figure 3-10. The Map3D model of the mine with the ore passes, (looking North-West).

46


The stress analysis was then followed by the investigation and observation of rock mass quality encountered around each ore pass along its length. The investigation was completed by ore pass performance considerations.

3.4.1 The 1000SFR ore pass

The 1000 South Fill Raise (1000SFR ore pass) was commissioned in the early 1990s. It was developed as a 3 m diameter, unsupported raise bored ore pass comprising two long sections. The first section extends from the 425 3-sublevel to the 725 3-sublevel, and is 325 m long with an orientation of plunge 65° trend 273°. The second section extends from the 725 3-sublevel to below the 850 2-sublevel and stands 226 m long, with a plunge of 75° at a trend of 125°. It is generally kept full up to the 725 6-sublevel.

The 1000SFR was developed within the "Quartz Feldspar Augen Schist unit". This is a foliated metasediment with a dip of almost 75° and a dip direction toward the West. Figure 3-11 shows the geological map of a section at 725 5-sublevel which represents the original and the dimension of 1000 SFR in 2007. The dip and dip directions of thr fractures have been identified on the map. The yellow and green areas of the map represent respectively, the Quartz Feldspar Augent Schist and Metasediment rock types respectively. Although limited field data are available, the host rock is relatively weak with a UCS of approximately 60 MPa.

1000 SFR (ROADBED)

Actual Size

Original Size

1 Of*eO

Figure 3-11. The geological plan of the 725-5sub and the expanded dimensions of 1000SFR.

47


The ore pass has displayed significant degradation in the upper section (425 3sub to 725 3sub). This has been documented by the mine using CMS, performed by the mine personnel. Figure 3-12 presents a part of the ore pass upper section where the CMS results, performed in 2006, demonstrates significant expansion along the hanging wall and foot wall of the ore pass.

Figure 3-12. Cross sections of 1000SFR ore pass in different levels of the upper section.

The geometry of the 1000SFR was integrated into the mine's Map3D model. The stress analysis around the 1000SFR ore pass was performed by considering two horizontal stress measuring grids in the Map3D model. The horizontal grids were placed in the upper and lower ore pass sections in elevations 2060 m and 1934 m. The stress analysis displays that the rock mass around the ore pass has been destressed by extensive mining. Figure 3-13 represents the results of induced stress distribution around the upper section of the ore pass. For both upper and lower ore pass sections, the 3D stress analysis models suggested that the induced stresses (ai) around the ore pass were along the East-West side walls and were estimated at 60-75 MPa. Although comparatively higher than the strength of the rock mass, there was no significant degradation in the lower ore pass section.

48


Figure 3-13. Induced stress conditions around the 1000 SFR ore pass in 2007.

A closer review of all available data suggested that the relative orientation of the ore pass with respect to the observed structure might have been responsible for the degradation of the upper section of ore pass. The importance of structure has been identified by Stacey & Swart (1997) and demonstrated by Hadjigeorgiou et al. (2005) based on case studies from Quebec mines, Figure 3-14.

a)

\ Uk WI,

U ' / \ \

\ *̂*> u .,-.,.-

Figure 3-14. Case studies of ore pass development with respect to rock bedding: a) favorable orientation; b) poor orientation, after Hadjigeorgiou et al. (2005).

49


Figure 3-15 illustrates where the upper section of the 1000SFR was developed sub-parallel to the host rock foliation and bedding. This poor orientation was accentuated by the induced stresses and by the high impact velocity of flowing particles, which is the consequence of the long ore pass sections. The upper section of the 1000SFR ore pass has a 325 m length and is operating as a pass through ore pass (empty ore pass section). Figure 3-15b demonstrates the foliation planes of the Quartz Augen Schist in the vicinity of the 1000SFR ore pass. The spacing of the foliation planes vary between 30 cm to 1 m with a dip angle of 75°.

Foliation

(QAS)

Section Looking North b) m

Figure 3-15. a) The upper and lower sections of the 1000 SFR with respect to the rock mass foliation, b) Foliation in Quartz Augen Schist, (photo looking North).

On the other hand the lower part of the ore pass was developed in a favorable orientation that has allowed the ore pass to maintain its integrity. Another factor was that it was easier to maintain the lower part of the ore pass full. This was a case where a favorable ore pass geometry and good practice mitigated the problems associated with ore pass wear.

3.4.2 The #15 ore pass

The #15 ore pass was developed in the late 1970s by the Alimak method. It delivers ore to the No. 3 crusher on 1000 3-sublevel. It was developed as a 2.7 m x 3.3 m rectangular,

50


unsupported ore pass comprised of three long sections and one short circular section of 6m in diameter, which served as a silo. The first section extends from the 575 8-sublevel to the 575 4-sulevel, and is 90 m long with a plunge 75° and a trend 274°. The second section extends from the 575 4-sublevel to the 725 1-sublevel, and is 204 m long with an orientation of plunge 62° and trend 257°. The third ore pass section extends from the 725 1-sublevel to the 850 sill and stands 170 m long, with a plunge of 65° and a trend of 298°. The last section is 30 m long and is approximately vertical. A control chain is used between the second and the third sections of the ore pass to control the flow of materials.

The #15 ore pass intersects three different rock types along its length, Table 3-5. The two upper sections were developed in the footwall metasediment, which is a weak (UCS 69 MPa), laminated rock mass with a dip varying between 70° to 80° toward the West. Figure 3-16 is a geological map of the mine at 575 5-sublevel, in which the metasediment rocks are indicated in green color. The dip and dip direction of joints have been delineated on the map. Figure 3-17 shows the metasediment rock mass at 725 6-sublevel with its foliations dipped 70° - 80° toward West and has a spacing vary from less than 10 centimeter to close to 40 cm. The configuration (plunge and trend) of the two upper sections of the #15 ore pass are relatively sub-parallel to the foliation of metasediment rocks. This condition is similar to that of the upper section of 1000SFR. The results of CMS surveys in the two upper sections of the ore pass indicate that the ore pass has been expanded approximately 5 to 9 times of its original dimension, Figure 3-18. The third and last sections of the ore pass are located in the competent massive sulphide rock mass. The results of CMS surveys for the lower sections indicate that some minor degradation has occurred in the lower sections, Figure 3-18.

Table 3-5. Rock units distribution along the #15 ore pass.

From To Rock Type

575-8 sub 575-3 sub Metasediments

575-2 sub 575-1 sub Chlorite Iron Formation

575 sill 725-2 sub Metasediments

725-1 sub 850-1 sub Massive Sulphides

51


Figure 3-16. The #15 ore pass intersects metasediments on 575-5sub.

H B H H ^ ^ ^ ^ ^ ^ ^ B b)l Figure 3-17. Metasediment rock on 725-6sub, photo: a) looking North; b) looking East.

52


► N

Elevation 1*62

850 sW

Figure 3-18. Degradation of the #15 ore pass along its length.

The stress analysis around the #15 ore pass was done using three horizontal measuring grids which were considered in the Map3D model and were placed in the elevations 2116 m, 1994 m and 1873 m respectively. The results of stress analysis obtained from the three different elevations reveals that the rock mass around the ore pass has already been de-

stressed and the ore pass sits, in large part, in the shadow of extensive mining. Investigation of the actual shape of the ore pass, in different elevations, indicates that the ore pass has been expanded uniformly around all the directions or in some elevations mostly along its footwall side (towards the East). As the maximum in-situ stress in the mine is along the East-West direction then it has been expected that a stress induced failure occurs along the North-South direction of the ore pass, perpendicular to the direction of the maximum in-situ stress. Such shape of expansion was not observed along the #15 ore pass. Thus, it can be concluded that the influence of stress induced damage in the degradation of the #15 ore pass was not significant.

The maximum induced stresses (o"i) around the ore pass were along the East-West side walls (due to the stress shadow) and were estimated at 35-50 MPa, Figure 3-19. This magnitude of stress could be critical for the two upper sections of the ore pass, where the strength of rock mass is not competent. However, for the lower sections where the rock mass is quite strong (UCS- 205 MPa), no stress induced failure was expected.

53


Figure 3-19. Induced principal stress (a,) around the upper section of the #15 ore pass (elevation 2116 m), using Map3D elastic model.

Based on case studies at underground mines in Quebec, Lessard & Hadjigeorgiou (2006) suggested that the finger raises are often associated with operational problems such as ore pass degradation. They explained that, in mines where multiple ore producing levels share the same ore pass, material dumped from the upper levels is more likely to damage the walls of the main ore pass. Moreover, the wall damage attributed to impact loading is most often localized at the intersection of finger raises to the ore pass. In practice it is sometimes difficult to differentiate among damage due to impact, wear or the presence of structural defects in the rock. It is most probable that the presence of structural defects in the rock mass accentuates the influence of impact loading, resulting in more pronounced degradation.

In upper sections of the #15 ore pass (575-8sub to 725-lsub), ten finger raises used to funnel rock fragments into the ore pass. Six out of the ten finger raises are oriented in direction in which the dumped rock fragments hit the footwall of the ore pass, opposite the junction points between the ore pass and the finger raises. The interaction of rock mass foliation and the impact induced damage has resulted in significant degradation, particularly in the footwall side of the ore pass. Figure 3-20 displays the extent of the #15 ore pass degradation zone (red lines) in several sub-levels of the upper sections. The finger raises that were used for the dumping of materials are numbered from 1 to 10. The figure indicates the relationship between the ore pass expansion shape and the direction of material dumping into the ore pass. The ore pass expanded mostly along the direction that broken material dumped through the finger raises (mostly toward the footwall).

54


•£>■ ■£=

575 NO.8 SUB P L A M v i e w 500 SCALE

Q

dcrM Jui f t5th.20O4 July 22nd,2004

Sapt 20*1,2005 luty 22nd,2004

575 HO. 7 SUB PLAH VIEW 5 0 0 SCAtC

' iSOP Dump Is not occaaJbf* d u * te fa of ground on tl N».« ■ ,■> U *

8

975 N0.6 SUB PIAM VIEW 5 0 0 SCALE

4=~

GMZZLr WAS PEMOVED BASE IS CONE WTO OP ( G.ft ANC HARD)

575 NO 5 SUB PLAN VIEW 5 0 0 SCALE

0

57S NO 4 SUB PL>N VIEW 5 0 0 SCALE

£=~

r o DUMP GRI22LV A BASE COWE INTO OP (C.BtAKCHAPD)

575 NO 3 S PLAN VIEW 500 SCALE

•£=>■ {£=

w ,GRIZZLY ft OASE

WITH EXHAUST TAN MOUNTED ON TOP STILL THERE AUG. I I In 2004 (DALE PETRIE)

575 NO 2 SUB PLAN VIEW 500 SCALE

©

575 LEVEL PLAN VIEW 5 0 0 SCALE

•£= , . ^ I M T don» 5*p i 25th.2003

0

725 l « . 6 SUB PLAN VIEW 5C0 SCA1£

£>• $ ■

725 N0.5 SUB P U N VIEW 500 SCALE

■£=Nov. 19tti,2003

^

725 HO.4 SUB PLAN VIEW 5 0 0 SCALE

©

7 2 5 N0.3 SUB PLAN VIEW 5 0 0 SCALE

Figure 3-20. The extents of expansion zone around the upper sections of the #15 ore pass developed based on CMS results taken in 2003 and 2004.

55


3.4.3 The ore pass complex 18-21

The other case study is related to the 18-21 ore pass system that serves mining zones 20 and 21. Mining zones 20 and 21 are not advancing directly toward each other, but rather past each other, Figure 3-21. This has resulted in high stress state concentration around the ore pass complex 18-21 and seismic activity nearby.

Zoo*2A

J > N

Figure 3-21. Advancing directions of the mining zones 20 & 21 with respect to the 18-21 ore pass complex, after Andrieux et al. (2006).

By 2004, ore passes #19 and #21 were abandoned due to considerable enlargement of their cross sections (in some elevation they were expanded 3 to 6 times of their original cross section). The ore pass #19 was merged with #19A, which necessitating the backfilling of #19 ore pass in order to prevent further expansion. Undertaken laser cavity surveys demonstrated that damage to ore pass #18 was considerably less than that of the other ore passes. This was explained by the presence of an adequate distance (10 to 15 times the ore pass minimum dimension) between this ore pass and other infrastructure. Furthermore, it was suggested that the expansion of the other ore passes in the vicinity resulted in a stress shadow region for the #18 ore pass. The original and present dimensions (according to the data was available in 2008) of ore pass complex 18-21 are summarized in Figure 3-22.

This ore pass complex was constructed to a large degree in massive sulphide rock, characterized by a competent rock mass. The stress analysis for the ore pass complex was done using two measuring grids placed in the elevations 1582 m, Figure 3-23, and 1554 m. Undertaken stress analyses of this ore pass complex indicated high stress concentration in the North and South walls of the #19, #19A and #2 lore passes. Figure 3-24 presents the distribution of maximum induced stresses (ai) around the ore pass complex in the 1582 m elevation. The maximum induced stresses (ai) in the pillars between #21 and #19A ore passes and between #19A and #19 ore passes were estimated at 140 to 200 MPa. Although the massive sulphide rock mass around the ore pass complex 18-21 is quite strong (UCS- 205 MPa), the magnitude of stresses concentrated in the pillars between the three ore passes #19, #19A and #2lis critical for the stability of the ore passes. However, the #18 ore pass is located in a stress shadow area.

56


1125-4Sub

1125-2Sub

a)

O P . 19A O P . 21

O P . 18 O P . 20

Exhaust Raise

1125-4Sub

1125 2Sub

b)

Figure 3-22. a) Original dimensions of the ore pass complex 18-21, b) Dimensions of the ore pass complex 18-21 after degradation.

Zone 21

Figure 3-23. Stress measuring grid placed in elevation 1582 m for analysis of induced stresses around the 18-21 ore pass complex.

57


Figure 3-24. Induced stress states around the ore pass complex 18-21.

The walls of #19 and #21 ore passes were subjected to high stress conditions and at the same time they were impacted by high velocity of material flowing through these ore passes. The resulting high velocities were in part due to the absence of material flow control mechanisms in these ore pass systems. The logistics of the operation involved dumping material into the #21 ore pass from a dump point located South of the ore pass system. This dumping resulted in material impact on the North wall side of the ore pass, where the higher induced stresses had resulted in greater damage to this zone. Consequently, the combination of material impact and high stresses resulted in expansion of the #21 ore pass towards the #19A ore pass, Figure 3-25. This problem may have been mitigated if a different orientation of the finger raise and ore pass were chosen to prevent the interaction of stress and material impact act on the North wall side of the ore pass. Brummer (1998) has suggested that ore passes should be oriented so that stress induced damage occurs on the sidewalls of the pass, Figure 3-26. Ore pass #21 was inclined toward the southern direction, resulting in both abrasion and wear on the floor of the ore pass.

5 S


1125-4 sub

Dump Point i3> Trend of Ore Pass

Stress Concentration Zones Damage Zone

1125 W0.4 VJ» PUW wr»

Figure 3-25. Expansions of the ore pass # 21 due to interaction of stress and particle impact degradation mechanisms.

Max Principal Stress Direction

Scour and abrasion at bottom of orepass

Max Principal Stress Direction

Stress-induced damage at top 4 bottom of orepass

Scour and abrasion at bottom ot crêpas s

Stress-induced damage in sidewalls of orepass

Figure 3-26. Undesirable orientation for ore pass (left) and preferred orientation (right), after Brummer (1998).

3.5 Lessons and Recommendations

Based on the analyzed ore pass design data, and the reported case studies from Brunswick Mine, the several lessons were gained. It is recognized that these comments are based on ore pass practice over several years and decisions were dictated by a variety of reasons including technical, operational and economic.

59


• Ore pass length: the majority (-75%) of the ore passes that experienced significant degradation problem were longer than 70 m in length. Therefore, design of ore pass section length smaller than 70 m is recommended.

• Ore pass orientation: ore pass sections intersecting foliated rock masses at acute angles (less than 30°) can result in less stable ore pass sections (ie. ore pass #15 and #1000SFR). It is strongly suggested that ore pass sections be oriented to intersect the rock mass foliations and beddings as near to perpendicular as possible. Furthermore, it is highly recommended that ore pass sections be oriented in a way that stress induced damage zones occur along the sidewalls of the ore pass instead of the footwall. This can prevent the interaction of stress and material flow acting on the footwall. Some of the problems reported at ore pass #21 may have been avoided if this was taken into consideration.

• Ore pass location: construction of ore pass systems in low quality rock masses can result in significant stability problems (ie., upper sections of the ore pass #15 have been constructed in weak metasediment rock mass). Ideally, ore passes should be located in the best possible rock mass quality. In addition, in high stress conditions, ore passes should be located in a stress shadow area. For example, the #19A ore pass which was subsequently constructed between the #19 and #21 ore passes resulted in high stress concentrations in the rock pillars between the ore passes by progressing the 20 & 21 mining zones towards the ore pass complex.

• Ore pass inclination: it is recommended that ore passes be inclined in excess of 75°, provided they are kept full during operation. Ore passes that are used as flow-through should be inclined at 65° to 75° to reduce the impact damage on the walls, and facilitate material flow. The inclination angle depends on material size distribution. For coarser material, shallower angles are recommended.

• Ore pass dimension versus particle size distribution: it is recommended that the minimum dimension of scalpers (d) be selected depending on the minimum ore pass dimension (D) to respect the acceptable ratio of (3<D/d<5) and consequently reduce the number of hang-ups in ore passes. This will mitigate the need for intrusive techniques to restore flow which inadvertently damage the ore pass walls.

• Flow control and block size control infrastructures: most of the ore pass sections at Brunswick Mine are not equipped with flow control devices. Installation of control chains, grizzlies and scalpers in ore passes provide better flow control along the ore pass sections and reduce the high speed impact of rock fragments on the ore pass walls.

• Finger raise: design of a finger raise orientation should consider the orientation of the associated ore pass section. It is recommended that material should be dumped so as to hit the side walls of the ore pass instead of the ore pass footwall. This has been respected for most of the finger raises designed at Brunswick Mine. However, there are some exceptions such as: ore passes #15, #21, #11, #25, #3 and #2. In

60


addition, the angle of intersection between a finger raise and an ore pass can have a significant influence on the magnitude of impact loads on the ore pass wall. Some recommendations concerning this issue have been provided in chapter 8.

3.6 Conclusions

This chapter reports on the work undertaken at Brunswick Mine to develop a site specific database of ore pass systems and performance. Close to 25 ore pass systems with a total length of 7200 m were identified which were classified in 98 ore pass sections.

The majority of the ore pass sections (49 sections out of 98 sections) driven by the Alimak raise with square or rectangular cross sections. This is similar to the experience of ore pass configurations in Quebec underground mines, reported by Lessard & Hadjigeorgiou (2006). In addition more than 43% of the ore pass sections at Brunswick are longer than 70 m, a design fact which increases the risk of ore pass degradation because longer ore passes have a greater chance to intersect poor ground zones.

There is a notable difference in support systems and flow control infrastructures between the ore pass database collected for Brunswick Mine and the Quebec mines ore pass database reported by Lessard & Hadjigeorgiou (2006). In Quebec over 92% of ore pass sections were supported in some ways and 79% of the sections were equipped by flow control infrastructures. However, at Brunswick Mine no support system is employed for ore pass systems and only 46% of the ore pass sections have control over the flow of material.

In the context of this chapter three case studies were selected for more detailed investigation of their stability problem. In the first case study, the influence of ore pass orientation with respect to rock structures was determined. It was revealed that in foliated rocks, the ore pass sections intersecting the rock structures at acute angles are more unstable.

In the second case study, the unfavorable orientation of ore pass with respect to the rock mass foliation has been exacerbated by impact of rock fragments funneled to the ore pass via finger raises. Furthermore, the weak to moderate rock mass strength present in the two upper sections of the ore pass resulted in additional expansion of the ore pass dimensions.

The third case study indicated the influence of high mining-induced stress state on degradation of an ore pass complex. However, this situation has been aggravated, in some cases, by both the orientation of the ore pass and the impact of material that has fallen down the ore pass.

The investigated case studies at Brunswick Mine demonstrated that ore pass degradation is complex. It is the outcome of interaction of several failure mechanisms acting

61


simultaneously. In the investigated case studies stress, rock structural defects and flow of rock fragments in the ore pass systems were identified as important ore pass failure mechanisms. In each case study one of these failure mechanisms was the controller of the degradation phenomenon.

This thesis aims to develop a methodology to integrate the interaction of stress, structure and material flow on geotechnical analysis of ore pass systems. The data collected from the Brunswick Mine, and the lessons learned from the reported case studies helped to design the rest of the thesis. The resulting methodology provides an insight into the ore pass degradation in underground mines. This has significant implications for understanding of some of the complex mechanisms that control the structural integrity of ore pass systems.

In the context of this chapter, Map3D model was used to evaluate the distribution of induced stresses in the vicinity of different ore pass systems at Brunswick Mine. Although the Map3D allows the integration of the influence of large discrete discontinuities such as faults into the stress analysis model, it is not possible to incorporate all structural complexity of a fractured rock mass with several numbers of non-persistent fractures into a Map3D model. This is considered as a limitation of boundary element methods.

The methodology aims to develop reliance on the generation of discrete fracture systems to better capture the structural complexity of a rock mass. The resulting fracture system is consequently linked into a distinct element stress analysis called Particle Flow Code (PFC). The PFC was selected as it potentially allows greater flexibility in representing a fracture system and also allows simulation of dynamic flow of granular materials.

The #19A ore pass, located in the 20 & 21 mining zones, was selected as the case study for development of the new methodology. This ore pass has been constructed in a high stress area, in the competent massive sulphide rock mass. The next chapter presents the fracture data collection for the massive sulphide rock mass at Brunswick Mine.

62

Chapter 4: Fracture System Modeling

4 Fracture System Modeling

4.1 Introduction

This chapter presents the fracture data collected for the massive sulphide rock mass at Brunswick Mine. The collected field data was subjected to statistical analysis to determine fracture characteristics including orientation, spacing and size of each fracture set and their statistical distribution. A fracture system model was generated to capture the structural complexity of the rock mass. The code Fracture-SG, developed at Université Laval by Grenon & Hadjigeorgiou (2008a), was used to generate a 3D fracture system model based on qualitative and quantitative field data collected at Brunswick Mine. A calibration process was then followed to validate the generated fracture system model based on the statistical analysis of the field data.

The influence of structure on the stability of ore pass systems can be addressed using rock mass classification methods. The Rock Mass Rating (RMR), developed by Bieniawski (1989), was used by Joughin & Stacey (2005) to find a relation between ore pass degradation and angle of intersection between ore pass and rock structural defects. The Q system, Barton et al. (1974), was extended by McCracken & Stacey (1989) to provide a complete design methodology for raising bored vertical excavation. Hadjigeorgiou & Lessard (2003) report that the Q system can be used as part of a process to establish a threshold to delineate stable and unstable ore pass systems in Quebec underground mines.

If an ore pass intersects large scale (major) structures, it is common to use a series of deterministic tools to evaluate the stability of the excavation. On the other hand, if quality field data on structural characteristics of a rock mass are available, it is advantageous to capture the full structural complexity of the rock mass. This can be achieved by modeling fracture systems for the zone of interest. The fracture systems have been employed successfully in geotechnical analysis of ore passes, Stacey et al. (2005) and Hadjigeorgiou & Grenon (2005). The fracture system modeling that is considered in the context of this thesis, is a more realistic approach to determine the influence of structure on the stability of ore pass systems.

The fracture system which is developed in the present chapter for massive sulphide rock mass, will be linked to the Particle Flow Code, a distinct element numerical model, to develop a synthetic rock mass model. The mechanical behavior of the synthetic rock mass model will be investigated in chapter 6 and the stability of the #19A ore pass constructed in the synthetic rock mass model will be evaluated in chapter 7.

63


4.2 Fracture mapping

The validity of results of numerical models depends on the quality and quantity of the input data. To generate a representative fracture system model of a rock mass it is essential to measure and quantitatively represent the relevant characteristics of fractures. This includes description of the location, orientation, sizes and shapes of all the fractures.

Rock mass properties are three dimensional entities, but observations of rock structure are usually one dimensional (as in boreholes) or two dimensional (when data are collected from outcrops or exposed rock walls). Therefore, the three dimensional properties have to be extrapolated from the one or two dimensional data, based on some assumptions.

Hadjigeorgiou et al. (1995b) provided a literature review of different sampling methods of rock mass. They have demonstrated that the usual scanline mapping is a proper approach to collect all essential fracture data. Details on the scanline mapping approach have been provided by Priest (1993). In reality, it is a relatively simple method to perform in the field. The statistical analysis of data which are gathered by this method is easy.

Fracture data were collected at Brunswick Mine during two mine visits in 2007 and 2008. The fracture data were collected by scanline mapping. The scanlines were stretched along the face wall of underground drifts using a measuring tape. All fractures intersecting the scanlines were taken, to measure their orientation, place of intersection with the scanline and their trace length on the face of excavation wall.

Scanline mapping was undertaken in the massive sulphide exposures in the lower block of the mine in the drifts located at 1000 and 1125 levels. In all six scanline mappings were performed. The specifications of the scanlines are summarized in Table 4-1. The locations of Scanlines #1 to #6 are presented in Figure 4-1, 4-2, 4-3, and 4-4, respectively. The light brown color in the map represents the massive sulphide rock.

Table 4-1. The specification of the scanline mappings was done at Brunswick Mine.

Scanline # Location Length (m) Trend (°) Plunge (°)

1 1125-5sub 26.0 142 00

2 1125-4sub 34.0 015 00

3 1000-2sub 43.5 182 00

4 1125-2sub 11.0 236 00

5 1125-2sub 9.5 152 00

6 1125-2sub 16.5 010 00

64


Figure 4-1. Location of the scanline #1 at 1125-5sub-level.

Figure 4-2. Location of scanline #2 in 1125-4sub-level.

65


is^fiL

Figure 4-3. Location of scanline #3 in 1000-2sub-level.

Figure 4-4. Location of scanlines #4, #5 and #6 in 1125-2sub-level.

A Silva Ranger compass was used to measure the orientation of fractures. For the dip direction measurement a declination angle (the angle between magnetic North and true North) of 20° West was considered. The magnitude of declination angle at Brunswick

66


Mine area was determined using the magnetic field calculator of Natural Resources of Canada (http://geomag.nrcan.gc.ca/apps/mdcal_e.php).

Using scanline mapping for fracture data collection has several limitations. The scanline mapping can lead to underestimation of the fracture intensities because only fractures which are intersected by the tape are considered. In some cases it may be difficult to distinguish whether a fracture is a natural discontinuity or a blasting or stress induced fracture. Furthermore, the access to the rock mass exposure is sometimes difficult if not impossible. This was one of the restrictions encountered at Brunswick Mine. Accessibility to the massive sulphide rock mass exposures near the #19A ore pass sections were restricted either temporarily, due to the working of mine haulage system nearby, or permanently for several reasons including abandonment of the drifts, or the coverage of drift walls by shotcrete. For these reasons only scanlines #5 and #6 were undertaken near the ore pass sections. These limitations have resulted in relatively short scanline lengths. Figure 4-5 and 4-6 shows the rock mass exposure at 1000-2sub and 1125-2sub (scanlines #3 and #1). Another issue in fracture mapping at Brunswick Mine was the presence of support systems (wire meshes) on the wall exposures. In some cases, where there was a risk that reading of fracture orientation angle from the compass be influenced by the magnetic property of support systems, like wire meshes, the fracture orientation was measured in certain distance from the wall surface, Figure 4-7.

Figure 4-5. Rock mass exposure along the Scanline #3.

67

http://geomag.nrcan.gc.ca/apps/mdcal_e.php


Figure 4-6. Rock mass exposure along Scanline #1.

Figure 4-7. Orientation measurement along scanline #6.

68


4.3 Statistical analysis of field data collected at Brunswick Mine

Statistical analyses can be carried out to summarize the raw data which are collected in the field. This enables quantification of fracture orientation, spacing and trace length. By this approach, each fracture set has its own representation of the statistical distributions for the orientation, spacing and size of the fracture population, based on the results of field mapping. For additional information about statistical analysis of fracture data, visit Appendix-B.

4.3.1 Fracture orientation

A total of 151 fractures were mapped along the six scanlines placed on the exposures of the massive sulphide rock mass. Figure 4-8 represents the stereographic projection of the collected fracture data using DIPS-V5.1, Rocscience (2006). The stereonet was employed to identify the number of fracture sets and calculate the mean orientation for each set. A 2% contouring limit was used for identification of each fracture set. The results are summarized in Table 4-2. The K value in the table represents the "Fisher constant" or dispersion factor. The bigger the K value the more clustered the fracture set, Priest (1993). The equation for K value calculation presented in Appendix-B.

Godin (1987) suggested four fracture sets for the massive sulphide rock mass at Brunswick Mine. He classified them as two sub-horizontal and two sub-vertical fracture sets at the site of interest.

In the undertaken mapping three fracture sets were distinguished in the rock mass. The fracture set # 1 and # 2 are relatively sub-vertical while the fracture set # 3 is sub-horizontal, see Figure 4-9. The fracture set # 1 has a strike toward the East-West while fracture set # 2 has a strike toward North-South direction. Fractures that are not belonging to a particular fracture set were defined as random fractures. Only 3.9% of the measured fractures were classified as random fractures. Figure 4-10a and 4-10b shows the major planes of the three fracture sets and the poles of all measured fractures respectively.

69


Figure 4-8. Stereonet constructed from scanlines data.

Table 4-2. Orientation characteristics of fracture sets.

Fracture Set#

Dip O

Dip Direction o

Fisher's K (unweighted)

Variability (68.3%)

(unweighted)

Number of

Fractures Percentage

of Fractures

1 89 007 17 21 68 45.0%

2 89 274 12 25 57 37.7%

3 17 227 57 11 20 13.2%

Random - - - - 6 3.9%

70


Figure 4-9. Fracture sets in massive sulphide rock, Photo looking West.

Figure 4-10. a) Stereonet shows great circles for all fracture sets; b) Stereonet shows poles of all fracture sets.

71


The number of fractures in each fracture set intersected by scanlines, is presented in Figure 4-11. The fracture set #3 is sub-horizontal, therefore few number of fractures (20) belonging to the set #3 were intersected by the horizontal scanlines, particularly when the length of scanline is short (i.e., scanlines #4 and #5). Although limited number of fractures intersected by the scanlines #4 and #5, they are the only scanlines which was possible to consider close to the ore pass complex 19-21. The drift walls in the vicinity of the ore passes are mostly covered by shotcrete or wire meshes.

Errors from fracture sampling biases are presented in Appendix-B. In the present work to account for orientation bias, the mapping was done in different directions. This procedure was considered as sufficient, since the investigated rock mass consists mainly of two vertical fracture sets and one sub-horizontal fracture set.

The distribution of dip angle for all fracture sets is presented in Figure 4-12. The fractures in set # 1 and set # 2 are semi-vertical with dip angles vary from more than 60° to 90°. The dip angle of fractures in the set # 3 varies between 0° and 40° (semi-horizontal).

A significant dispersion of dip direction angles for fractures in set #1 and set #2 is observed. The fractures in set #1 and set #2 are semi-vertical and their dip angles are directed toward two opposite sides. The dip direction of fractures in set #1 varies between 330° and 60° toward the North side and also from 150° to 240° toward the South direction. On the other hand, the fractures belonging to set #2 have the dip direction between 60° to 150° toward East side and also from 210° to 330° toward the West. The dip direction of the fractures in set #3 varies from 180° to 300°. The dip and dip direction of random fractures are arbitrary.

Scanline #1 Scanline #2 Scanline #3 Scanline #4 Scanline #5 Scanline #6

Figure 4-11. Number of fractures intersected by different scanlines.

72


>-u c QJ =J CT 01

80

70

60

50

40

30

20

10

0 L □ Random

DSet#3

■ Set #2

■ Set#l

0-10 >10-20 >20-30 >30-40 >40-50 >50-60 >60-70 >70-80 >80-90 Dip(°)

Figure 4-12. Distribution of the dip angle of fractures.

A Fisher univariate distribution is generally used for modeling the distribution of fracture orientation. For this purpose, the approach proposed by Mardia (1972) and summarized in Appendix-B was used to find the distribution of fracture poles around the mean pole value for each fracture set. Figure 4-13 presents the distribution of fracture orientation for set #1, #2 and #3. This was followed by fitting a bell-shaped normal distribution over the field orientation data. Finally, the Kolmogorov-Smirnov (K-S) test was employed to evaluate if the orientation of fracture sets follows a normal distribution. The test investigates the null hypothesis that the fracture orientation data has a normal distribution, against the alternative hypothesis that the data does not have a normal distribution. A maximum significant level can be selected for the test to accept or reject the null hypothesis. In the present work the maximum significant level of 5% was chosen. The results of K-S test for the orientation data of fracture set #1, #2 and #3 indicates that the normal distribution for the fracture sets orientation cannot be rejected at 5% significant level. The details of the test results are listed in Table 4-6, see page 80. The P-

values in the table indicate the significance levels at which the considered probability distribution is suitable to represent the statistical distribution of data: the higher the P-

value, the better the probability distribution function in representing the distribution of data.

73


o o : S*t* 1

006 V

C M

f0M

7

V

\

o o ;

/ /

V

\ V

001

/ /

V

\

V 0 r~

/

V

\

V 0

3 > 10 18 20 28 »

tas 0 048

/

H o ^ i r , 0 048

/

004

0038

£ 003 /

\ 004

0038

£ 003

J /

\ s

/

/

002

0 018 /

/

\, J

/

/

\ N 0008

0 /

/

/

\ N

9 . 10 18 20 28 30 M 40

Orientation difference f ) b) Orientation difference (°)

0 09

0 03

0 07

006

% 005 c Q 004

0 03

oo; 0 01

o

' ' S « t # 3 "■ Normal Ft

S « t # 3 "■ Normal Ft

■

\ \

c) 5 10 15 20

Orientation difference (°)

Figure 4-13. A normal distribution fit over the histogram of the fracture sets orientation data for a) set #1, b) set #2 and c) set #3. Orientation difference is the angle between the mean pole of

a fracture set and a given pole of a fracture belong to the same fracture set.

4.3.2 Fracture spacing

Total spacing of fractures determined from the scanline mappings is presented in Table 4-

3. The frequency plot for the total fracture spacing values obtained from the mappings is presented in Figure 4-14. The results indicate a characteristic of the exponential distribution. The exponential distribution for total fracture spacing has been also reported by Priest & Hudson (1976), Baecher et al. (1977), Kulatilake et al. (1993). A Kolmogorov-Smirnov test was used to determine if the total fracture spacing data follows an exponential distribution. The result of KS test indicates that the null hypothesis that the total spacing data has an exponential distribution cannot be rejected at 5% significant level.

74


Table 4-3. Total spacing values for scanline mapping data.

Scanline Mapping Length (m) Total Fracture Spacing (m)

1 26.0 0.97

2 34.0 1.10

3 43.5 0.75

4 11.0 1.63

5 9.5 1.04

6 16.5 1.06

Combined data 140.5 0.90

0,5 1 1,5 2 2,5 3 3,5 4 4,5 Total Spacing (m)

Figure 4-14. A histogram for total fracture spacing from scanline mapping data.

The fracture set spacing was also measured along each scanline. This was followed by calculation of the normal set spacing for each fracture set along each scanline, using the method discussed in Appendix-B. For this purpose the acute angle between the scanline orientation and the orientation of the line normal to the mean orientation of the fracture set in question was determined. Finally, the normal set spacing was obtained using the

75


equation (B-4) in Appendix-B. The results of fracture set spacing and normal set spacing are listed in Table 4-4. Based on the ISRM suggested methods, Brown (1981), the fracture sets in the massive sulphide rock mass can be described as widely spaced.

The results of set spacing and normal set spacing indicate that there is a minor difference between the mean set spacing and the mean normal set spacing for the fracture set # 1. On the other hand, the difference between mean normal set spacing and mean set spacing of fracture set # 2 and set # 3 are significant. This can be justified by the orientation of the scanlines used for the data collection. Most of the scanlines are oriented in North-South direction relatively normal to the fracture planes of set #1.

The normal fracture frequency for each set is the reciprocal of the value of mean normal set spacing, Table 4-4. This value is used to validate the volumetric fracture intensity of the generated model.

Table 4-4. Spacing characteristics of the fracture sets.

Fracture Set #

Mean Set spacing (m)

Mean Normal set spacing (m)

Mean normal set frequency (m1)

1 1.7 1.52 0.65

2 2.1 1.12 0.89

3 5.2 1.23 0.81

Although the total fracture spacing data follow an exponential distribution, the set spacing and normal set spacing of fracture sets may have different or the same distribution type. The frequency plot for the normal set spacing values of the sets #1, #2 and #3 are presented in Figure 4-15. The negative exponential distribution was fitted against the normal set spacing values. The Kolmogorov-Smirnov goodness of fit test was conducted separately for normal set spacing data belonging to each fracture set, Table 4-6, see page 80. The results of K-S test indicate that the negative exponential distribution is a best fit to represent the normal set spacing of the three fracture sets.

76


S 04

03

02

0 1

0,

«t»cin9_««1 dat. Ei<x»Mitf

V

\ ^ \

1 2 3 4 5

0 9 &VOMT4W

0 8

07 "

C S 04

03 \ 02 ■

01 ^ ^ ^ ^ ^

a) Spacing (m) b) Spacing (m)

C) Spacing (m)

Figure 4-15. An exponential distribution fitted over the histogram of normal set spacing values, a) set #1, b) set #2 and c) set #3.

4.3.3 Fracture trace length

The trace length of fractures on the rock mass exposures along the scanlines was measured using a 5 m steel measuring tape. A few small fractures (5 in total) with trace lengths smaller than 0.5 m were intersected by the scanlines. It was difficult to recognize if these fractures are natural or blasting induced fractures. Finally, it was decided to disregard these small fractures in the fracture mapping process. This decision has resulted in a truncation limit in the collected field data. As the numbers of these small fractures compared with the total numbers of measured fractures is small (5 versusl51), no correction was applied for this bias in estimating mean trace length of fracture sets. All sampling errors from biases and the methods of bias correction have been discussed in Appendix-B.

Three types of fracture traces were observed on the finite exposures of the sampling drifts in the mine, intersected by the scanlines, including (a) both ends of fracture trace visible, (b) one end of fracture trace censored and (c) both ends of the fracture trace censored.

77


More than 96% of the recorded fractures had two defined terminations. This comprised of 94% for fracture set #1, 97% for fracture set #2 and 100% of the fracture set #3. Therefore, the effect of trace length sampling biases (censoring bias) on the estimated mean trace length is not significant. Consequently, no attempt for correction of fracture censoring bias was undertaken.

The results of mean trace length and the standard deviation of each fracture set are summarized in Table 4-5. According to the ISRM suggested methods, Brown (1981), the fracture sets characteristics in the massive sulphide rock mass can be described as low persistent.

The frequency plot of the fracture trace lengths for the fracture set # 1, #2, and #3 is presented in Figure 4-16. The measured trace lengths belonging to each fracture set mapped on the walls were subjected to Kolmogorov-Smirnov goodness of fit test to find the best probability distribution to represent the statistical distribution of the observed trace lengths. Lognormal distribution turned out to be the best fit for the fractures trace length distribution. The lognormal distribution of fractures trace length has been also reported by Baecher et al. (1977) and Villaescusa & Brown (1992). The results of the Kolmogorov-Smirnov test are listed in Table 4-6, see page 80.

Table 4-5. The mean trace lengths of fracture sets.

Fracture Set #

Number of fractures

mean trace length (m) Std. (m)

1 68 1.81 0.70

2 57 1.69 0.54

3 20 1.23 0.22

Villaescusa & Brown (1992) suggested that the size bias (probability of fractures of intersecting the rock surface) has an influence on the simulated average fracture size in 3D fracture system models. Since in the following analysis, only the measured and simulated trace lengths were compared, not the fracture sizes, the correction of this aspect of the size bias is of minor interest for the present work and was disregarded.

78


\

Trace_*W»1 dat* LognomMlF*

/ \ \

j

\

\

\

\

\

\

\ 1 !

/

\

\

\ S

" 1 !

/

\

\

\ 1 !

/

\

\

"~ " t a)

18 2 25 Trace Length (m) b)

0 8 r rK«_i«t»2daU LognofrnM Fit rK«_i«t»2daU LognofrnM Fit

0 7

\ 0 6

\ ^ 0 5 IS

& 04 J

i \

\ 0 3

/

■

/ 0 2 /

0 1 i /

y

° y ° 15 1 15 2 25 3 3 !

Trace Length (m

2 Traca_s«t « data

1 8

/ LognomwIFl \

1 6 / 1 4 / \

>.12

/ /

0 8 / /

0 6 / . 0 4

\j \ j 0 2 \j L ^

c) Trace Length (m)

Figure 4-16. A log-normal distribution fitted over the histogram for trace length of fractures in a) set # 1, b) set # 2 and c) set # 3.

79


Table 4-6. Results of Kolmogorov-Smirnov goodness of fit test for distribution of sampled fracture sets characteristics.

Fracture Set # Parameter Distribution K-S test P-value

1

Orientation Fisher Univariate Accepted 0.21

1 Spacing Negative exponential Accepted 0.08 1

Trace Length Lognormal Accepted 0.82

2




3




In summary, the results of statistical analysis of field data collected at Brunswick Mine identified three fracture sets. These fracture sets were classified as two semi-vertical sets which are relatively perpendicular (set #1 and set #2), directed along East-West and North-South, and one semi-horizontal fracture set (set #3) which has less dispersion compared to the other fracture sets. Table 4-7 provides a summary of the structural characteristics of the massive sulphide rock mass. The statistical distribution of the orientation, spacing and trace length data for the three fracture sets were tested using the Kolmogorov- Smirnov test. The results of the K-S tests indicated the Fisher univariate distribution for the fracture sets orientation data, the exponential distribution for the fracture sets spacing data and the lognormal distribution for the trace length data of fracture sets. The information on the structural characteristics of the rock mass was employed for generation of a fracture system.

80


Table 4-7. Summary of fracture set characteristics measured in the field.

Fracture set#

Fracture Characteristics

Fracture set#

Orientation Normal Set Spacing Trace Length Fracture set#

Dip(°) Dip

Direction o K Mean (m) Std.

Linear frequency

(m l) Mean (m) Std.

1 89 007 17 1.52 1.8 0.65 1.81 0.73

2 89 274 12 1.12 1.0 0.89 1.69 0.54

3 17 227 57 1.23 0.8 0.81 1.23 0.22

4.4 Fracture system modeling

An important development in representation and visualization of rock mass has been the development of stochastic models for representation of fracture systems. These models have found many applications in rock engineering field including mining, civil, environmental and reservoir engineering.

The fracture system model, which is also known as Discrete Fracture Network (DFN), is developed based on specific relationships between fracture characteristics such as orientation of fracture sets, fracture shape, size, and termination, Dershowitz & Einstein (1988). The mean values and statistical distribution of these fracture characteristics are generally evaluated from analysis of field mapping. In each model a particular combination of these parameters is considered. In general terms, the main difference between the various fracture system models is a function of the way fracture characteristics are considered.

Various conceptual models have been developed to provide the geometry of fracture systems. Dershowitz & Einstein (1988) discussed the concept of stochastic fracture system models. They demonstrated that fracture system models can be used to describe all the fracture characteristics as an entity. Moreover, they present detailed descriptions of the Orthogonal, Baecher, Veneziano, Dershowitz and Mosaic Tessellation models. Some other conceptual models that have been developed since and can be used for the modeling of fracture geometry were reported by Staub et al. (2002).

The developed fracture system models assume that fractures are planar. The shapes of fractures in most of the developed models are circular, rectangular or polygonal, mostly for convenience purpose since the real shape of fractures cannot be fully known. In majority of the models fracture locations are defined stochastically by a Poisson process. Fracture size refers to trace length of fractures on the exposed surface of rock or to the

81


surface area of individual fractures. Fracture sizes are usually stochastic either specified directly or indirectly through stochastic location and orientation. Bounded fracture size implies that the fractures smaller than the region under consideration can be represented. Fracture termination is recognized in most of the models. Fractures can either terminate at the intersection with other fractures or against intact rock. Co-planarity implies that a number of fractures can be located in the same plane.

Staub et al. (2002) summarized the applicability, advantage and disadvantage of the fracture systems models. The majority of the models have not been adequately verified for several engineering applications. In practice, the choice of a model will depend on how it can be related to the available field data and to the engineering needs of a project, Grenon & Hadjigeorgiou (2008b).

On the other hand the reliability of a fracture system model depends on the quality of mapping and sampling. As there are several limitations in sampling process, hence the adequacy and reliability of a constructed fracture system model is difficult to be evaluated, Jing & Stephansson (2007).

Several fracture system generators have been developed in the recent years with varying of complexity and ease of use. FracMan, (Golder, 2005) is a commercially available code which allows one to model a wide range of fracture systems. On the other hand some codes can model specific, such as Stereoblock which is based on the Baecher model, Hadjigeorgiou et al. (1995a). For generation of a fracture system model for massive sulphide rock mass at Brunswick Mine the Fracture-SG code was used. This code has been successfully used for slope stability analysis, Grenon & Hadjigeorgiou (2008b).

4.4.1 Model generation

The Fracture-SG code, developed by Grenon & Hadjigeorgiou (2008a) was used for generation of a fracture system model (FSM) for the massive sulphide rock mass at Brunswick Mine. This code is based on the Veneziano model. Other details about the Fracture-SG code has be presented in Appendix-B. Figure 4-17 summarizes a methodology for using a fracture system model. The Figure is based on the Stereoblock model, Grenon & Hadjigeorgiou (2003a), which relied on the Baecher model. The Fracture-SG code used in this thesis uses the same methodology but instead use the Veneziano model to generate fracture system. In this code, statistically analyzed field data including the information about fracture sets orientation, trace length and spacing are employed as necessary input data for fracture system generation.

82


îi n 5

f l II

Back

Discontinuity position along the scanline

*

*i

♦ N ,

I Borehole axis Discontinuity

Orientation (')

Spacing (m) Trace length (m)

l] OJ o

Figure 4-17. Methodology of the Fracture-SG, after Grenon & Hadjigeorgiou (2003a).

A model of 40 m x 40 m x 40 m was generated based on the field data obtained at Brunswick Mine and summarized in Table 4-7. Figure 4-18 demonstrates the fracture system model with the Y axis representing North.

83


Figure 4-18. Visualization of the generated fracture system.

The fracture system generation is an iterative process. The process is repeated until statistical agreement is reached between the field data and simulated data. For the current work several fracture system models were generated to arrive with an acceptable fracture system model which met all the acceptable criteria. Consequently, the validated fracture system model was accepted as a plausible representation of the in situ fracture network. Figure 4-18 shows one possible representation of the fracture system, populated with 72816 fracture polygons. The fracture set #1, set #2 and set #3 were identified with red, blue and green color, respectively.

4.4.2 Validation of the fracture system model

In order to quantify the characteristics of the generated fracture system model and compare it with the field data, six scanlines parallel to the orientation of the scanlines used for the field data collection were introduced into the model. The orientation (dip and dip direction), trace length and total spacing of the fractures intersected by the six scanlines were measured. Figure 4-19 presents the corresponding stereonet of the sampled fractures by the six scanlines in the model. For each fracture set, the angles between the orientations of the fracture poles and the orientation of the mean pole were measured, using the equation B-3 in Appendix-B. Finally, the distribution of fracture poles around the mean pole value was determined for each fracture set, Figure 4-20.

84


The validation procedure comprises of comparison of field input data and the information resulting from simulated model. The distribution of fracture characteristics for simulated fracture system model is derived and compared to the distribution of field observed data, using a Kolmogorov-Smirnov test. This includes the distribution of fracture orientation, spacing and trace length of the simulated model.

Figure 4-19. Stereonet constructed from the six scanlines introduced in the fracture system model.

The comparison of the distribution of the fracture sets orientation for the field and simulated data are presented in Figure 4-20. A two-sample Kolmogorov-Smirnov test was employed to compare the distribution of the values of the fracture sets orientation for the simulated and field observations. The null hypothesis is that the field and simulated data are from the same continuous distribution. The alternative hypothesis is that they are from different distributions. At 5% significant level, the results of the Kolmogorov-Smirnov test, summarized in Table 4-8, confirm the null hypothesis for the orientation data of the three fracture sets.

85


8'

FMd data-Sat t 1 FSM data-Sat #1 FMd data-Sat t 1 FSM data-Sat #1 FMd data-Sat t 1 FSM data-Sat #1

a) Orientation difference (")

0 05

0045

004

0035

, 0.03 i

0025 I

002

0 015

001

0005

0

b)

- FiaW dat»Sat r z FSM data-Sat »2


8

01

0 09

0 08

0 07

0 06

005

004

0 03

0 02

0 01

0

' Fiald data-Set 03 FSM date-Sat * 3

-

Fiald data-Set 03 FSM date-Sat * 3

-

Fiald data-Set 03 FSM date-Sat * 3

- -

c) 5 10 15 20


Figure 4-20. Comparison of the distribution of the fracture sets orientation obtained from the field and the simulated data for a) set #1, b) set #2, c) set #3. Orientation difference is the angle

between the mean pole of a fracture set and a given pole of a fracture belong to the same fracture set.

The six scanlines used to measure the fracture orientations were also employed to characterize the trace length of fractures in the fracture system. The trace length of all fractures intersected by each scanline was measured.

To compare the distribution of fracture trace length in the model with those obtained from the field data, the same truncation limit considered in the field mapping was taken into account for the simulated data. The fractures with trace length less than 0.5 m were disregarded from the data collected in the model. Figure 4-21 compares the distribution of fracture trace lengths obtained from the field and simulated data. Consequently, the Kolmogorov-Smirnov goodness of fit test was used to investigate if the distribution of fracture trace length for the model and the field is the same. The raw fracture trace data from the field and the model were compared. The K-S test results for trace length of fractures are listed in Table 4-8, see page 91. The results of the K-S test confirm the null

86


hypothesis indicating that the field and simulated fractures trace length have the same distribution for the three fracture sets.

a) 05 1 15

Field dataSet al FSM dataSet «1

25 3 35 Trace length (m)

0 t ' ^—FialddataSetaj

— F S M dataSet «2 ^—FialddataSetaj

— F S M dataSet «2

I f

05

a? |

0 4

05

a? |

0 4

J ■

■

1 ■

1 ■

1 ■

1 ■

1 b)

25 Trace length (m)

I f . FteW aaaS*a #3

— FSM dataSet K3

' J

1

Q

0(

04

02

0 l . l I 0 05 1

Trac e length (m) 25 3

c)

Figure 4-21. Comparison of the distribution of the fracture sets trace length obtained from the field and simulated data for a) set #1, b) set #2, c) set #3.

Total spacing of fractures in the fracture system model was determined using the six scanlines introduced into the model parallel with the orientation of scanlines in the field. The frequency plot for the total fracture spacing values obtained from the simulated model is presented in Figure 4-22. The distribution of total spacing of fractures obtained from the field data (Figure 4-14) was compared with the total spacing data from the model, using the Kolmogorov-Smirnov test. The null hypothesis which indicating that the total spacing of fractures for simulated and field data have the same distribution was not rejected at 5% significant level.

87


160

u c OJ =! CT OJ

0,5 1,5 2 2,5 3 Total Spacing (m)

3,5 4,5

Figure 4-22. A histogram for total fracture spacing obtained from the simulated model.

To measure the values of normal set spacing for the fracture system model, 15 additional scanlines were introduced into the model. For each fracture set five scanlines normal to the mean orientation of the fracture set in question were considered, Figure 4-23.

«m*^P

15 10 .5 "

a) s 10 15" 2 0 ' " »

■ b)

: ■ :

15 10

5

0

-5 10 15

O 10 "~~~——— .10 °

20 20

Figure 4-23. Scanlines introduced normal to the orientation of each fracture set, a) set #1, b) set #2 and c) set #3.

88


The normal spacing was measured for each fracture set along the five associated scanlines. The distributions of normal spacing data obtained from the field mapping were compared with those from the fracture system model, using the Kolmogorov-Smirnov test. Figure 4-24 demonstrates the distribution of normal set spacing for the three fracture sets obtained from the field and simulated data. The results of the K-S test, presented in Table 4-8, indicate that for the fracture set #1 the null hypothesis which states that the normal set spacing of the field and the simulated data have the same distribution is rejected in 5% significant level. However, the results of the K-S test for the set #2 and set #3 confirm the null hypothesis.

I

Field dataSet n 1 PSM dataSet «1

1 1

a) Noonal set spacing (m)

' ^—Fie*ddata3e l«2 FSM data Set «2

1

• !

C Q

| • . 02

i l

1 1 1

b) Normal set soacinq (m)

C)

■

— Field dataSet «3 > 3 . ' C M » ' «

1)5

1

Î 0 3

02

1

.' .' J 3 4 5 5

Norma set spac ing (m)

Figure 4-24. The probability density function for the fracture set normal spacing obtained from the field and simulated data; a) set #1, b) set #2 and c) set #3.

89


Table 4-8. Results of Kolmogorov-Smirnov goodness of fit test to validate the simulated fracture network.

Fracture Set # Parameter K-S test P-value

1

Orientation Accepted 0.78

1 Spacing Not-Accepted 0.039 1

Trace Length Accepted 0.63

2


2 Spacing Accepted 0.35 2


3


3 Spacing Accepted 0.6 3


The structural properties of the generated fracture sets were compared with the field data. The detail results of validation process presented in Table 4-8 which corroborate that 8 out of 9 distributions are statistically similar and the simulation can be considered acceptable. The P-values in the table illustrate the significance levels at which the probability distribution of data obtained from the simulated model is suitable to represent the statistical distribution of the field data.

4.5 Summary and conclusions

This chapter summarized the fracture data collected at Brunswick Mine. A statistical analysis of the collected fracture data for the studied region identified three fracture sets including two sub-verticals and one sub-horizontal fracture set. Based on the results of statistical analysis for the fracture spacing and trace length obtained in the field, the structural characteristics of the massive sulphide rock mass can be described as widely spaced with low persistent.

A fracture system was generated for the site at Brunswick Mine based on the fracture sets characteristics estimated from the field data. The orientation, spacing and trace length of the three fracture sets obtained from the statistical analysis of fractures were used as input data for the Fracture-SG code, to generate a fracture system. The generated fracture system was validated through a calibration process by comparing the distribution of

90


fracture sets characteristics obtained from the model and the field data. Finally, a fracture system which has resulted in statistical agreement between the field and simulated data was selected as the acceptable system.

In the next chapter, synthetic rock mass model is presented based on incorporation of a fracture system into a stress analysis distinct element numerical model. In chapter 6 the generated fracture system was employed to construct a synthetic rock mass model for the massive sulphide rock mass at Brunswick Mine. This synthetic rock mass was then characterized for its structural and mechanical behavior. In chapter 7 stability of the #19A ore pass section constructed in the massive sulphide rock mass at Brunswick Mine and represented by a synthetic rock mass model was investigated.

91

Chapter 5: Synthetic Rock Mass Model

5 Synthetic Rock Mass Model

5.1 Introduction This chapter presents an approach in simulating fractured rock mass behavior. This includes a more realistic incorporation of both the fracture network and solid rock matrix behavior. A synthetic rock mass model is defined as a hybrid numerical model which is constructed by linking the Particle Flow Code (PFC), a distinct element method (DEM) developed by Itasca (2008a), and the fracture system. The synthetic rock mass model attempts to simulate intact rock failure together with fracture movement. The input parameters for a synthetic rock mass are intact rock properties, fracture properties and geometrical characteristics of fractures representing by a fracture system. The intact rock is simulated by an assembly of particles bonded together at their contact points. A calibration process is followed to assign the mechanical properties of intact rock to the bonded particle model. The fractures are represented by a smooth joint model which allows particles to slide through one another along the fracture plane. Thus, the behavior of a synthetic rock mass depends on the combined behavior of the solid rock matrix and the embedded fracture network. Loading of a synthetic rock mass allows for a fully integrated analysis of interaction of stress and structure. Depending on the imposed level of stress and strain, new cracks can initiate and develop in the rock bridges between the pre-existent fractures along with the sliding and opening of pre-existent fractures can occur.

In the following sections of this chapter, a discussion is presented on the numerical modeling of fractured rock masses, with particular attention to the distinct element method. The possibility of linking different distinct element stress analysis codes with the fracture system model is reviewed and the selection of the Particle Flow Code to link with the fracture system model is justified. This is followed by a presentation of the main characteristics of the Particle Flow Code (PFC), where procedures of intact rock and rock fracture simulation in the 2D and 3D PFC models are discussed. Finally, the steps of generating 2D and 3D synthetic rock masses are explained.

5.2 Numerical methods for fractured rock mass simulation

A rock mass is largely discontinuous, anisotropic and inhomogeneous and non-elastic, Harrison & Hudson (2000). Numerical modelling of a fractured rock mass is difficult due to the complex and non-homogeneous geological conditions of a rock mass. According to Jing (2003), the complex combination of fractured rock mass constituents makes it a

92


difficult material for mathematical representation using numerical modelling. Starfield & Cundall (1988) suggested that numerical models can simplify complex real problems and enhance the understanding of mechanisms that govern them. A numerical model does not have to be complete and perfect; it only has to be adequate for its purpose, Jing (2003).

The excavation of an underground opening in a fractured rock mass can result in a series of complex perturbations in the surrounding rock mass; involving modification of the stress states, the deformation and fracturing of solid rock matrix, the displacement of large individual rock blocks and the possibility of pre-existent fracture moving relative to one another. Hoek et al. (1990) suggest that a numerical model does not have to represent such processes in their entirety, but rather that the objective of the analyst is to determine which process needs to be explicitly considered and which can be represented in an average way.

In recent decades, a number of numerical methods have been developed to simulate the mechanical behavior of rock masses, with a particular interest on numerical modeling of fractured rock masses. Therefore, it is essential to fully understand the varying applications, assumptions, limitations and advantages inherent in each of the numerical methodologies.

Two main approaches are employed in the numerical modeling of rock masses. Based on the deformation of a rock mass under applied external loads, the rock mass can be treated as being either continuous or discontinuous. The most commonly employed continuum based numerical models are the finite element method (FEM), the finite difference method (FDM) and the boundary element method (BEM). The basic assumption in continuum numerical models is that the material is continuous through the loading process and it cannot be torn or broken into pieces, Jing & Stephansson (2007). The large-scale displacement or macroscopic slip cannot occur in a continuum model of a fractured rock mass. This implies that, in a continuum model, deformation along or across the fractures has the same order of magnitude as that of the solid rock matrix near the fractures. Some of the popular continuum codes which are used in rock mechanics problems are: BEM: Map3D, Mine modeling (2008) and Examine 2D and 3D, Rocscience (2008), FEM: Phase2D, Rocscience (2007) and FDM: FLAC, Itasca (2008c).

On the other hand, the discrete element method is the most generally adapted discontinuum based numerical method. In a discrete element model the problem domain is treated as an assembly of independent elements. The equation of motion is solved for all the individual elements based on the continuous evaluation of contacts between the independent elements. For a fractured rock mass simulation the independent elements can be the rock blocks formed by connected fractures in the problem domain. Large displacements due to the motion of individual rock blocks are straightforward in the discrete element models. This includes block rotation, fracture opening and complete detachment of a rock block from its original domain, Jing (2003). The most representative explicit discrete element modeling is the distinct element method (DEM). UDEC, Itasca (1999) and 3DEC, Itasca (1999) are popular DEM computer codes for 2D and 3D analysis of rock mechanics problems.

93


Problem scale and fracture system geometry influence the choice between continuum and discrete methods. Figure 5-1 presents the suitability of the continuum methods or discrete element methods for analysis of an opening in different rock mass structural regime, Brady (1987). The mechanical behavior of a solid rock mass can be investigated using a continuum approach. Moderately fractured rock masses, where displacements of individual rock blocks are possible, are better described by a discrete approach. Discrete element modeling allows for a more refined consideration of the role of fractures within the rock mass. Finally, a highly fractured rock mass can be assumed to behave like a continuous body, thus a rock mass of this type could also be treated by continuum models.

Sets of discontinuities

Persistent discontinuities

Figure 5-1. Suitability of different numerical methods for analysis of an excavation in a rock mass, a) continuum methods, b) either continuum or discrete methods, c) discrete methods, d)

continuum method with equivalent properties, after Brady (1987).

The combination of continuum and discontinuum methods for numerical modeling of rock mechanics problem is possible by using hybrid models. Hybrid methods are used to eliminate the undesirable characteristics of continuum and discontinuum methods while maximizing their advantages as many as possible, Jing (2003). Different hybrid codes have been developed for rock mechanic problems by linking BEM/DEM, FEM/BEM and FEM/DEM and other hybrid models. ELFEN, Rockfield (2008), is a hybrid FEM/DEM code which has found application for fractured rock mass simulation.

Although all the continuum, discontinuum and hybrid models can be used for simulation of fractured rock masses, there are important limitations on how to best introduce and interpret fracture characteristics and behavior in the models. Most of the continuum

94


models only allow the accommodation of the effect of limited number of major rock structures in the model. However, the complex fracture network can be integrated into the discrete models and in some hybrid models. Grenon et al. (2001) provided one of the earlier examples of linking a fracture system to a distinct element model. In a modeling of an underground drift, they used the comprehensive field data to generate a 3D fracture system. They extracted a series of 2D trace planes, along the simulated rock mass, and they introduced these 2D planes into UDEC. Although this analysis was successful in explaining the influence of structure in a stress model, it was limited by the distinct element model used. It was not possible in UDEC to represent fractures terminating in intact rock. Consequently, the represented fracture systems were simplified so that fractures terminating in intact rock not considered during the analysis. Other examples of linking fracture systems to distinct models are reported by Staub et al. (2002) also linking fracture system to UDEC.

Pine et al. (2006) introduced 2D traces from a 3D fracture system representing a mine pillar into a 2D hybrid model (ELFEN). A similar approach was used by Elmo & Stead (2009) to characterize the mechanical properties of rock pillars constructed in fractured rock masses. Although incorporation of a fracture system model into the hybrid models like ELFEN allows for a more realistic simulation and representation of fractures network, the models are generally limited in 2D. A fracture system model is three dimensional and can be better represented in a 3D model.

Some of the earlier efforts to link fracture systems with 3D stress analysis packages have been reported by Kulatilake et al. (2004) and Olofsson & Fredriksson (2005). In both cases, they linked a 3D fracture system to a 3D distinct element model (3DEC). In order to overcome computational limitations, Kulatilake et al. (2004) limited the maximum number of fractures linked to the 3DEC model to 16. This small number of fractures was chosen from the 8100 generated fracture in a 3D fracture system model for a 30 m cube. Furthermore, in order to be able to discretize the domain into polyhedral, as required by 3DEC, it was necessary to introduce fictitious joints in the generated rock mass. These fictitious fractures were used in tandem with the defined fractures. Although this was a clever way to overcome inherent software limitations, this approach cannot be considered adequate. A different approach to the problem, which consists in extracting fracture traces and representing them in 2D sections, was followed by Olofsson & Fredriksson (2005). These trace data were then used as input data in a 3DEC model which was loaded in plain strain. It can be argued that neither of these approaches provide for a true 3D link between fracture system models and stress analysis packages.

More recently, the synthetic rock mass approach has been developed based on the link between fracture system model and bonded particle model. The bonded particle model is based on Particle Flow Code, proposed by Potyondy «fe Cundall (2004). Park et al. (2004) reported an early example of extracting 2D trace sections from a 3D fracture model for the Àspô ZEDEX tunnel and integrating them in a 2D PFC model. They consequently measured the mechanical properties of the 2D synthetic rock mass model. Hadjigeorgiou et al. (2007) used a similar approach to investigate the stability of vertical raises. Traces of fractures formed rock wedges along the walls of a vertical raise were introduced into a

95


2D PFC model to simulate influence of stresses on the stability of each individual wedge. The 3D synthetic rock mass model was first introduced by Pierce et al. (2007). They generated spherical synthetic rock mass samples to simulate the failure in a block cave at Northparkes Mines under site-specific stress paths.

An important advantage of a synthetic rock mass model over ELFEN, UDEC and 3DEC codes is that the behavior of a rock mass is not defined "a priori" through constitutive equations, Cundall (2000). In addition, isolated fractures or dead-end segments of fractures in the UDEC and 3DEC models are generally discarded because the initiation and propagation at tips of fractures cannot be effectively simulated. This limitation has been overcome using the synthetic rock mass model, where it is possible to fully integrate a fracture system model into a PFC model, in both 2D and 3D. Using the synthetic rock mass approach it is possible to simulate not only failure along pre-existing fractures but also fracturing through the solid rock matrix.

The objective of current work is not only to simulate the interaction of stress and structure on the stability of ore pass systems, but also to address the effect of material flow on the ore pass integrity. Previous works on material flow in ore pass systems by Iverson et al. (2003) and Hadjigeorgiou & Lessard (2007) successfully employed the PFC. Consequently, it would be a further privilege to use the Particle Flow Code not only as a stress modeling tool, but also as a tool for modeling of material flow.

5.3 The Particle Flow Code The particle flow code, developed by Itasca (2008a) uses the distinct element method (DEM). It models the behavior of distinct rigid circular (2D) and spherical particles (3D). The particles move independently and they interact with each other through forces that develop uniquely at contact points. The physical properties of particles at contact points are described by a linear stiffness model and a slip model. The stiffness model represents elastic relationship between force and displacement. The slip model introduces friction into contact behavior. Any arbitrary particle shapes, such as clump, can also be created by attaching particles together. The particles may represent grains in granular material like sand, or they may be bonded together to represent a solid material like rock, which is considered as bonded particle model.

The bonded particle model (BPM) was presented by Potyondy & Cundall (2004). The BPM simulates an intact rock as a packing of non uniform circular or spherical rigid particles that are connected together at their contact points with parallel bonds. The micro mechanical properties of the rigid particles are shear and normal stiffness and coefficient of friction. The micro-properties of the parallel bonds are normal and shear stiffness, tensile and shear strength and parallel bond radius multiplier. A bonded particle model can represent a homogeneous rock, or it can be divided into a number of discrete regions or blocks by fractures. The properties of particles and bonds along the fracture planes are usually different than those that exist in the solid part of the model. Loading of a BPM

96


can result in breaking of bonds between particles which generally simulates rock fracture and failure. Detailed description about the bonded particle model, its formulation and generation was presented in Appendix-C.

The calculation method in the PFC is a time-stepping, explicit scheme. It involves the repeated application of the law of motion to each particle and a force-displacement law to each contact. At each calculation cycle, Newton's second law (force = mass x acceleration) is employed to determine the change in velocity and position of each particle came up from the contact forces, applied forces and body forces acting upon it. Based on these new particle positions, the linear force-displacement law at contacts is used to determine the resultant contact forces from the relative motion at each contact, Figure 5-2.

p a r t i e ^ 3 " positionsandsetof^

Law of Motion (applied to each particle)

• resultant force + moment

Force-Displacement Law (applied to each contact)

• relative motion • constitutive law

Figure 5-2. Calculation cycle in PFC, after Itasca (2008a).

In particle flow code, rigid walls are available to load the assembly of particles. The walls have their own set of contact properties including their normal and shear stiffness and coefficient of friction. By prescribing a wall velocity, the particle assembly can be loaded. The resultant forces acting on the wall can be measured.

During a simulation, particles and walls in a model can be deleted or have their properties modified at any time. In addition, other particles or walls can be added to an existing model, provided that the new particles and walls are not generated on the existing model.

97


5.3.1 Simulation of intact rock properties

It is customary to assign the mechanical properties of intact rock in a bonded particle model, and to compare the results of mechanical properties from laboratory tests to those obtained from computational tests. This inverse calibration method is used to establish the appropriate micro-mechanical parameters of a BPM that will result in representative intact rock properties.

The computational tests which are used for the simulation of intact rock properties are unconfined compressive strength test, the Brazilian indirect tensile test, the direct tensile test and the triaxial compressive strength test.

The process of calibration of micro-mechanical properties for intact rock simulation has been discussed by Autio et al. (2002) and Potyondy & Cundall (2001). A major challenge in the generation of a bonded particle model is to select the necessary micro-mechanical parameters of particles and bonds that result in representative intact rock properties. Different combinations of micro-mechanical properties can produce similar macroscopic mechanical properties. In order to reproduce the mechanical properties of a solid material (intact rock), an iterative process is employed, as suggested by Potyondy & Cundall (2004). This process involves selecting a particle size, and the attempt to match an appropriate macro property, such as the elastic modulus to this particle size. The first step in the iteration process is to match the desired elastic modulus at the macro level. The elastic modulus is controlled by the particle contact modulus Ec, the particle normal/shear stiffness (kn/ks), the parallel bond modulus Wc and the bond normal/shear stiffness (kn l ks ). The second macro property to establish is the Poisson's ratio of the BPM. This ratio is influenced by the particle normal/shear stiffness (kn/ks) and the bond normal/shear stiffness (k n lk s ) ratios. The final parameter to calibrate is the uniaxial strength of the intact rock. This property is controlled by the average normal and shear strength of the particle bonds.

Potyondy & Cundall (2004) simulated the mechanical properties of Lac du Bonnet granite rock using the 2D and 3D bonded particle model. The results that they obtained indicate that for both the 2D and 3D BPM the Brazilian tensile strength is greater than that of the real rock samples. They have related this discrepancy to the use of circular and spherical particles and have suggested the use of irregular grain shapes to reduce the Brazilian tensile strength. They furthermore investigated the influence of particle size on the mechanical response of the 2D and 3D bonded particle models. The result of their investigation demonstrated that particle size is an essential part of material characterization which can also controls model resolution. In a PFC2D model the elastic constants (Elastic modulus and Poisson's ratio) and the unconfined compressive strength are independent of particle size. However, the Brazilian strength exhibits a clear dependence on particle size. The Brazilian tensile strength decreases as particle size is reduced. In PFC3D model, the Poisson's ratio is independent of particle size, while both elastic modulus and unconfined compressive strength increase as particle size decreases. The Brazilian strength decreases with decrease in particle size.

98


5.3.2 Simulation of mechanical properties of fractures

The strength of an in situ rock mass is influenced by the mechanical properties of its fracture network. In a BPM model, fractures can be defined as planes along which clusters of particles can slide and separate. Fracture planes can be defined individually or as a set of planes. The fracture sets can be combined to create a jointed or blocky system with multiple joint intersections.

In defining fractures in a PFC model, it may be necessary to assign a different contact model and properties, for all particles that are situated on opposite sides of a fracture plane from those assigned to other particles not on the fracture plane. The original contact model for particles along a fracture surface produces a bumpy surface. Sliding is inhibited by the roughness of these planes. Small particles can be employed to decrease the bumpiness of the surface, but this is not feasible when the model requires a large number of interfaces. This problem has been overcome in the new version of the Particle Flow Code (PFC-V4), by assigning smooth-joint contact models to all contacts between particles that lie on opposite sides of a fracture plane, Mas Ivars et al. (2008). This model effectively eliminates the bumpiness of the fracture surface and allows for the specification of macroscopic joint properties.

This smooth-joint model simulates the behavior of an interface regardless of local particle contact orientations along the interface. A pair of particles joined by a smooth-joint contact model may overlap and pass each other, instead of being forced to move around one another, Figure 5-3. The smooth joint model can act as a set of elastic springs which are uniformly distributed over a circular cross-section, centered at the contact point and oriented parallel to the fracture plane. During each time step, the increase in relative displacement of the two particle surfaces relative to each other is analyzed in terms components that are normal and tangential to the fracture surfaces. These components are multiplied by the smooth-joint normal and shear stiffnesses to produce increment of fracture force. The contact force-displacement law acts in the fracture coordinate system and provides either Coulomb sliding with dilation or bonded behavior. Details on smooth joint contact model are available in Itasca (2008a).

99


joint smooth-joint contact

cross-section

Figure 5-3. a) Notation used to define joint and smooth-joint contact, b) Large shearing motion results in the creation of smooth-joint contacts along the fracture plane, after Mas Ivars et al.

(2008).

Laboratory tests are usually used to determine the mechanical properties of fractures. This includes using the direct shear test, the tilt test or the triaxial test. Wang et al. (2003) simulated a direct shear test with the PFC2D model in order to determine the mechanical properties of fractures. They measured the shear stress and shear displacement along the fracture plane at different normal stresses. They plotted both the maximum and residual shear stress versus the normal stress in order to determine the cohesion and internal friction angle of the fracture plane. A similar approach was employed by Kulatilake et al. (2001) to simulate the mechanical properties of fractures using PFC3D.

Triaxial test of a fractured intact rock sample is a common technique for determining the mechanical properties of fracture. In this test the axial stress, which initiates sliding along the fracture surface, is recorded through the use of a low confinement stress. Figure 5-4 demonstrates a schematic of fractured triaxial test specimen with elliptical fracture plane, and displacement orientation along the fracture surface, Rosso (1976).

100


I I SAMPLE AXIS

ELLIPTICAL JONT SURFACE

Figure 5-4. Fractured triaxial test specimen showing fracture plane, strain, and displacement orientations, after Rosso (1976).

By means of the equations provided by Rosso (1976), both the axial and confinement stresses can be broken down into the shear and normal stresses acting along the fracture plane.

Shear Stress: T = (a-, — <r3) Sind Cosd (5-1)

Normal Stress: an = a3 + (o^ — oz)Sin26 (5-2)

Where 0 is the angle between the fracture surface and the sample axis, o\ is axial stress and 03 is confining pressure.

Simulation of a triaxial test on specimens having different fracture inclinations can be used to calibrate the mechanical properties of fractures. Using the bonded particle model a specimen with height to diameter ratio of 2 can be generated. An inclined fracture with inclination angle ranging from 30° to 60° is then created in the center of the specimen using smooth-joint model. Finally, the fractured specimen is subjected to a triaxial test with low confining stress (less than or equal to 1 MPa). Figure 5-5 shows a triaxial test of a fractured rock specimen having a fracture plane inclined at 45°.

101


Fracture

Figure 5-5. Triaxial test for a BPM sample having a joint inclined at 45°.

The axial stress that results in initiation of slip along the fracture surface is recorded. The test is repeated with different fracture orientations. For each orientation, the recorded axial stress and the confining stress are transformed to normal and shear stresses along the fracture planes using the equations 5-1 and 5-2. The results can be plotted as a shear stress-normal stress graph in order to calibrate the mechanical properties of fractures (friction angle and cohesion).

5.4 Generation of a synthetic rock mass model

A synthetic rock mass use Particle Flow Code to represents a jointed rock mass as an assembly of fractures inserted into a rock matrix. This approach necessitates a link between a fracture system model and a bonded particle model. The fractures are represented in the bonded particle model where a smooth joint model is applied to particle contacts along the fracture planes. Loading of a synthetic rock mass can result in new fracture initiation and propagation and sliding along the pre-existent fractures. Cundall et al. (2008) suggested that the pre-peak properties (elastic modulus, crack initiation, Poisson ratio, and peak strength) and post-peak properties (brittleness, dilation angle, residual strength, fragmentation) of a rock mass at different scales can be quantitatively estimated by a synthetic rock mass model. This can have a significant application in characterization of rock mass behaviors in rock mechanics problems.

In order to construct a synthetic rock mass model, the geometrical characteristics of the fracture system are identified. For a two dimensional synthetic rock mass model, first fracture traces on an objective section are determined. This is followed by measuring the

102


coordinate of centre point, dip angle (measured CCW from positive X-axis) and trace length of each fracture trace on the section in question. These parameters are then used to transfer the fracture traces on a 2D bonded particle model. Figure 5-6 shows an example of a two dimensional synthetic rock mass construction based on fracture traces on a section that has intersected a 3D fractures system model. Figure 5-7 demonstrates the application of a 2D synthetic rock mass model in rock slope stability.

Figure 5-6. a) Fracture traces on a 2D section cut through a 3D fracture system model, after Grenon & Hadjigerogiou (2003a); b) the resulting 2D synthetic rock mass model with fracture traces.

Sliding joint (discontinuity)

Bonded parades (" inuct" rock)

Figure 5-7. A 2D synthetic rock mass model for high slopes, after Cundall (2007).

For a three dimensional synthetic rock mass model, the geometrical characteristics of all fractures generated in a 3D fracture system model is transferred to a 3D bonded particle model. This includes fracture centre point, dip angle, dip direction and fracture radius.

103


The direction of the positive Y-axis is considered as North. The disc shape fractures are generated in PFC3D models. Figure 5-8 shows an example of a three dimensional synthetic rock mass model in which three fracture sets, identified by different colors, inserted in a bonded particle model.

Figure 5-8. An example of a 3D synthetic rock mass model.

Once the geometrical characteristics of a fracture system model are embedded into a bonded particle model, mechanical properties of particles and bonds that exist on the fracture planes are modified so that the mechanical properties of fractures are assigned to. The micro-mechanical properties required for characterization of the interface of a smooth-joint fracture are normal and shear stiffness of particles per unit area, friction coefficient of particles, dilation angle, bond tensile and shear strength, Itasca (2008a). In order to assign the smooth-joint model to the contacts on a fracture plane a series of numerical routine was developed by Itasca (2008a). For this mean a set of smooth-joint contacts is created through the fracture surface one at a time, starting from the first fracture. All the bonds exist at the contacts along each fracture can be removed and new bonds can be created for sooth-joint contacts that can either break in tension or shear or can remain intact.

The smooth-joint contact model can be also added to particles when a synthetic rock mass is subjected to loading. Once a large amount of shearing occurs along a pre-existent fracture, the fracture may extend into the rock bridges. It may then be necessary, in the synthetic rock mass model, to trap contacts that form near the fracture, and to assign a

104


smooth-joint contact model to them. Otherwise, these contacts will behave as interacting asperities because they will be assigned the default contact model, and spurious contact forces will develop between them. Such asperity lockup can be prevented by setting the function smooth-joint-add to be invoked whenever a new contact is created, Itasca (2008a).

5.5 Summary and conclusions

This chapter reviewed different numerical methods used for modeling the behavior of fractured rock masses. The continuum and discontinuum numerical methods were introduced and their applicability, advantages and limitations were briefly explained. It was argued that the relative scale of an engineering work (the size of excavation versus the geometry of the fractured rock mass) is an important parameter in selection of a suitable numerical method. Discontinuum models can better represent moderately fractured rock masses while continuum models are used for sparsely or highly fractured domains.

The Synthetic Rock Mass (SRM) model was presented as a hybrid distinct element model/ fracture system model. To set up a synthetic rock mass model, the PFC is used to construct a solid rock as a bonded particle model. The bonded particle model simulates the intact rock by representing it as an assemblage of bonded circular or spherical particles. The bonded particle model is calibrated to establish the mechanical properties of intact rock measured in the laboratory. These properties include elastic modulus, Poisson's ratio, UCS and tensile strength. A network of fractures simulated by a fracture system model based on the fracture mappings in the field are then inserted in the bonded particle model. The smooth joint model is assigned to the particle contacts along the fractures. This allows associated particles to slide through each other along the fracture plane rather than pass over each other.

Several advantages were cited for the synthetic rock mass over other numerical codes which are used for fractured rock mass simulation. The properties of a synthetic rock mass are not assigned directly to the constitutive material. It is the combined behavior of the solid rock matrix (represented by a bonded particle model) and the integrated fracture fabric (smooth-joints) that determine the synthetic rock mass properties. In addition, development of the synthetic rock mass approach is an important progress in simulation of the inherent discontinuous nature of a rock mass in a numerical model. This modeling approach allows the investigation of the mechanical response of a rock mass including a network of intersected and isolated fractures. The rock mass can respond to loading by fracturing through the intact rock bridges between the pre-existent fractures or by sliding and opening of the pre-existent fractures.

Standard mechanical testing of the SRM samples (UCS test, direct tensile test and triaxial test), have been developed by Itasca (2008a) to quantify the mechanical behavior of a fractured rock mass. In the next chapter, synthetic rock mass models of different sizes are

105


generated for the massive sulphide rock mass at Brunswick Mine. The generated models are then subjected to uniaxial compressive test. The results are used for the characterization of the massive sulphide rock mass.

106

Chapter 6: Structural and Mechanical Properties of a Synthetic Rock Mass; a Case..

6 Structural and Mechanical Properties of a Synthetic Rock Mass; a Case Study from the Brunswick Mine

6.1 Introduction

The preceding chapter presented the synthetic rock mass approach, a link between a fracture system and a bonded particle model, to simulate the behavior of a fractured rock mass. This approach was used as part of this research to assess the interaction of stress and structure on stability of ore pass systems. Before using this approach in ore pass stability analyses, it was important to investigate how a synthetic rock mass represents the mechanical and structural behavior of a fractured rock mass. To evaluate the suitability of a synthetic rock mass for modeling of rock mass behaviors it was decided to devise a series of computational tests on the synthetic rock mass model.

This chapter describes the methodology that was adopted to characterize a synthetic rock mass model. The massive sulphide rock mass at Brunswick Mine was used to construct several synthetic rock mass models of varying size. For this purpose the 40 m x 40 m x 40 m fracture system, generated and validated based on the field data, collected for the massive sulphide rock mass and described in chapter 4, was employed. The fracture system model was sampled to procure 40 cubic specimens of height to width ratio of 2 and of varying widths (0.05 to 10 m). These smaller specimens were subsequently introduced into 3D bonded particle models to create synthetic rock mass (SRM) samples.

Both the structural and mechanical properties of the synthetic rock mass samples were measured. This allowed quantifying the scale effect on the rock mass behavior. The structural properties of the synthetic rock mass samples were included number of fractures in each sampled volume (P30) and the volumetric fracture intensity (P32) of the samples. The mechanical properties were uniaxial compressive strength (UCS) and elastic modulus (E) of the synthetic rock mass samples. Based on the results of synthetic rock mass properties obtained for the different sample sizes, a Representative Elemental Volume (REV) size was determined for the rock mass, accounting for both mechanical and structural properties. This chapter builds on previous work by Esmaieli et al. (2009a and 2009b).

107

Chapter 6: Structural and Mechanical Properties of a Synthetic Rock Mass; a Case....

6.2 Structural and mechanical behavior of rock masses

A jointed rock mass is defined by fracture defects of different size. Large scale geological structures, such as faults and dykes, can extend to tens or hundreds of meters while medium scale fractures, such as joints, bedding planes and foliations, can be from a few centimeters to tens of meters. Finally, micro scale fractures, such as micro cracks and grain boundaries, are generally distributed randomly in rock matrices. From an engineering perspective it is necessary to characterize the rock mass and investigate its impact on excavated structures. Consequently the importance of the size of geological fractures can vary depending on excavation size.

Determining the mechanical properties such as strength and elastic modulus of intact rock is relatively straightforward and well documented in the International Society of Rock Mechanics (ISRM) suggested methods, Brown (1981). Extrapolating these laboratory tests into field estimates for jointed rock masses is still a major challenge. This is due to the inherent complexity of a rock mass and in our limitations in quantifying large scale rock mass properties. It is commonly assumed that the mechanical properties of jointed rock are lower than that of intact rock. The mechanical behavior of a rock mass not only depends on the strength and characteristics of the fracture system, but also depends on the strength and deformability of the intact rock bridges between fractures. Fracture properties such as fracture roughness, spacing, alteration, orientation, etc. can all influence the strength and deformability of a rock mass.

Krauland et al. (1989) reported four groups of methods for the determination of mechanical properties of rock mass: mathematical models, rock mass classification, large scale testing and back-analysis of failures.

Mathematical models are based on simulation of both intact rock and fractures as discrete elements of the rock mass. They often require determination of a large number of parameters and are often based on simplifying assumptions. Classification methods use selected rock mass properties to determine a representative index of rock mass quality. This is often linked to rock mass failure criteria to predict the behavior of a rock mass. Rock mass classification techniques are popular even though the assumption that a unique index can capture the behavior of a structurally complex rock mass is debatable. Large in situ tests can measure the relevant mechanical properties of a rock mass. These tests are difficult to undertake and can be very expensive. Large scale laboratory tests are limited by sampling considerations and the inherent difficulties of manipulating large rock mass samples. The use of back analysis of reported failures is an interesting approach to determine the mechanical properties of a rock mass. This requires that the failure mechanism is well-established, which is not the case in complex geological environments.

Figure 6-1 presents a conceptual representation of the transition from intact rock to a heavily jointed rock mass, Hoek (2001). The Hoek and Brown failure criterion is only applicable for isotropic rock specimens or isotropic heavily jointed rock mass. It is of limited applicability for other structural conditions.

108


The criterion is applied to rock masses with sufficient number of closely spaced fractures and, with similar surface characteristics. Isotropic behavior involving failure on fractures can then be assumed. When the block size is of the same order, as that of the structure being analyzed (relative scale is low) or when one fracture set is significantly weaker than the others (presence of schistosity in the rock mass), the rock mass is considered highly anisotropic and the Hock-Brown failure criterion is inapplicable.

Heavily jointed rock mass -use Hoek-Brown criterion for rock mass

Figure 6-1. Transition from intact to heavily jointed rock mass with increasing sample size, after Hoek (2001).

Rock mass behaviors are scale dependent, Krauland et al. (1989), Schultz (1995), Heuze (1980). If the volume of rock mass in Figure 6-1 is increased or if the dimensions of the constructions (tunnel or slope) are changed, it can be noticed that the scale of the construction versus the rock mass and its block sizes has a great influence on the rock mass behavior. This relative scale (scale of construction versus the scale of fracturing) must be considered to choose appropriate strength and deformation properties for rock mass, Schultz (1996), Hudson & Harrison (1997).

The synthetic rock mass approach which was introduced in the previous chapter was developed based on the integration of a fracture system model to a bonded particle model. The resulting synthetic rock mass model aimed to better represent the structural and mechanical behaviors of a fractured rock mass. In order to investigate the mechanical

109


and structural behaviors of the rock mass a synthetic rock mass model was constructed for the massive sulphide rock mass at Brunswick Mine. This was followed by steps and procedures used for characterization of the structural and mechanical behaviors of the synthetic rock mass model.

6.3 Fracture system modeling (FSM)

The use of 3D fracture systems is a powerful tool for an improved representation of jointed rock masses. The fracture system generation methodology was described in Chapter 4 and a fracture system of 40 m x 40 m x 40 m was generated and validated based on the structural data collected at Brunswick Mine from mapping carried out on the massive sulphide exposures, Figure 6-2. The fracture system is presented in Figure 6-3 with the Y axis representing North. Since fracture system models are based on stochastic generation, the fracture system used in this analysis is only one of many possible systems.

Figure 6-2. Photo of massive sulphide rock mass looking North- West.

110

Chapter 6: Structural and Mechanical Properties of a Synthetic Rock Mass; a Case.

Z(m) o

|Fia:rureSeî*l I ■ Fis -..-- : - : » :

X(m) 20 20 Y(m)

Figure 6-3. Visualization of the generated fracture system for massive sulphide rock mass at Brunswick Mine.

6.3.1 Sampling of the fracture system

The validated fracture system model was then subjected to random spatial sampling. In all 40 cubic samples of constant height-width ratio of 2, and of varying base width (0.05 m, 0.1 m, 0.2 m, 0.5 m, 1.5 m, 3.5 m, 7.0 m and 10.0 m) were collected, Figure 6-4. Table 6-1 summarized the dimensions of the collected samples. For each sample size, five samples were extracted from within the initial FSM "master" volume. It was observed that not all generated fracture sets were present in the smaller samples. Furthermore, depending on the size and location of the samples within the master volume, a fracture can be completely engulfed or it can be delineated (truncated) by the sample boundaries. By increasing the sample size, the degree of random distribution of fractures in the collected sample is increased.

Table 6-1. Dimensions of samples collected from the Master fracture system model.

Sample Dimension Sample Dimension

Size#l 0.05 m x 0.05 m x 0.1 m Size #5 1.5 mx 1.5 mx 3 m

Size #2 0.1 mx 0.1 m x0.2 m Size #6 3.5 m x 3.5 mx 7 m

Size #3 0.2 m x 0.2 m x 0.4 m Size #7 7 m x 7 m x l 4 m

Size #4 0.5 m x0.5 mx 1 m Size #8 lOmx 10mx20m

111


\ 0 0 5 m x 0.05m xO 1 m 0 1 m xO 1 m x0.2 m 0.2 m xO.2 m x 04 m 0.5 m x 0.5 m x 1 0 m 1.5 mx 1.5m x3.0 m

3 5 m x 3 5 m x 7 0 m 7.0m x7 Omx 14 0 m

W - - * -"Z* v - 'v 11

v v ; n r _ „ " a r s£ j

100 mx 10.0 m x 20 0 m

Figure 6-4. Rock mass samples drawn from the fracture system model, (not to scale).

6.3.2 Estimation of the rock mass structural properties

In a jointed rock mass, the number of intersected fractures increases with sample size. The relationship between sample size and the number of fractures in any sample is illustrated in Figure 6-5. The same graph illustrates the observed variation in the number of fractures for each sample size. The number of intersected fractures is influenced by the fracture set orientation, spacing and size.

Another way to quantify fracture intensity is the cumulative fracture area per rock volume (P32). As illustrated in Figure 6-6, variations in calculated P32 decrease with increased sample size. The P32 for this rock mass converges to a mean value of 2.65 when sample size reaches 7 m. There is less variation in the calculated P32 for samples greater than 7 m of width dimension.

The volumetric fracture intensity P32 does not depend on the fracture orientation and size distribution and as long as it is representative of the fracture system, P32 is independent of

112


the size of the sampled region, Zhang & Einstein (2000). Fracture spacing, however, depends on fracture orientation and the orientation of the mapped rock mass exposure.

4000

Sample Size (m)

Figure 6-5. Relationship between sample size and number of fractures, including variations.

CL

Sample Size (m)

♦ Size * 1 ♦ Size #2 ♦ Size #3 ♦ Size #4 ♦ Size #5 • Size #6 ♦ Size » 7 Size * 8

Figure 6-6. Fracture intensity (P32) of different sample size, P32 <aVe) =2.65.

113


6.4 Generation of a Synthetic Rock Mass model

The preceding sections focused on generating 3D fracture systems models. These were used to estimate the structural properties of rock mass based on number of fractures per rock volume (P30) and cumulative fracture area per rock volume (P32) values. In order to predict the mechanical behavior of these rock masses it is necessary to load these large samples. An interesting approach to tackle this problem is the use of "synthetic rock mass models" whereby a jointed rock mass is represented as an assembly of fractures inserted into a rock matrix. A practical way to accomplish this goal is by linking a fracture system model to a bonded particle model. A bonded particle model was constructed using the 3D Particle Flow Code (PFC3D). An intact rock was simulated as an assembly of rigid distinct spherical particles, bonded together at their contact points. A linear contact and a frictional sliding model were used to describe the physical behavior of the contacts. Finally a disc-shape fracture system was incorporated into this synthetic rock, which was then put under load.

6.4.1 Intact rock simulation

The intact rock samples were generated using procedures proposed by Potyondy & Cundall (2004). A bonded particle model is characterized by the particle density, shape, size distribution, assembly and the micro-properties of particles and bonds used in the model. The macroscopic mechanical properties of an intact rock sample can be developed by several combinations of micro-mechanical properties. For the purposes of this work the geomechanical database of Brunswick Mine was used to define characteristic properties for intact rock samples. An inverse calibration method discussed in section 5-3-1 was then used to establish the necessary micro-mechanical parameters that would replicate the representative intact rock properties.

The micro-mechanical properties of particles and bonds used to generate the bonded particle model of intact rock are summarized in Table 6-2. Laboratory and PFC3D results for Brunswick Mine's massive sulphide rock are summarized in Table 6-3.

114


Table 6-2. Micro-properties of the PFC3D models.

Sample size (m)

*Mnin

(cm) Ec

(GPa) Ob-Tb

(MPa) Std.

(MPa)

0.05 0.16 117 189 45

0.1 0.31 112 183 45

0.2 0.6 119 181 45

0.5 1.4 114 167 45

1.5 2.5 123 162 45

3.5 3.6 122 151 45

7.0 7.0 121 140 45

10.0 10 120 130 45

Rmir,: Minimum particle radius; Ec: Particle contact and parallel bond modulus;

Ob, ib: Bond normal and shear strength. Std.: Standard deviation of bond normal and shear strength

Table 6-3. Mechanical properties of intact rock and bonded particle models.

Mechanical Properties

UCS (MPa)

E (GPa)

Poisson Ratio

Specific Gravity

Massive Sulphide 205 104 0.29 4.3

PFC3D Sample size

(m)

0.05 203 104 0.29 4.3

PFC3D Sample size

(m)

0.1 204 102 0.29 4.3

PFC3D Sample size

(m)

0.2 205 104 0.27 4.3

PFC3D Sample size

(m)

0.5 204 103 0.28 4.3

PFC3D Sample size

(m)

1.5 205 104 0.28 4.3 PFC3D Sample size

(m) 3.5 204 104 0.27 4.3

PFC3D Sample size

(m) 7.0 204 105 0.29 4.3

PFC3D Sample size

(m)

10.0 205 103 0.28 4.3

UCS: Uniaxial Compressive Strength; E: Modulus of elasticity.

115


The construction and calibration of large intact rock samples in PFC3D imposes high computational requirements. In order to overcome this obstacle, the calibration process was simplified for the larger samples, using the pre-compacted bricks approach proposed by Mas Ivars et al. (2008). Bricks were assembled together to construct rapidly the large intact samples. In this study 0.5 m width bricks were used for the 1.5 m sample, 0.7 m for the 3.5 m, 1.4 m for the 7.0 m and 2 m for the 10 m sample. In all samples the friction coefficient of the particles was fixed to 2.5 and the ratio of particle and bond normal to shear stiffness was equal to 3.6.

6.4.2 Fracture properties

The introduction of fractures in a PFC model has been somewhat problematic in the past because of the inherent roughness (bumpiness) of the interface surfaces. In earlier versions of the PFC, fractures were presented by a band of low-strength material. Consequently, particles that fell along opposite sides of a fracture plane were forced to move around each other in order to slide. In this work fractures were introduced using the smooth joint model (SJM) which was applied to particle contacts. The SJM is a major improvement over prior versions of the PFC software as it simulates the behavior of an interface, regardless of local particle contact orientations along the interface. This ensures sliding and unraveling of rock blocks along the fracture surface, Mas Ivars et al. (2008).

For the purposes of this study all fractures were assumed to be cohesion less and having an angle of friction of 30°. The calibration process described by Hadjigeorgiou et al. (2009) was used to assign the necessary micro-mechanical properties to the particles along the fracture planes, in order to achieve the desired macro-properties. This was done by conducting a series of triaxial tests in the PFC3D model. Four specimens, each intersected by a fracture at a different inclination of 45°, 50°, 55° and 60° were tested under a confinement stress of IMPa. The axial stress, initiating sliding along the fracture surface, was recorded during each test. In designing these tests, the bond strength along the fracture surface was set to zero, and a range of coefficients of friction of adjacent particles (0.1, 0.15, 0.2, 0.25 and 0.3) along a fracture plane were used to arrive at the desired strength. Based on these iterations a coefficients of friction of adjacent particles of 0.2 resulted in zero cohesion and an angle of friction of 30° for the fractures. Another part of the iteration process involved selecting appropriate normal and shear stiffness for the particles along the fracture surface. An initial stiffness value was selected based on the elastic modulus and average particle size. Following a series of iteration a value of 2.5x1012 N/m3 was selected for the normal stiffness and 0.5x1012 N/m3 for the shear stiffness of particles along the fracture surfaces.

This calibration exercise resulted in fractures in the synthetic rock mass samples with the desired macro-mechanical properties of zero cohesion and an angle of friction of 30°. This configuration facilitated the generation of the desired macro-mechanical properties of the fractures. The same mechanical properties were assigned to all fractures in the

116


synthetic rock mass although it is acknowledged that the peak shear strength of fractures can decrease as fracture size increases, Barton & Bandis (1980).

6.4.3 Sample generation

The micro-mechanical properties in Table 6-2 were used to generate rock samples in PFC3D-V4, Itasca (2008a). These are illustrated in Figure 6-7 and the number of particles used to generate the different sample sizes listed in Table 6-4. Once the intact rock samples were generated, all fractures generated using Fracture-SG, were introduced in the corresponding bonded particle models.

The generated fracture system using Fracture-SG and illustrated in Figure 6-3 is based on the Veneziano model which results in polygonal shape fractures. In PFC3D, however, fractures are represented as disk shape features, identified by their center point and radius. Consequently it was necessary to approximate the generated polygonal fractures as disks. This was done using the fracture center point and its equivalent radius. FishLab -V.l developed by Itasca (2008b) was used to visualize the constructed fracture systems in the synthetic rock samples, Figure 6-8. The influence of sample size on the number of encountered fractures is evident.

Table 6-4. Number of particles in different sample size.

Sample size (m)

Nb.of Particles

Sample size (m)

Nb.Of particles

0.05 5512 1.5 39014

0.1 6063 3.5 92300

0.2 6689 7.0 115455

0.5 8228 10.0 142376

117


y \

0.05 m X 0.05 m X 0 1 m 0 1 m X 0 1 m X 0 2 m 0.2 m X 0.2 m X 0.4 m 0 5 m X 0 5mX 1 0m 1.5 m X 1.5 m X 3.0 m

f

vu

3 5 m X 3 5 m X 7 . 0 m 7.0 m X 7.0 m X 14.0 m 10.0m X 100 m X 20 0 m

Figure 6-7. Bonded particle models of intact rock samples.

118


3 5 m X 3 5 m X 7 0 m 7 0 m X 7 0 m X 1 4 0m 1 0 0 m X 1 0 O m X 7 O 0 m

Figure 6-8. Synthetic rock mass samples generated by linking the PFC3D and fracture system models.

6.5 Mechanical properties of Synthetic Rock Mass samples

A series of uniaxial compressive tests were performed on all 40 synthetic rock mass samples. The PFC3D code was run on a computer using a Pentium IV processor with a speed of 3.3 GHz.

The wall-based loading procedure in PFC3D was used to load the 0.05 m, 0.1 m, 0.2 m and 0.5 m synthetic rock mass samples. In this process, the walls that confine the specimen in the bonded particle model generation were used to load the specimen with the top and bottom walls acting as loading platens, Figure 6-9a. In order to speed up testing of large samples a new testing methodology developed by Itasca was used. This approach facilitates uniaxial, triaxial and direct tension testing of large samples along different coordinate directions. A grip-based procedure and a rapid loading methodology were used for the larger samples. This involved identifying as plates, grip spheres on the top and bottom of each synthetic rock mass sample. Once the grip particles were identified, loading was performed using internal-based or platen-based methods. Internal-based loading assigns linearly varying axial velocities to all particles of assembly. This can result in a specified gradual development of induced axial strain in the assembly of particles over a certain number of calculation steps. At each stage of induced strain, the grip particles are stopped, the axial velocity of all other particles is set to zero and they

119


are free to displace until static equilibrium is re-established. The platen-based approach is performed by moving the grip platens toward one another, resulting in the desired velocity of the grip spheres. This approach was used to load the larger synthetic rock mass samples shown in Figure 6-9b. Some of the larger samples were randomly selected for comparing the results of both loading methods. The results obtained by the platen-based and internal-based methods are almost the same. Two samples defined by a base width of 7 m and 10 m were tested by both internal based approach and platen-based methods. The results are less than 2 MPa apart.

What is important in both platen and grip based approaches, and is that the loading rate, applied to the particle assembly, must be slow enough to allow time for the particle assembly to adjust to the force redistribution that accompanies slip and bond breakage.

a) b)

Figure 6-9. Uniaxial compressive test of synthetic samples: a) using wall for loading, b) using spherical grip particles for loading.

6.5.1 Uniaxial compressive strength of synthetic rock mass samples

The relationship between sample size and normalized uniaxial compressive strength (UCS of tested samples/UCS of intact rock), calculated from the PFC3D loading tests is illustrated in Figure 6-10 including the 90% confidence limits. The average strength, for

120


the smallest sample of 0.05 m was approximately 160 MPa which is somewhat lower than the uniaxial strength of the intact rock. At the other extreme, the strength of a 10 m rock mass sample was approximately 45 MPa which is close to 20% of the UCS of the intact rock. The stress-strain curves for all synthetic rock mass samples were presented in Appendix-D.

An interesting observation in Figure 6-10 is that the mean uniaxial compressive strength and its variation seems to diminish with the increase in sample size. The higher variation for the smaller samples is attributed to the very small number of fractures present. This can attributed to failure of the rock samples along several possible preferential planes of weaknesses. If an unfavorable orientation is present this will result in much lower strength. This problem illustrated in Figure 6-11 where the uniaxial compressive strength of two samples having the same dimension of (0.05 m x 0.05 m x 0.1 m) was compared. In the first sample the unfavorable orientation of fracture with respect to the loading axis has resulted in a low strength, while in the second sample the semi-vertical orientation of fracture has resulted in higher strength of the synthetic rock sample. Moreover, the failure mechanism in the first sample is the propagation of the pre-existent fracture, while for the second sample the initiation of cracks in the solid rock matrix and the coalescence of these cracks with the pre-existent fracture plane were resulted in the rock sample failure.

1,2

CO o 3

1 -

0,8 -

o a _c CO O D

J 0,6 H Q. E CO

0,4

0,2

« o

o + 90 % O

Level

/ ■

jnfidf ▼ « o

o + 90 % O

Level

/ ■

jnfidf ;nce

i ...A—

+ 90 % O Level

/ ■

* o

o A

■ !■

►

■

■ 1

♦

It 1

o + ■

!■

►■

■

1

♦

It 1

♦ X

1

o + ■

!■

►■

■

1

♦

It 1

240

200 ro 0_

CL

160 | CO

o

120 I

c 80 CO

CO o 3

40

0,01 0,1 1 10

Sample Size (m) 100

♦ Size#l «Size #2 o Size #3 + Size #4 «Size #5 Size #6 «Size #7 «Size #8 A Average

Figure 6-10. Influence of specimen size on the strength of rock mass.

121


a) UCS = 39 MPa

b) UCS =154 MPa : " :>!

Figure 6-11. The uniaxial compressive strength and failure pattern of two same size samples (0.05 m x 0.05 m x 0.1 m); Yellow: tension cracks, Black: shear cracks.

6.5.2 Elastic modulus of the synthetic rock mass samples

The elastic modulus of synthetic rock mass samples was measured during the uniaxial loading tests. Figure 6-12 presents the normalized elastic modulus (elastic modulus of tested samples/elastic modulus of intact rock) versus sample size. The 90% confidence limit illustrates that, the elastic modulus for the largest sample size is about 35% to 40% of the intact rock elastic modulus. The average elastic modulus for the smallest sample of 0.05 m in width was close to 90 GPa, while for the largest sample of 10 m width it was about 38 GPa. In general, it seems that the elastic modulus of the rock mass is about 38% of the intact rock elastic modulus. In general, the mean elastic modulus and variance decrease with sample size increase.

122


0,01 0,1 10

Sample Size (m)

125

104

83

62

42

21

100

ro C3

_0j Q.

u O _̂ u

'■I, 01

C c >

1/1

3 T3 O

♦ Size # 1 x Size # 2 o Size # 3 + Size # 4 ■ Size # 5 i Size 6 * Size # 7 • Size # 8 A Average

Figure 6-12. Influence of specimen size on the elastic modulus of rock mass.

6.6 Estimation of the REV size

The undertaken procedure allows a quantitative estimation of what is a Representative Elemental Volume (REV) for the investigated area in massive sulphides at the Brunswick site. Based on definition proposed by Hudson & Harrison (1997), the REV is the volume for any given body at which the size of the sample tested contains a sufficient number of inhomogeneities for the "average value" to be reasonably consistent with repeated testing, Figure 6-13.

123


a, c

Inhomogeneous medium

Homogeneous medium

REV Size

Figure 6-13. The REV concept; after Hudson & Harrison (1997).

In order to investigate the REV size of the rock mass two approaches were considered. In the first approach the T and F statistical tests were performed to quantify REV size. The tests were performed for both structural and mechanical properties of the rock mass. This included the results of the P30, P32, UCS and Elastic modulus. The T-test (Student test) assesses whether the means of two groups are statistically different from each other. In this analysis the mean value of P30, P32, UCS and Elastic modulus for the sample of 10 m width (largest sample size) were compared with the means of the smaller samples. The T test investigates the null hypothesis that data in two groups (i.e. sample sizes) are independent random samples from normal distributions with equal means and unknown variances. This is tested against the alternative hypothesis that the means are not equal.

The results of the T-test for number of fractures per unit volume of rock mass (P30), fracture volumetric intensity (P32), uniaxial compressive strength and elastic modulus of the synthetic rock mass samples are summarized in Table 6-5. All calculations were based on a maximum significant level of 5%. The maximum significant level at which the null hypothesis cannot be rejected is the P-value. The higher the P-value, the more probable the mean values of different sample sizes are equal.

124


Table 6-5. The results of T-test for P30, P32, UCS and elastic modulus of synthetic rock samples, (5 samples per each sample size).

Sample Size (m)

T-Test

Sample Size (m)

P30 P32 UCS E Sample Size (m)

P-Value Result P-

Value Result P-Value Result P-

Value Result

0.05 0.178 Accepted 0.33 Accepted 0.02 Rejected 0.0064 Rejected

0.1 0.071 Accepted 0.44 Accepted 0.1 Accepted 0.018 Rejected

0.2 0.004 Rejected 0.22 Accepted 0.07 Accepted 0.009 Rejected

0.5 0.007 Rejected 0.79 Accepted 0.16 Accepted 0.088 Accepted

1.5 9e-5 Rejected 0.51 Accepted 0.10 Accepted 0.10 Accepted

3.5 2e-4 Rejected 0.11 Accepted 0.02 Rejected 0.035 Rejected

7 0.051 Accepted 0.77 Accepted 0.49 Accepted 0.42 Accepted

The F-test was used to determine whether the variances of the calculated mechanical and structural properties for different sample sizes, were statistically different from the 10 m sample (largest sample size). An F test is performed to validate the null hypothesis that two independent samples were from normal distributions with the same variance, against the alternative that they were from a normal distributions with different variances.

Considering a maximum significant level of 5%, the results of the F-test for fracture intensity, uniaxial compressive strength and elastic modulus of the synthetic rock mass samples are summarized in Table 6-6.

125


Table 6-6. The results of F-test for P30, P32, UCS and elastic modulus of synthetic rock samples, (5 samples per each sample size).

Sample Size (m)

F-Test

Sample Size (m)

P30 P32 UCS E Sample Size (m)

P-Value

Result P-Value Result P-

Value Result P-Value Result

0.05 4e-18 Rejected le-7 Rejected 4e-5 Rejected le-3 Rejected

0.1 3e-14 Rejected 5e-6 Rejected 2e-5 Rejected le-3 Rejected

0.2 3e-10 Rejected le-4 Rejected 3e-5 Rejected 2e-3 Rejected

0.5 le-8 Rejected 3e-4 Rejected 5e-5 Rejected 9e-4 Rejected

1.5 0.006 Rejected 0.01 Rejected 3e-4 Rejected 0.001 Rejected

3.5 0.53 Accepted 0.07 Accepted 0.009 Rejected 0.024 Rejected

7 0.42 Accepted 0.4 Accepted 0.11 Accepted 0.2 Accepted

The following conclusions can be drawn for the REV size of the rock mass, for structural and mechanical rock mass properties, based on the results of the T-test and F-test.

• The F-test results for the number of fractures per unit volume of the rock mass (P30) indicate that the equal variance is achieved when the sample width is 3.5 m. The T-test results for the two smaller size samples (0.05 m and 0.1 m) indicates that the null hypothesis (the mean values are equal) cannot be rejected. This is due to the high variances of the fracture numbers per unit volume of rock mass for these samples. For the mid size samples (0.2 m to 3.5 m), the average number of fractures per unit volume of rock mass are bigger than for the largest sample (10 m). Finally, for the 7.0 m samples the null hypothesis cannot be rejected. Although the sample size 3.5 m x 3.5 m x 7 m has an equal variance with the biggest sample size, the mean P30 property for this sample size is bigger than for the 10 m x 10 m x 20 m sample. The REV size of the rock mass, based on the P30 property, defined by an equal mean and variance values, is 7 m x 7 m x 14 m.

• For the fracture intensity (P32) values, the results of the F-test show that the sample variances are equal when the sample size reaches a width of 3.5 m. The results of T-test demonstrate that the null hypothesis (the mean values being equal) cannot be rejected at the 5% significant level. Then in this example the P32 based REV size, defined by an equal mean and variance values for different sample sizes, is 3.5 m x 3.5 mx 7 m.

126


• For the UCS values, the F-test results show that the sample variances are equal only when the sample size reaches a width of 7 m. The results obtained for samples smaller than 7 m display a much larger variance. The results of T-test statistically demonstrate that for the smallest sample size, the mean UCS value is not equal to the mean UCS value of the largest sample. For the mid range sample sizes (0.1 m to 1.5 m), T-test indicate that the null hypothesis (the mean values are equal) cannot be rejected due to the high variances of uniaxial test results of these samples. For the 3.5 m size samples the mean UCS value is higher than for the 10 m size samples (the null hypothesis can be rejected) and for the 7.0 m samples the null hypothesis cannot be rejected. Finally, the REV size of the rock mass, defined by an equal mean and variance values for different sample sizes, is 7 m x 7 m x 14 m based on UCS values.

• For the values of elastic modulus, the F-test results indicate that the sample variances are equal only when the sample size has the 7 m width. The samples smaller than 7 m width have greater variances. The results from the T-test demonstrate that for the three smaller sample sizes (0.05 m, 0.1 m and 0.2 m), the mean E values are higher than for the 10m sample. Based on the large observed variances the null hypothesis (the mean elastic values are equal) cannot be rejected for the 0.5 m and 1.5 m samples. The mean E value of the 3.5 m size samples is higher than the 10 m size samples. Finally, the null hypothesis cannot be rejected for the 7.0 m samples. In conclusion the elastic modulus based REV size, defined by an equal mean and variance values for different sample sizes is as 7 m x 7 m x 14 m.

The second approach which was employed to estimate the REV size of the rock mass was based on the coefficient of variation. The coefficient of variation (CV) is the ratio of the standard deviation to the mean value. The coefficient of variation of the mechanical properties (UCS and E) and structural properties (number of fractures and P32) were plotted against sample size, Figure 6-14.

127


> o

5? ô

1 O

Sample Size (m)

• Fracture Intensity (P32) ■ Nb. of Fractures ♦ UCS A Elastic Modulus

Figure 6-14. Coefficient of variation vs. sample size.

The structural and mechanical properties of sampled rock mass are listed in Tables 6-7 and 6-8. Referring to Table 6-7, some of the smaller size samples such as 0.05 m and 0.1 m may end up not intersecting any fractures. In other words our sampling resulted in intact rock samples. Consequently when these samples were tested they give the same results as intact rock. This explains the somewhat lower coefficient of variation for the mechanical properties of these sample sizes.

It is interesting to observe that the coefficient of variation values, for both mechanical and structural properties, converge when the sample size has a width dimension between 3.5 and 7 m. Min & Jing (2003) had suggested that the REV size for a given rock mass can be determined according to a chosen 'acceptable variation'. In this study, following the proposed guidelines, the acceptable variations for CV were 10% and 20%. The selected REV sizes, for both mechanical and structural properties based on these criteria are summarized in Table 6-9. If only the structural properties were considered the selected REV would be 3.5 m x 3.5 m x 7 m. If the decision is based on the mechanical properties of the rock mass then a 7 m x 7 m x 14m REV size is a more appropriate choice.

The REV sizes based on the coefficient of variation criteria and the T and F tests are in agreement. In this case study, based on the available conditions the rock mass can be described by a global REV size, considering both structural and mechanical properties, of 7 m x 7 m x 14 m.

128

Vi a U cd vï Vi

o o

OJ

■S c

CO cd

' + O Vi <U

n. o

u

J3 o

crj

r/3 s©

> OJ

u

s

es E

c O L.

«

si _Çj

S

00 c o Os o

in Tf

o CN

o in SO

d ro d

CN d

ro d

Mea

n

Val

ue

m SO

CN Tt

Os Os ro

CN oo CN

ro CN

o ro CN

in SO CN

CN SO CN

'S co

. & G <D a

i o {S (H

o o o

m Os «ri

00 iri

ro in (N

SO SO in

CN O t-; CN

O 00 CN

'S co

. & G <D a

i o {S (H

Tt o o o

SO CN o in

in oo 00

00 CN CN

■«t 00 "t

CN O sq CN

'S co

. & G <D a

i o {S (H

CO SO ON

o o ©

Tt d

ro » — i

00 o CN

so sq CN

o Os CN

in Tt CN

'S co

. & G <D a

i o {S (H

CM CO

o o o in

Os o in

00 ro

OO CN

O sq CN

O Tt CN

O CN

'S co

. & G <D a

i o {S (H

» — 1 CO o

CN

O o d

so ro

CN in

ro ro ro

Os O 00 CN

in in CN

CO in o

in d

Tt d

SO CN CN

O Os

O d OO

CN in

•§13 TT o

so d

(N CN SO

00 d ro

Tt Os .-H

Tt CN

CN oo o ro

1 o li PH o 11 w 1

in co o - -< SO ro

Os Os .-H

CN in CN

so Tt ro ro

1 o li PH o 11 w 1

T * co o - (N r-- 00

CN CN 00

Tt Os

00 Os O ro 1

o li PH o 11 w 1

CO o - ro ro ro

in o CN

Tt o ro

O 00 Os CN

1 o li PH o 11 w 1 CN

CO o - l - ^ in oo CN

ro O CN

O Os O

SO CN O ro

1 o li PH o 11 w 1

co o - O ro

00 Os

CN CO CN

SO 00 Os CN

Sam

ple

S

ize

(m)

o o d

(N d

in d

in in ro r̂ O

Os CN

CJ V i C3

r j cd V i V i a

y o

c CO

C3

O V i CD

' ■ f i

OJ c o H

O. "3 o

'S c3

C O CJ

T3 C C3

"3 3 o S

co sC t

o -4-» CL C3

o

es

E

o

U

E

o ai oo sO

^̂ ^

Std

.

(GP

a <n r o Tt" Os es >n rs o

Std

.

(GP

a

<N <N

>n es Csl

SO Csl

i n es 1

—■ sO r o

B o 9 CS <N O TT o r s O T t U 3 CU

2 > w

O _̂ r o • n r s T f (—. 0 0 U 3 CU

2 > w ON oo 0 0 SO sO m TI r o

■n o o T f rs OS CS i n j j i CO o

T f 0 0 m Os r o T f r o o o K o

TT T T O o o r» 0 0 oo Os SO

g J 3

00 T T O 0 0 <n VO r~ SO T f r o g

J 3

us

of S

yn

se

s (G

Pa)

us

of S

yn

se

s (G

Pa)

r o <N CN O

Tt; O 0 0 es r r̂

us

of S

yn

se

s (G

Pa)

oo <n CN O

Tt; O Os SO TT r o T I

3 % o'S 3 % o'S CN TI

CS Csl o o

. CS m Tf CS

3 % o'S

co TICS

oo o o 0 0 ■n >n T i Tt

o

SS SS

U P H r~ CN O o r~ SO CN r o 0 0

0 0 0 0 CN O oo <N ( N SO r o r o

•—s

Std

.

(MP

a * p i oo <N r̂ 0 0 Os sO

Std

.

(MP

a

r» Csl 0 0

rs) SD

Csl T f

f oo r o

Mea

n

Val

ue

(MP

a)

T f 0 0 TT T * o Tt 0 0 r

Mea

n

Val

ue

(MP

a)

o so

Csl CN

T f CNl SO

Os >n 0 0 f

rT t 5

>n r o O Csl

O r̂ Os T? Tt o r o 0 0

r o O Csl ( N _ •n T?

m >n T f

O

ttffi ttffi T l " oo

r o O t N

O CS Cs|

Csl T t sO

>n O Os

oo i n Os

r o oo « <U U I

.S 3 r o 00

Os r o

Tjo Csl

r o O CN

Csl OS ON

m SO

OS r o

r^ t

P. U

o 05 o 05 U o CS r o

O (N

sO <n 0 0

OS 0 0 r T t Os

a u 00 r o O (N SO

<n 0 0 i n m >n Tl

Un

iax

i S

ynth

U

nia

xi

Synth

U

nia

xi

Synth

rt T f

m Csl i n oo O O 0 0 m

00 T f

m Csl Os Csl r o Os r o T i

■fi *i o O o

Ela

sti

od

ulu

s ta

ctR

c (G

Pa)

S CN O

T f

o r o O

TI CS

T f O

i n O

r o O

Ela

sti

od

ulu

s ta

ctR

c (G

Pa)

Ela

sti

od

ulu

s ta

ctR

c (G

Pa)

S .5 <»

t f l Ï J J O r o Tj >n TT ■n T f T f i n

U a " £ O t O O O O O © O t N CN CM CN r s CS r s CS

o

U , v ex Ë

O —. Csl m m ■n r^ o

§ '> O o O o — r o Ô5.S

00

o

c o

> ■ ■ *

o

u £ o CJ c o

■o

CJ

o

N

>

ai

c o

E eu eu

Ci Cjs'

i sO

"S

cd o s P S JH o u T t CN

X X

6

SS r~ o X X

S S S o r» o ê 1

—* co

> e S o

£ T t CN X

"3 'S S 6

o ct$ r^ ~̂ a x X o eu E Fi s t » o

> S r

S

s x B

T t

X

s i n r o

E o X

è E E C/D i n

E r o

T W

c u O / * .

e^ o "—* o C <J 2 O

CN o

Js* t3 V v I as a > H < + H

s o <

Chapter 6: Structural and Mechanical Properties of a Synthetic Rock Mass; a Case ....

6.7 Conclusions

This chapter presents the mechanical behavior of fractured rock masses represented by synthetic rock mass models. The fracture system model constructed based on the field data collected from the massive sulphide rock mass at Brunswick Mine was subjected to sampling. Samples of various sizes ranging from (0.05 m x 0.05 m x 0.1 m) to (10 m x 10 m x 20 m) were collected. Finally, it has been possible to integrate the fracture system samples into bonded particle models to construct the synthetic rock mass models. The uniaxial compressive loading of the synthetic rock mass samples indicated that with increase of sample size the strength and elastic modulus of rock mass samples decrease. The REV size of the selected rock mass properties, was determined using the T-test and F-test and by plotting the coefficient of variation against sample size. Both approaches suggest that, in this particular site, the structural REV size is 3.5 m x 3.5 m x 7 m and the mechanical REV size i s 7 m x 7 m x 14 m. Consequently, the largest REV constitutes the global REV beyond which both the mechanical and structural properties of the rock mass are reasonably consistent with repeated testing.

The REV size obtained for the massive sulphide rock mass at the Brunswick Mine implies that for large scale excavations, such as mining stopes, the equivalent continuum approach can be used in the selection of appropriate numerical models. In these cases, the REV-derived mechanical behavior can be employed as input data for the massive sulphide rock mass. On the other hand, when the scale of excavation size versus rock mass block size is small, such as ore passes, drifts, etc. the equivalent continuum approach is not applicable. The results reported in this chapter were used for a stability analysis of an ore pass system at the Brunswick Mine.

131

Chapter 7: Stability Analysis of the #19A Ore Pass at Brunswick Mine

7 Stability Analysis of the #19A Ore Pass at Brunswick Mine

7.1 Introduction This chapter investigates the stability of the #19A ore pass at Brunswick Mine. This ore pass is a part of 18-21 ore pass complex located in the 20 & 21 mining zones. This is of interest as it is an area of high stress and recorded seismic activity. The area has been investigated in the past using stress analysis models.

The innovation in this chapter is that the undertaken investigation employs fracture system and particle flow code. This involved the construction of a Synthetic Rock Mass (SRM) model to capture the stress-structure interaction. Furthermore, this chapter addresses the influence of other mechanism on the ore pass stability.

A two-stage numerical approach was used to investigate the stability of the #19A ore pass. A global boundary element stress analysis was provided the necessary boundary conditions to build a local distinct element model. The local numerical model was constructed using the synthetic rock mass approach discussed in the precedent chapters. Finally, the dimensions of the #19A ore pass were introduced in the synthetic rock mass model. The resulting analysis aims to provide a better understanding of the interaction of structure and stress in an operating mine. This has significant implications on understanding some of the complex mechanisms that control the structural integrity of ore pass systems. The numerical model was compared to observed field performance of the ore pass system. Preliminary results of this work were presented by Esmaieli et al. (2008).

During the work towards this dissertation, the stability analysis of a vertical raise constructed in hard rock was investigated, using the synthetic rock mass approach. The details of this generic problem are presented in full in appendix-E and were also presented by Hadjigeorgiou et al. (2009). The information and knowledge gained, was applied for a more sophisticated approach to investigate the stability analysis of the #19A ore pass at Brunswick Mine, presented in this chapter. In particular, the work presented in this chapter considered all the advantages of recent developments in the Particle Flow Code (PFC), especially, the "smooth joint model" for contacts of particles along fractures, in both the 2D and 3D analysis.

132


7.2 Ore pass degradation in high stress states

Ore passes located in deep underground mines or in the abutment of mining stopes are generally subjected to high stress states. This can result in serious stability problems for such infrastructures. In an anisotropic stress field, a predictable stress is concentrated tangent to the ore pass wall. The stress concentration zone is formed perpendicular to the axis of major in-situ stress. When the induced stress exceeds the strength of rock mass, a damage process begins in the stress concentration zone. Martin et al. (1999) suggested that stress induced fracturing around underground openings initiates when the maximum tangential stress exceeds approximately 0.4 times the uniaxial compressive strength of intact rock. The damage then is a function of rock mass characteristics, in-situ stress states and excavation geometry.

A typical characteristic of the stress induced damage zone around ore passes, constructed in brittle rocks, is the formation of a V-shape (notched-shape) failure region. The failure in the notched-shape zone is governed by initiation and growing of new tension cracks usually parallel to the ore pass boundary. These cracks are merged together to form fractures. The process results in creation of rock slabs in the damage zone which is called spalling and slabbing. Martin et al. (1997) reported that the thickness of slabs, create in the damage zone, can vary from a few millimeters to tens of centimeters.

Presence of pre-existent rock structural defects and their distribution and persistence in a rock mass can influence the formation and ultimate shape of a damage zone which is formed around an underground opening excavated in the rock mass. Sellers & Klerck (2000) conducted laboratory tests to investigate stress induced failure around a hole drilled in a cubic rock sample containing several horizontal fractures. They furthermore simulated the similar problem using finite-discrete element method (ELFEN). The results of both laboratory and numerical experiments indicated that the pre-existent fractures can alter the state of stresses around the excavation and subsequently can change the pattern of damage zone which is different from classical notch shape failure.

Based on observation of damage zones around tunnels constructed in jointed rock masses, Goodman et al. (1972) suggested that pre-existent fractures induce additional fracturing which can modify failure pattern in the damage zone. Other observations in Underground Research Laboratory of Canada in Manitoba by Everitt & Lajtai (2004) reveals that damage in stress concentration zone is more extensive where the foliation planes coincides with slabbing directions. Other investigations in deep Canadian underground mines by Castro et al. (1997) suggested that, with an increase in stress magnitudes, pre-existent fractures in the rock mass surrounding excavations are clamped, particularly, in the stress concentration zone. In this situation, displacement and sliding of rock wedges along fractures has minor influence on initiation of rock mass damage around the excavations. According to Castro et al. (1997) damage process starts with intact rock fracturing for example in the form of breaking rock bridges between the existent fractures.

133


Experiences from underground openings suggest that rock slabs formed in a damage zone around an underground excavation are in a meta-stable state, Potyondy & Cundall (2004) and Martin et al. (1997). These rock slabs do not fully detach from their rock mass. Therefore, in the damage zone the rock mass become weaker. According to Martin et al. (1997), if the rock slabs are removed from the excavation wall the extent of damage zone increase. In ore pass systems flow of rock fragments and their impact with damage zone around the ore passes can easily remove the stress-induced rock slabs which can lead to significant expansion of damage zone around the ore passes, Rech et al. (1992).

Most of the geotechnical investigation methods developed for ore pass stability analysis in high stress states are based on the estimation of induced stresses around the ore pass and the choice of a proper failure criterion for the rock mass. The majority of the developed failure criteria to estimate the extent of damage zone are based on the ratio of the maximum tangential stress to the uniaxial compressive strength of intact rock. Analytical and numerical methods have been employed to estimate the induced stresses around the ore passes. Joughin & Stacey (2005) used analytical method (closed form solution) for estimation of induced stress around the ore pass walls in several South African underground mines. Vieira & Durrheim (2005) used a boundary element numerical method to estimate the maximum tangential stress around the ore passes in deep South African gold mines. The advantage of using the numerical method is that the influence of the abutment openings and mining sequences on the induced stress concentration surrounding the ore passes can be evaluated. Although these methods provide some useful tool to estimate and quantify the extent of damage zone around ore passes, none of these geotechnical methods accounts for the influence of rock structures and their interaction with stress states on the degradation of ore pass systems.

If an ore pass intersects large scale structural features it is convenient to use limit equilibrium analysis packages such as Unwedge, available from Rocscience (2003). This is acceptable if the structures are assumed to be of infinite length. Quite often, however, the structural regime surrounding an ore pass is characterized by finite length fractures, then it is arguably best to try to simulate a 3D rock mass representing all potential wedges that may form along the ore pass. Stacey et al. (2005) and Hadjigeorgiou & Grenon (2005) used such an approach to investigate the stability of vertical or near vertical excavations by first generating a series of fracture systems and then determining the stability of every individual wedge exposed along the walls of the excavation. Although such approaches are arguably more realistic than assuming infinite length fractures they can be inadequate when excavation stability is controlled by stress rather than structure.

There are several stress analysis packages that can be used to investigate the stability of ore passes. Of interest is the work of Sjoberg et al. (2003) where they looked at stress but also the resulting modified geometry due to wear. Using FLAC, Itasca (2000) they investigated the impact on the integrity of an ore pass of a groove on the floor caused by wear during material flow. The model was based on observations from the Kiirunavaara mine in Sweden. In another example for an ore pass stability analysis at Creighton Mine in Sudbury Canada, Kazakidis & Morrison (1994) used the capacity of Ubiquitous joint analysis in Examine2D to determine the influence of a shear zone orientation on mining

134


induced stresses. A significant limitation of these stress analysis packages is that the complexity of rock structural behaviors cannot be entirely integrated in them.

7.3 Definition of the #19A ore pass problem

The #19A ore pass is belongs to the ore pass complex 18-21. The ore pass complex 18-21 is located on the North side of 1125 mining block, between mining zones 20 and 21. It consists of the ore pass systems #18, #19, #19A and #21, Figure 7-1. Andrieux et al. (2006) suggested that mining zones 20 and 21 are not advancing directly toward each other, but rather past each other, Figure 7-2. This has resulted in high stress state concentration around the ore pass complex 18-21 and seismic activity nearby.

l^ ,

U«.H> i . " " ' ■ ' H

Figure 7-1. Schematic longitudinal section of the Brunswick Mine looking North-West, indicates the location of Zones 20 & 21 at the bottom mining block of the mine and the ore pass

complex 18-21.

135


f complex 18-21

- F ^ N

Figure 7-2. Advancing directions of the mining zones 20 & 21 with respect to the ore pass complex 18-21, after Andrieux et al. (2006).

Ore passes #19 and #21 were abandoned in 2004 due to considerable enlargement of their cross sections. In fact ore passes #19 and #19A merged, thus necessitating the paste backfilling of the #19 ore pass in order to inhibit further expansion. Undertaken laser cavity surveys demonstrated that damage to ore pass #18 was considerably less than the other ore passes. This was explained by the presence of an adequate distance between the ore pass and other infrastructure. Furthermore, it was suggested that the expansion of the other ore passes close by resulted in a stress shadow region for the #18 ore pass. The original and current dimensions of ore pass complex 18-21 are summarized in Figure 7-3.

I 125 4Sul .

M25-2Sul>

1125 4Sub

1125-2Sub

;.)

Figure 7-3. Ore pass complex 18-21, a) Original dimensions, b) Dimensions after degradation.

With mining expected to progress in the 20 & 21 mining zones, it was necessary to investigate the long-term integrity of ore pass #19A. This ore pass has a sub-vertical orientation of 85° and an upper section of 50 m long with an initial diameter of 3 m while

136


the bottom section has a length of 70 m and initial diameter of 6 m. By 2007, almost four million tons of broken ore has been transferred through the #19A ore pass. In December 2007, the decision was made to abandon the upper section of the #19A ore pass and fill it with paste fill, Harrisson (2008). The mine plans to keep in operation the segment between the silo below the second sublevel and the crusher (sill). Furthermore, a grizzly will be installed to control material flow inside the silo. The #18 ore pass remains a critical infrastructure that serves the North part of the mine between the 1125 sill and the 1000 -3 sublevel. This decision was influenced by the transfer of significant resources into reserves in August 2007. A portion of these reserves are located above the #19A ore pass on the 4 sublevel.

7.4 Stress estimation around the ore pass complex 18-21

For the purposes of this thesis a two stage numerical modeling approach was implemented. For practical purposes a 3D boundary element stress analysis package, Map3D-V50, developed by Mine Modeling (2008), was used to simulate all mining sequences between early 2002 and early 2007 in zones 20 and 21. This numerical model provided a reasonable estimate of the state of principal stresses around the #19A ore pass that were used as boundary conditions to the local ore pass stability analysis.

The input data for the Map3D model are presented in Appendix-A, Figure A-9. The gradients of principal in-situ stresses used for the Map3D model are as follows o~N_s — 0.044 MPa/m, aE_w = 0.055 MPa/m and ov = 0.028 MPa/m . The other input parameters for the Map3D model were the ones employed by the mine in order to undertake stability analysis and to compute the influence of different mining sequences. The results obtained by these Map3D simulations have been extensively used by the mine to interpret the observed performance of the rock mass. This is valuable empirical evidence.

The mining sequences between early 2002 and early 2007 in zones 20 and 21 has been summarized in Table 7-1. In order to estimate the stress states around the ore pass complex 18-21 the Map3D was used. First, the dimensions of the # 18, # 19 and # 2 lore passes were integrated in the Brunswick Mine Map3D model. The mining sequences in the zone 20 and 21 were then simulated in the model. In all 18 mining sequences were considered. Finally, the stress states due to the mining sequences were measured on three grid sections. One horizontal grid and two vertical grids were placed in the model. The vertical grids considered perpendicular to each other in East-West and North-South directions, Figure 7-4.

The stress states on the three grid sections were recorded during each mining step. Consequently, the measured stresses on the grids were decomposed in the three principal directions of x, y and z to estimate the average stress states along these directions for each mining step. In the Brunswick Mine model the Y axis is parallel to the North direction and the X axis is parallel to the East direction. Figure 7-5 illustrates the variation of average stress states in the different directions throughout the mining step 1 to 18.

137


Table 7-1. Mining Sequences at Zones 20-21 between June 2002 and Feb 2007.

Stope Name Year of Excavation

Date of extraction

Mining Sequence

Mining Zone

336-4 2002 Jun 2002 1 20

241-6 2002 Aug 2002 2 20

335-6 & 336-6 2002 Oct 2002 3 20

341-6 2002 Oct 2002 3 20

550-3 2002 Oct 2002 3 21

350-3 2002 Oct 2002 3 21

235-4 2002 Dec. 2002 4 20

536-6 2003 Jan 2003 5 20

242-4 2003 Apr 2003 6 20

342-6 2003 Jul 2003 7 20

244-4 2003 Sep 2003 8 20

249-3 2003 Sep 2003 8 21

536-4 2003 Oct 2003 9 20

542-6 2003 Oct 2003 9 20

241-4 2003 Dec 2003 10 20

350-2 2003 Dec 2003 10 21

245-4 & 246-4 2004 Feb 2004 11 20

550-2 2004 Feb 2004 11 21

249-2 2004 May 2004 12 21

344-6 & 345-6 2005 Jan 2005 13 20

544-6 & 545-6 2005 Apr 2005 14 20

346-6 2006 Mar 2006 15 20

546-6 2006 May 2006 15 20

349-2 2006 Jul 2006 16 21

247-4 & 248-4 & 249-4 2006 Sep 2006 17 20

536-2 2006 Dec 2006 17 20

549-2 2007 Feb 2007 18 21

138


Zone 20 Zone 21

Figure 7-4. The vertical grids placed perpendicular to each other in the Map3D model.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Mining Sequence

Figure 7-5. Variation of stress states in different directions around the ore pass complex 18-21 during the mining sequence 1-18.

139


The results of global stress modeling suggested that the employed mining sequence had resulted in stress variations including progressive relaxation in the area of interest between the #19 and #21 ore passes. The information extracted from the stress analysis was employed in the local model by first employing the stresses during mining step #6 (o"xx = 75 MPa, Oyy = 40 MPa, a^ = 45 MPa). Once the local model reached to an equilibrium condition, it was decided to introduce the influence of mining induced relaxation after mining step #13, (oxx = 70 MPa, ayy = 22 MPa, a^ = 50 MPa). The procedures of stress installation and relaxation for the local models are discussed in the following sections.

7.5 Stability analysis of the # 19A ore pass

The #19A ore pass has been constructed in the competent massive sulphide rock mass. There is no evidence of major geological structures near the ore pass as the main mine dyke passing about 45 m of the North-West side of the ore pass section. Nevertheless, the massive sulphide rock mass is characterized by three dominant fracture sets, two of which are sub-vertical and one is sub-horizontal. The characteristics of fracture sets are listed in Table 7-2. A practical decision was taken to use the 3D boundary element model Map3D to construct the global model and the PFC for the local model.

Table 7-2. Fracture sets characteristic for Massive Sulphide rock mass.

Fracture set#

Fracture Characteristics

Fracture set#

Orientation Normal Set Spacing Trace Length Fracture set#

Dip

0 Dip

Direction

0 K Mean

(m) Std.

(m) Linear

frequency (m1)

Mean (m) Std.

1 89 007 17 1.52 1.8 0.65 1.40 0.73

2 89 274 12 1.12 1.0 0.89 1.46 0.54

3 17 227 57 1.23 0.8 0.81 1.16 0.22

In order to investigate the interaction of stress and structure on degradation of the # 19A ore pass, the 3D fracture system model developed for the massive sulphide rock mass (Chapter 4) was used to construct 2D and 3D synthetic rock mass models. Once the synthetic rock mass models were constructed, the dimension of the #19A ore pass was introduced in the models. Finally, the response of the models to the interaction of stress and structure was monitored by measuring the induced damages initiates and accumulates around the ore pass walls.

140


7.5.1 2D and 3D synthetic rock mass model

In order to construct the 2D and 3D synthetic rock mass models, the validated fracture system developed for the massive sulphide rock mass was employed, Figure 7-6. A horizontal plane was intersected with the 3D fracture system model to obtain fracture traces in the plane, Figure 7-7.

Z 0

20 -20

Figure 7-6. Fracture system model of the massive sulphide rock mass, (40 m x 40 m x 40 m).

141


20

10

' .-' • * - ■■

■ . y - - " ' \ -■ y ■>

p Safe

Figure 7-7. Trace of fractures on the horizontal section.

7.5.1.1 Simulation of intact rock properties

The inverse calibration method was used to establish the necessary micro-mechanical parameters for the generation of the 2D and 3D bonded particle models (BPM) that result in appropriate intact rock properties.

A minimum particle radius of 0.005 m was used for the intact rock simulation in the 2D BPM. The same particle size was then employed for the large synthetic rock mass model. The micro-mechanical properties of particles and bonds used in the 2D BPM are summarized in Table 7-3. The mechanical properties for the massive sulphide rock at Brunswick Mine obtained from laboratory tests, compared with the results of computational PFC2D, Itasca (2004) tests, Table 7-5.

142


Table 7-3. Micro-mechanical properties of 2D BPM model.

Particle Micro-Mechanical Parameter Value Bond Micro-Mechanical Parameter Value

Particle Density (kg/m3) 4300 Parallel Bond Modulus (GPa) 85

Particle Contact Modulus (GPa) 85 Bond Strength (MPa) 144

Particle Friction Coefficient 0.5 Bond Strength Standard Deviation (MPa) 32

Particle Normal/Shear Stiffness 2.5 Bond Normal/Shear Stiffness 2.5

For the 3D BPM model the minimum spherical particle radius was fixed at 0.07 m for both the intact rock simulation and 3D synthetic rock mass generation. The micro-mechanical properties of particles and bonds used in the 3D BPM are listed in Table 7-4. The results of the computational PFC3D tests, for the massive sulphide rock at Brunswick Mine, are summarized in Table 7-5.

Table 7-4. Micro-mechanical properties of 3D BPM model.

Particle Micro-Mechanical Parameter Value Bond Micro-Mechanical Parameter Value

Particle Density (kg/m3) 4300 Parallel Bond Modulus (GPa) 121

Particle Contact Modulus (GPa) 121 Bond Strength (MPa) 140

Particle Friction Coefficient 2.5 Bond Strength Standard Deviation (MPa) 45

Particle Normal/Shear Stiffness 3.6 Bond Normal/Shear Stiffness 3.6

Table 7-5. Mechanical properties of intact rock in laboratory and 2D and 3D BPM.

Mechanical Properties Massive Sulphide 2D BPM 3D BPM

Uniaxial Compressive strength (MPa) 205 206 204

Modulus of Elasticity (GPa) 104 104 105

Poisson Ratio 0.29 0.28 0.29

Brazilian Tensile Strength (MPa) 4 45.5 52

143


The calibration process, in the present work, resulted in a greater tensile strength of the 2D and 3D bonded particle models than the tensile strength of the intact rock. The small ratios of uniaxial compressive strength to tensile strength (oc/at), along with the low internal friction angle for the BPM are the two common limitations in PFC intact rock modeling. Potyondy & Cundall (2004) suggested that using a smaller particle size can improve the (oc/ot) ratio. However, particle size reduction can result in increasing of the total number of particles in a model and subsequently increase the computational time of the numerical modeling.

Several strategies have been verified to overcome these limitations in intact rock modeling employing PFC. This includes the using of irregular particle shapes and deformable particles. Ting et al. (1993) used elliptical particles to increase the internal friction angle of granular particle assemblies. Potyondy & Cundall (2004) suggested the using of cluster of particles to create irregular shape of particle grains in which neighboring particles are bonded together with high bond strength. Fakhimi (2004) proposed slightly overlapped circular particle assembly approach to overcome the inherent limitations of low (cc/ct) ratio and low internal friction angle in BPM. Although partial successes were reported using these approaches, none of these techniques could fully resolve the problem. More recently Cho et al. (2007) suggested using clumped particles to give an irregular shape to the particles assembly. Using PFC2D they showed that assembly of clumped particles can result in better agreement between laboratory tests and computational tests. This approach has yet to be successfully implemented in PFC3D. Other issues such as compatibility of this approach with fractured rock mass simulations must also be reviewed.

The influence of the (ac/o"t) ratio on initiation and extent of damage zone around underground excavations is arguable. Fakhimi (2004) suggested that "in the vicinity of an unlined excavation, the state of stress is uniaxial. Therefore, calibration of the synthetic rock for elastic modulus, Poisson ratio and uniaxial compressive strength should be good enough to simulate surface spalling and rock damage similar to those observed in actual situations".

7.5.1.2 Simulation of fracture properties

The present analysis employed the Smooth Joint model to simulate the mechanical behavior of fractures. Using the results of intact rock calibration a biaxial test in 2D and a triaxial test in 3D was conducted on four specimens, each intersected by a fracture at a different inclination of 45°, 50°, 55° and 60°. Using a confinement stress of 1 MPa the axial stress initiating sliding along the fracture surface was recorded.

In this study for 2D model, zero bond strengths were assumed across the fracture surface and the normal and shear stiffness of particles along the fracture surface were modified to 3xl012 N/m3 and 3xl012 N/m3 respectively. Particles directly adjacent to a fracture were assigned a friction coefficient of 0.1. This has resulted in the desired macro-mechanical properties of fractures, zero cohesion and an angle of friction of 30°.

144


For 3D model the bond strength along the fracture surface was assigned to zero and all particles directly adjacent to a fracture were assigned a friction coefficient of 0.2. The normal and shear stiffness of particles along the fracture surfaces were 2.5xlO,2N/m and 0.5xl012N/m3 respectively. This calibration exercise resulted in fractures in the synthetic rock mass samples with the desired macro-mechanical properties of zero cohesion and an angle of friction of 30°.

7.5.1.3 Synthetic rock mass generation and in-situ stress installation

In order to construct the 2D synthetic rock mass model a square assembly of particles of 12 m x 12 m was generated. The dimension of model was chosen to be large enough with respect to the 3 m ore pass diameter. The assembly was populated by 117792 particles of varying size, with minimum particle size of 0.005 m. The ratio of maximum to minimum particle size was 1.66. The micro-mechanical properties obtained from the intact rock calibration, listed in Table 7-2, were used for the particles and bonds in the model. Once the bonded particle model was constructed, the results of the global stress model were applied to the 2D bonded particle model using a wall based stress installation. The rigid walls used in the generation of the bonded particle model were employed to confine the bonded particles until the desired boundary stresses were achieved. In the first step the stresses during mining step #6 (axx = 75 MPa, Oyy = 40 MPa) was applied to the 2D model, Figure 7-8. Following the in-situ stress installation, the trace of fractures obtained by intersecting the horizontal plane with the 3D fracture system model was transferred to the pre-stressed bonded particle model. Consequently, a 2D synthetic rock mass model of 12 m x 12 m was constructed, Figure 7-9.

145


* ayy = 40 MPa

i

12m

Y

. 12m X

axx = 75 MPa

12m X

r

t

axx =75 MPa

ayy = 40 MPa

Figure 7-8. In-situ stress installation on the 2D bonded particle model.

Figure 7-9. Construction of the 2D synthetic rock mass model by linking the 2D bonded particle model and the trace of fractures on the horizontal plane.

146


Modeling the full length of the ore pass section (50 m) in a 3D model would have resulted in prohibitively long modeling times. Consequently a tactical decision was made to construct a 3D synthetic rock mass model for only 10 m length of the ore pass.

For the 3D synthetic rock mass model, a 1 2 m x 1 2 m x 10m bonded particle model containing of 167576 spherical particles of minimum size of 0.07 m was generated. This synthetic rock mass size is bigger than the REV size of the rock mass calculated in the previous chapter. In order to generate a uniform particle size distribution, the ratio of maximum to minimum particle size was fixed at 1.66. The micro-mechanical properties obtained from the intact rock calibration, summarized in Table 7-3, were used for the particles and bonds in the model. The state of stresses during mining step #6 (oxx = 75 MPa, Oyy = 40 MPa, azz = 45 MPa), from the global model constructed in Map3D, were applied to the 3D model, Figure 7-10. Once the stresses were installed in the model, the fracture geometry data from the 3D fracture system model (Figure 7-6) were transferred to the pre-stressed bonded particle model. Figure 7-11 illustrates the 3D synthetic rock mass model using FishLab-vl8, Itasca (2008b), for visualization.

In both 2D and 3D synthetic rock mass model the size of fracture system model is approximately three times greater than the bonded particle model to avoid introduction of boundary truncation-errors during the link procedure. The resulting fracture system is one of many possible systems. Consequently, it is necessary to investigate the range of possible fracture systems for any given set of data. This was not undertaken in this work as the main objective was to investigate the interaction of fracture systems and stress states on the stability of the #19A ore pass.

12m

a« = 75 MPa

ayy = 40 MPa

■ * ' ■ ' '

Figure 7-10. Installation of in-situ stresses on the 3D bonded particle model.

147


Figure 7-11. Representation of the 3D synthetic rock mass model.

7.5.2 2D Stability analysis of the #19A ore pass

The PFC2D has been successfully used to simulate stress induced damage around underground openings. Potyondy & Cundall (1998) employed PFC2D to estimate brittle failure zone around Mine-By Experiment tunnel in Underground Research Laboratory (URL) in Manitoba, Canada. The approach also was employed by Potyondy & Autio (2001) and Fakhimi et al. (2002) to predict damage formation around circular holes subjected to compressive loading. Tannant & Wang (2004) investigated the influence of liner on initiation and extent of stress damage zone around tunnels using PFC2D. These examples exhibit the potential of the PFC in simulation of rock mass damage in adjacent of underground openings.

In order to evaluate the stability of the #19A ore pass, a circular hole of 3 m of diameter was created in the 2D synthetic rock mass model (2DSRM) by deleting particles in the center of the model, Figure 7-12a. The model was run for 600000 cycles where it reached relatively to an equilibrium condition (no extra bond failure was recorded after 600000 cycles).

Once the dimensions of ore pass were introduced in the 2D fractured rock mass model, deformation of the rock mass was initiated with low frictional displacements along the

148


pre-existent fractures and in the form of rock matrix deformation around the ore pass boundary. Figure 7-12b shows the displacement vectors in the 2D model once the ore pass was excavated.

This initial deformation was followed by nucléation and development of shear and tension cracks, mostly in the South and North side walls of the ore pass (stress concentration zones). In the PFC models a crack is simulated when a bond between two neighboring particles is broken. In the investigated 2D model, the majority of the ruptured bonds (more than 80%) failed in tension. The tension cracks were oriented predominantly parallel to the excavation boundary. The pre-existent fractures served as stress concentrators. The fracture tips and the point of intersection between fractures were highly stress concentration spots. This has resulted in local change of stress states which led to grow of new extension cracks from these spots. Figure 7-13 illustrates the contact-force distribution in the ore pass model where the thickness of the lines is proportional to force magnitude. It can be seen that fracture tips and the points of intersection between fractures acted as stress concentrators. The failure process continues by the propagation of cracks between pre-existent fractures in the form of breaking of intact rock bridges between neighbor fractures and between fractures and the excavation boundaries. Finally, when sufficient damage is accumulated around the ore pass wall, the stress induced fractures together with pre-existent fractures creates rock wedges which move along the fractures or fall from the ore pass wall.

Figure 7-12. a) Simulation of a 3m diameter ore pass in the 2D synthetic rock mass model, b) The vectors of velocity in the 2D ore pass model, some instance after the excavation, Y axis

represents the North.

Figure 7-14 presents the evolution of damage zones around the #19A ore pass for different stages of the model. The damage was localized and concentrated in the North and South wall side of the ore pass and in the adjacent of pre-existent fractures in stress concentration zone. Tension cracks are identified in blue while shear cracks are red.

149


12m

12m

Figure 7-13. Contact force distribution in the ore pass model (the thickness of the lines is proportional to force magnitude).

Once the model reached equilibrium, the applied stresses were reduced to simulate stress relaxation (axx = 70 MPa, oyy = 22 MPa). For this purpose the confining walls which were used at the stage of bonded particle model generation, were retreated over a certain number of calculation steps until the desired stress states were achieved. This was done slowly enough to give enough time to the particle assembly to adjust to the force redistribution that accompanies slip and bond breakage. The process was accompanied by the development of further cracks around the ore pass wall and in the rock mass mostly in the form of tension cracks. Figure 7-15 exhibits a zoom view of the failure zone around the #19A ore pass after stress relaxation.

The displacement vectors in the 2D model, after stress relaxation, indicate that the damage zone has been extended deep in the rock mass, Figure 7-16a. Moreover, it can observe that the presence of fractures has changed the pattern of failure around the excavation which is different from the classical notch shape failure generally observed in massive rocks, Figure 7-16b.

150


Figure 7-14. Evolutions of damage zone around the 3 m of diameter #19A ore pass in the 2D model.

151


Figure 7-15. Failure zone around the #19A ore pass after stress relaxation (zoomed view).

Figure 7-16. a) Displacement vectors in the 2D model after stress relaxation, b) The pattern of developed micro-cracks around the # 19A ore pass.

An important advantage of the 2D model is the low execution time. The model was run for 600000 cycles for 162 hours. In addition, in the 2D model, the rock mass deformation and damage can be measured easier. This has facilitated the understanding and tracing of the failure mechanisms in the 2D model. However, the 2D model cannot fully represent the structural complexity of the rock mass. The fracture system and the stress regime are three dimensional which are better represented in a 3D model.

152


7.5.3 3D Stability analysis of the #19A ore pass

The stability of the #19A ore pass was simulated using the 3D synthetic rock mass model. For this purpose a gravitational acceleration of 9.8 m/s2 in vertical (Z) direction was applied to all of the particles. Finally, a 3 m diameter and 10 m long section of the #19A ore pass was simulated by deleting the particles in the center of the model. The model was run for 700000 cycles during 514 hours where it reached to an equilibrium condition and no more bond rupture was occurred.

Due to the three dimensional complexity of the fracture system, detailed visualization of the fractured rock mass deformation is difficult. The failure begins with initiation of tension and shear cracks in the ore pass boundary and in the fractured rock mass far from the ore pass. Failure zones develop in the South and North wall side of the ore pass by accumulation and growing of the micro cracks. Figure 7-17 represents a plane view of the model where the cracks are superimposed along the 10 m ore pass length. More than 80% of the ruptured bonds fail in tension. The tension cracks are blue in Figure 7-17 and the shear cracks are red.

• " la- •

/£ V->„ •' 4 •

Figure 7-17. A plane view of the failure zone in the 3D model (superimposed).

153


Figure 7-18 is a cross section of the 3D ore pass model, 2 m above the center point (Z = 2 m). The figure illustrates the extent of damage zone around the ore pass in different directions. It indicates the damage propagation until the model is reached to an equilibrium condition.

»

' • ,

• i

» i

Figure 7-18. A cross-section of the extents of damage zone around the #19 A ore pass in the 3D model (Z = 2 m), (Blue: Tension crack, Red: Shear crack).

The magnitude of induced stresses at a distance of one meter away from the ore pass boundaries was measured using twelve measuring spheres of one meter of diameter each. Three measuring spheres were placed at different elevations of (Z = -3 m, 0 m and 3 m) to calculate the average induced stresses along each wall side. In the North and South walls, the average magnitude of tangential stress (measured from six measuring spheres) increased to almost 130 MPa. Once sufficient damage was accumulated, adjacent of the ore pass walls, the tangential stress in the North and South walls decreased to approximately 90 MPa. The induced stress reduction in the damaged zone implies that this zone was gradually unloaded. Calculation of the average tangential stress in the East and West walls of the ore pass showed that the stress magnitude remained between 65 MPa to 75 MPa during the simulation.

In order to evaluate the stability of the ore pass walls along the North-South and East-West directions, two longitudinal sections of a uniform thickness of 1.5 m were considered for further stability monitoring. Figure 7-19a displays the longitudinal section of the #19A ore pass along the North-South direction immediately after excavation of the ore pass. Particle or clusters of particles which are identified by brown, orange and yellow colors represent rock wedges delineated by the ore pass walls and pre-existent

154


fractures. The rock mass is represented by grey. The figure shows that a small number of wedges (~ 4 wedges) exposed along the ore pass walls. This can be justified by the relatively small size of fractures in the rock mass. Figure 7-19b exhibits the same section once the stress induced damage was developed. Several large and small clusters of particles (rock wedges) have been created along the North and South wall sides of the ore pass. The resulting clusters of particles can fall and slide into the ore pass, Figure 7-19c. Finally the shear cracks (red color) and tension cracks (blue color) developed in the North-South section, for a uniform thickness of 1.5 m, was superimposed to the model, Figure 7-19d. A comparison of Figure 7-19d and Figure 7-19b indicates a correlation between the extents of stress induced cracks developed in the model and the damage zone represented by cluster of particle detached from the original rock mass. Figure 7-19d also shows that the extent of damage zones in North-South direction varies along the 10 m of the ore pass length. The extent of damage zone in some elevations is more than 2 m from the boundary of the ore pass wall.

10m

a)

C)

Figure 7-19. a) A longitudinal section of the #19A ore pass along the North-South direction a moment after the ore pass excavation (particle or cluster of particles with brown, orange and yellow

colors represent rock wedges), b) the same section after stress induced damage, c) The vectors of particles displacement along the North-South walls, d) The location of bond failures along the North-

South walls, (Blue: Tension crack, Red: Shear crack).

155


On the other hand, no significant stress induced damage was observed along the East-West direction. This is due to the relatively low induced stresses concentrated along these wall sides. Figure 7-20a represents a longitudinal section of the ore pass model along the East-West direction immediately after the ore pass was excavated. The section has a uniform thickness of 1.5 m. Similar to the North and South walls there are not many wedges exposed along the East and West wall sides once the ore pass is excavated in the fractured rock mass. By running the model to the equilibrium condition, some new rock wedges (cluster of particles) have been created along the ore pass wall and inside the fractured rock mass, Figure 7-20b. Figure 7-20c shows the displacement vectors of falling particles including the sliding of a pre-existent wedge in the East wall side. The new rock wedges created in the model are associated with initiation of shear and tension cracks in the rock mass. Figure 7-20d represents the shear cracks (red color) and tension cracks (blue color) developed in the East and West wall sides, for a uniform thickness of 1.5 m. The cracks have been developed locally around the new rock wedges. Based on these observations, the failure mode in the East and West wall sides are mostly structural failure.

3 m

10m

a)

c)

Figure 7-20. a) A longitudinal section of the #19 A ore pass along the East-West direction a moment after the ore pass excavation (particle or cluster of particles with brown, orange and yellow

colors represent rock wedges), b) the same section after stress induced damage, c) The vectors of particles displacement along the East-West walls, d) The location of bond failures along the East-

West walls, (Blue: Tension crack, Red: Shear crack).

156


Subsequently, as discussed earlier, the applied stresses were reduced to simulate stress relaxation (oxx = 70 MPa, Oyy = 22 MPa, a^ = 50 MPa). This was accompanied by the development of some further cracks, but these were not critical and did not extend the damage zone. However, the stress relaxation resulted in an acceleration in the movement of detached particles (rock wedges) formed along the North-South wall directions.

7.6 Comparison of the PFC numerical models to field observations for the #19A ore pass

The first cavity monitoring survey (CMS) of the #19A ore pass was done by the mine personnel in 2003. Figure 7-21 represents two cross-sections of the #19A ore pass obtained from the CMS results in 2003 and 2006. The ore pass expanded mostly along the South and North wall sides. Based on the 2006 CMS results, the diameter of the upper section of the ore pass expanded from 3 m to a maximum of 14 m along North-South direction and to a maximum of 6 m along the East-West direction at 1563 elevation, Figure 7-22a. In all the ore pass expanded 5 times of its original volume. However, the expansion is not uniform along the ore pass length, Figure 7-22b.

N

L 30/10/2003

Elevation 1584

27/11/2006

Figure 7-21. A cross section of the CMS results for the #19A ore pass in 2003 and 2006.

157


N

i

a) Elevation 1563 b)

Figure 7-22. a) A cross section of the CMS results for the #19A ore pass indicating the dimensions of expansion in different directions at 1563 elevation, b) A Longitudinal section of the

CMS results for the 19A ore pass (looking West).

Comparison of the results of 2D and 3D numerical models with the field observations obtained from the CMS revealed that although the pattern of crack propagation observed in the numerical models follows the orientation of the observed ore pass degradation in the field, the models fail to predict the actual magnitude of the failed zone. In this context, "failed" refers to actual excavation expansion for the #19A ore pass. In fact, the results of cavity monitoring observations suggest that the ore pass expanded to almost five times of its original volume and approximately 3 to 10 times of its original cross section area in different elevations.

There are two possible explanations for this discrepancy:

• The first possibility is that the constructed model may not have been adequately calibrated. Improved calibration would involve more simulations and further attention to the influence of particle size and shape and different micro-properties of particles and bonds on the macro-properties of a PFC model. A promising avenue for this problem is to account for the influence of stress corrosion as employed by Potyondy & Cundall (1998) and Chandler et al. (2002) for brittle failure simulation and stability analysis at Underground Research Laboratory (URL) in Canada. Based on PFC2D models they have suggested that using the stress corrosion approach can result in better approximation of brittle failure zone around the underground excavations. The stress corrosion is a time-dependent

158


weakening process implemented in the Particle Flow Code. This problem necessitated that they calibrate their model not only for short term mechanical properties (UCS, E and u) but also for long term mechanical properties. For calibration of the PFC2D model for the long-term mechanical properties they have simulated a static-fatigue test.

• A more plausible interpretation of the discrepancy, however, is that there are several degradation mechanisms in place. Ore pass degradation is also influenced by wall wear, abrasion, impact loading etc. during material flow. Furthermore, attempts to release flow, in hang-ups, using explosives can further lead to more ore pass degradation. This is the major difference between ore pass and other excavations such as shafts, ventilation raises and etc.

Figure 7-23 represents the results of a CMS for the ore pass complex 18-21 including a ventilation raise which has been constructed near the #19A ore pass. This ventilation raise has a diameter of 1.8 m and has been constructed in the massive sulphide rock mass. The in-situ stress states for this ventilation raise are almost the same as for the #19A ore pass. However, the results of the CMS show some minor failures in the North and South walls of the ventilation raise which is several times smaller than the extent of damage zones were observed in the ore passes.

^ N

21 OP

19AOP 19 OP

Exhaust Raise

Figure 7-23. The results of CMS for the #19, #19A and #21 ore passes compared with the results of CMS for an exhaust raise.

The impact of rock fragments with ore pass walls has a significant effect on expansion of the failure zones around the ore pass. Stacey & Swart (1997) stated that wear of ore pass walls due to the rock fragments impact is greater in the presence of stress scaling.

159


In order to evaluate the influence of particle impacts on the degradation of the #19A ore pass a simple numerical experiment was performed. The 3D model, in which the induced stresses have already damaged the rock mass, was chosen. A relatively large rock fragment of 0.9 m of diameter was generated inside the ore pass in the upper part. The rock fragment size was selected based on the dimensions of the scalpers (0.9 m x 1.4 m) employed in the Brunswick Mine. The same micro mechanical properties assigned to the particles in the 3D intact rock simulation were attributed to the rock fragment. The rock fragment was projected against the South and East walls of the ore pass with a constant velocity of 17 m/s in an angle of 30°. The results are presented in Figure 7-25.

12m

0.9 m

10m

3 m

Figure 7-24. A projectile particle impacting the South wall of the #19A ore pass.

Collision of the rock fragment with the South wall of the ore pass creates fourteen extra cracks in the rock mass. Five micro cracks (4 tension cracks and 1 shear crack) are created on the ore pass wall while nine other tension cracks are developed inside the rock mass in the stress induced damage zone. On the other hand, the impact of rock fragment with the East wall of the ore pass creates four tension cracks on the ore pass wall. Figure

160


7-25 shows longitudinal sections of the model along South and East walls, before impact and after rebound, respectively. The sections have a 1.5 m uniform thickness.

» .

a) b)

c) d)

Figure 7-25. a) Impact of rock fragment with the South wall side, b) Damage due to the rock fragment impact on the South wall side, c) Impact of rock fragment with the East wall side, d)

Damage due to the rock fragment impact on the East wall side.

This simple numerical model demonstrated that impact of a rock fragment with an ore pass walls can not only create damage along the ore pass wall surface but it can also accelerate the stress scaling in stress concentration zone. Due to the stress induced damage the rock mass in the stress concentration zones become weaker. Therefore, the damage zones can be more easily removed with the flow of rock fragments in the ore

161


pass. This can result in greater expansion of the ore pass along the damage zones. Finally, removal of the primary damaged zones, due to the flow of rock fragments, will result in re-concentration of stresses in the regions behind the primary damaged zones. This process will be repeated during the lifetime of an ore pass.

7.7 Conclusions

This chapter aims to address ore pass failure mechanisms in high stress states. The analysis is based on the synthetic rock mass approach where a fracture system was linked to a bonded particle model both in two and three dimensions. The approach allowed for a more realistic interpretation of the influence of stress and structure in the stability of ore pass systems.

In this analysis the #19A ore pass of the Brunswick Mine was selected as a case study. The results of 2D and 3D models demonstrated a failure zone development in the North and South walls of the ore pass where induced stresses were concentrated.

The 3D model has several advantages over the 2D model. These include better representation of the structural complexity of rock mass and more accurate interaction between a 3D stress regime and the defined rock mass structures. Moreover, the effect of gravity was integrated in the 3D model. Nevertheless, mechanisms of failure were traced in the 2D model. These mechanisms can be summarised as follow:

• Frictional displacements along the pre-existent fractures.

• Initiation of cracks (mostly tension cracks) in the intact rock in adjacent of the ore pass.

• Propagation of cracks between pre-existent fractures in the form of breaking of intact rock bridges between neighbor fractures and between the fractures and excavation boundaries.

• Creation of rock wedges and rock slabs which can displace along the fracture planes or can fall into the ore pass.

Interpretations of the result of the 3D model shows that the extents of the damage zone around the ore pass varies at different elevations along the ore pass length. Moreover, the impact of a flowing rock fragment with the stress induced damaged zone can lead to more failure along the ore pass wall.

It was recognized that the constructed models cannot yet fully explain the observed behavior in field. The models need further refinement and calibration in order to better predict the extent of damage zones. Moreover, it will be necessary to integrate the influence of material flow as a precursor to wall degradation.

162


Although there are some limitations in the proposed approach it can be used as a good tool to evaluate design alteration. Ore passes of different configurations (inclination angle, trend, cross section) can be simulated in the 3D model and the developed damage zones which result from different configurations can be compared.

The next chapter presents the application of the Particle Flow Code in quantification of the ore pass wall damage inflicted by material impact.

163

Chapter 8: Investigation of Ore Pass Wall Degradation Due to Material Impact

8 Investigation of Ore Pass Wall Degradation Due to Material Impact

8.1 Introduction

This chapter investigates in detail the problem of wall degradation in ore passes, as it is associated with the use of finger raises. Fragmented ore is transported from stopes, or production faces, to a tipping point, where this fragmented ore is dumped into the ore pass. When an ore pass intersects two or more production levels, finger raises are employed to funnel material into the ore pass. In this configuration, material flows into the finger raise and falls into the ore pass at the junction between the ore pass and finger raise. Material is subsequently drawn out of the ore pass through the use of a chute system.

Stacey & Swart (1997), Hadjigeorgiou et al. (2005), and Lessard & Hadjigeorgiou (2006) report that finger raises are often associated with operational problems. The drop of rock fragments from finger raises results in high impact loads acting on the walls of an ore pass, and can contribute to the degradation of the ore pass system. This can result in the enlargement of the area where a finger raise intersects the ore pass, Figure 8-1. This phenomenon has been confirmed by cavity monitoring surveys at several mine sites, Lessard & Hadjigeorgiou (2006).

The extent of the damage inflicted by impact loading of the ore pass walls is influenced by the type of material transferred, the finger and ore pass configuration, and the rock mass quality of the walls. The critical material characteristics are the rock fragment shape, hardness, density and size distribution. In addition, the capacity of ore pass walls to resist impact loads is determined by the rock mass characteristics and the induced stress regime. It is recognized that the presence of structural defects in the rock mass results in more pronounced wall degradation.

This chapter presents the results of numerical experiments that investigate the influence of ore pass and finger raise geometry on the level of ore pass degradation. The distinct element method (DEM), in particular the Particle Flow Code (PFC), was employed. A series of numerical experiments was used to simulate the influence of ore pass and finger raise configurations on impact loading of the ore pass walls. In these numerical models, the magnitude of impact loads generated by rock fragments on the ore pass walls were measured for a range of ore pass-finger raise configurations. An earlier work on this topic has been presented in summary by Esmaieli & Hadjigeorgiou (2009). This chapter provides full details and the considerable progress made in the work.

164


In the second series of numerical experiments, a block of rock mass along an ore pass wall was simulated with a bonded particle model. A projectile particle was thrown against the rock block, and the extent of damage zone inflicted by the projectile particle on the rock mass block was quantified.

Impact Damage Zone

Figure 8-1. Damage and wear zones in an ore pass, Hadjigeorgiou et al. (2005).

8.2 Degradation of ore pass structural integrity by rock fragment impact

In ore pass systems, the gravity movement of rock fragments includes rolling, sliding and inter fragment collision. The interactions of moving material and ore pass walls can result in the development of wear and/or impact damage zones. Wear is associated with the rolling and sliding of particles along a surface, resulting in the scouring of the wall surface. Damage attributed to impact loads can be caused by the fall of single boulders in the ore pass, or by a stream of rock or a large mass of material, Iverson et al. (2003). The mechanical properties of the rock mass along the ore pass wall can influence the extent of damage. Stacey & Swart (1997) note that wear of ore pass walls is greater in weak rock masses (RMR < 50) and in the presence of stress scaling. If the ore pass is located in a

165


rock mass with structural defects, the action of moving material can initiate further wall degradation, including falls of ground.

8.3 Finger raise configuration

Figure 8-2 illustrates a typical finger raise - ore pass configuration. Hadjigeorgiou et al. (2005) report that, in Canadian underground mines, finger raises have cross section dimensions of 1.5 m x 1.5 m and 1.8 m x 1.8 m. The fingers are linked to ore passes of larger cross section dimensions. Ore pass dimensions vary from 1.5 m to 4.4 m with a typical dimension of 2.1 m.

A well designed finger raise can minimize the ore pass wall damage and maximize ore pass longevity. Current practice is often based on empirical rules which are quite general and may not always be appropriate for site specific conditions. Hambley et al. (1983) and Ferguson (1991) recommend an inclination of 60° for finger raises in order to ensure free flow of rock fragments in the finger raise. This recommendation, however, does not seem to be respected at the Brunswick Mine, where a range of finger raise inclinations varying from less than 50° to more than 85° was observed. The guidelines proposed by Hambley et al. (1983) and Ferguson (1991) do not consider the importance of the ore pass inclination with respect to finger raise orientation.

The finger raise inclination influences the trajectory and interaction of rock fragments flowing in the ore pass and the resulting load on the ore pass wall. If the finger raises are steep, this will result in higher impact velocity on the ore pass walls. On the other hand, if the finger inclination is shallow, material flow is slow and can result in hang-ups. A steeply inclined finger raise results in a narrower pillar at the intersection of the ore pass and finger raise. This makes the pillar more susceptible to stability problems. Consequently, an operational design will use a finger raise inclination that will minimize impact load on the ore pass wall while will maintain free flow of material in the finger.

166


inger raise inclination angle |1

Impact Zone Angle of Intersection y

Figure 8-2. A typical finger raise-ore pass configuration.

8.4 Impact load simulation

This chapter reports on the use of numerical experiments to investigate the influence of ore pass and finger raise configuration on the impact of materials on the ore pass wall. Distinct element models and in particular, the Particle Flow Code (PFC), has been used by several authors, including Lessard & Hadjigeorgiou (2003), to investigate material transport in an ore pass system. This was justified given that the flow of granular material exhibits large-scale discontinuous dynamic behavior which is not well represented by conventional continuum-based approaches, such as the finite element methods. The distinct element method was also employed by Iverson et al. (2003), Nazeri & Rozgonyi (2003), Loughran et al. (2003) to evaluate the impact of rock fragments on several ore pass components.

Although the three dimensional distinct element methods have been used by several authors including Mustoe (2004) to investigate the impact problem in ore pass systems, most of the impact simulation problems associated with the ore pass systems have been done in two dimensions. This preference for 2D modeling is due to the simpler geometry and shorter execution time for the 2D models. In this study the 2D Particle Flow Code was used to model the movement and interaction of particles that represent rock boulders or fragments. By these means, it was possible to follow the trajectories of selected fragments and record their impact on the ore pass wall.

167


In order to construct representative models we relied on information on material properties collected in several Quebec underground mines, Hadjigeorgiou & Lessard (2007). Several ore pass and finger raise configurations were selected to quantify the influence of finger raise inclination on the resulting impact loads on the ore pass wall.

8.4.1 Simulation of rock fragments and ore pass wall

The physical and mechanical properties of rock fragments simulated in the PFC2D model include: rock size distribution, particle shape, normal and shear stiffness, density, friction coefficient and coefficient of restitution. The main source of input data for the numerical models was provided by Hadjigeorgiou & Lessard (2007) and Turcotte et al. (2003) where they described the methodology used to derive suitable material properties for PFC ore pass models. The material properties used in the context of the present work are summarized in Table 8-1.

Table 8-1. Material properties used in the PFC models.

Property Particles Ore pass wall

Density 4300 kg/m3 -

Normal stiffness 1.0* 109 N/m 1.0* 109 N/m

Shear stiffness 1.0* 109 N/m 1.0xl09N/m

Friction Coefficient 0.25 0.4

Particle size (radius) 0.12 m-0 .4 m -

Coefficient of restitution 0.3 -

Once the material properties were defined, it was necessary to select the material flow model to be used, whether single particle or multiple particles. Larson et al. (1998) simulated material flow in an ore pass through the use of a single moving particle in the ore pass. Although this provides a reasonable estimate of the resulting impact on the ore pass wall, it cannot account for collision between particles. Inter particle collisions can result in loads being applied to the ore pass wall which may be different than a single particle impact. Consequently, in this work it was decided to model multiple particle flow. This necessitated the choice of material distribution.

In the past, particle size distribution in an ore pass has been simulated by using an average particle size, Beus et al. (1998b) or by the generation of particle size distributions, Iverson et al. (2003). This approach provided for more comprehensive models, but it came at the expense of the model complexity and time of execution. Nazeri

168


(2001) and Lessard & Hadjigeorgiou (2003) used uniform size distributions and a maximum size deviation of 15% for their numerical models. This is justified if the mine has a good blasting practice. In the generated distribution, the smallest particle size was 30% of the largest particle size. Hadjigeorgiou & Lessard (2007) determined the size of the largest rock fragment in an ore pass based on the dimensions of the grizzly installed at the tipping point. In the absence of grizzlies the largest particle size was estimated as dm based on visual estimations and, in certain cases, image analysis from data collected in Quebec underground mines.

Blasted rock, results in fragments of various shapes. In PFC2D, the basic particle shape is circular, but it is possible to design different shapes by grouping circular particles together. For the purposes of the numerical analysis, circular rock fragment shapes were used in order to expedite the time necessary to run the models. The normal and shear stiffness of particles reported in Table 8-1 were based on a series of numerical experiments in which the model behavior was calibrated with respect to laboratory and field data. Nazeri (2001) demonstrated that the use of large contact stiffness values results in larger impact forces on the walls of the ore pass. The constructed PFC models used a slip-model, defined by a friction coefficient between particles to control their frictional characteristics. Based on previous work experiences by Turcotte et al. (2003), the friction coefficient was assigned a low value of 0.25 in order to prevent material hang-ups along the finger raise. It is recognized that the choice of friction value will influence material flow. The friction forces along with damping forces are responsible for a significant reduction in the kinetic energy of the flowing ore.

In order to simulate the collisions that would occur between particles flowing in the finger raise and the ore pass, it was necessary to establish appropriate values for the coefficient of restitution (COR) of rock fragments. Jung & Iverson (2004) noted that there are several acceptable definitions of COR. In the present work, the Coefficient of Restitution was defined as the ratio between the magnitudes of the rebounding and impacting velocities of the particle.

The coefficient of restitution of rock fragments falling or sliding along a surface depends on a variety of factors, including the size, shape, and type of the rock fragments, the geometry of the surface, the velocity of the rock fragments and the impact angle, Azzoni &deFreitas(1995).

The coefficient of restitution can be measured by both laboratory and in-situ tests. In-situ tests will provide values for the coefficient of restitution that take account of rock fractures and surface conditions as well as the rock material types. However, these tests are very expensive and there is no reference to the use of this method in a mining context. The rock drop test is the most popular test that can be done in the laboratory, Chau et al. (2002), Imre et al. (2008) and in the field, Azzoni & de Freitas (1995). This test can be simulated with numerical methods. The test is based on the initial height of the object before it is allowed to fall and the height of bounce after the impact. The COR for this test can be written as follows:

169


1 m x g x h i = - ( m x V l

2 ) (8-1)

CORv = y ( 8 - 2 )

CORv = (-r1)0 5 (8 - 3) hi

Where m is the particle mass (kg), Vj and Vr are the incoming and rebounding particle velocities (m/s), hj is the initial height of the particle before it is allowed to fall and hr is the height of bounce after impact (m).

In computer models like PFC, damping factors are used to simulate the coefficient of restitution of particles. There are several numerical damping methods such as local damping, viscous damping, and hysteretic damping. These methods are used to maintain numerical stability in PFC when simulating quasi-static processes. Itasca (2008a) suggested that local damping is inappropriate for particles in free flight under gravity or for impact of particles. When a dynamic simulation of compact assemblies is required, the viscous contact damping should be used.

There is no standard way of determining the coefficient of restitution of rock fragments, particularly in an underground infrastructure like an ore pass. For the purpose of these numerical analyses, a coefficient of restitution of 0.3 was assigned to the rock fragments. This is within the range of 0.2 to 0.6 reported by Iverson et al. (2003) based on physical pendulum tests. A PFC2D simulation of a single particle drop test on a rock mass with the properties of the ore pass wall was undertaken. In this experiment the viscous damping parameters were varied in order to arrive at the desired 0.3 COR reported in Table 8-1.

The ore pass walls absorb the dynamic impacts of rock fragments. A rigid wall property was considered for both the ore pass and finger raise simulation. This implies good rock mass quality of the ore pass wall where there is no structural defect. The stiffness and friction coefficient of the walls are listed in Table 8-1.

8.4.2 Simulation of ore pass and finger raise configuration

A total of 33 ore pass and finger raise configurations were modeled using PFC2D. Three different ore pass inclinations (a = 90°, 80° and 70°) were considered, Figure 8-3. The inclination of the finger raise (P) ranged from 30° to 80°, at 5° increments. The ore pass inclination (a) and finger raise inclination (P) result in different angles of intersection (y), as summarized in Table 8-2.

170


Table 8-2. Ore pass and finger raise configurations and resulting angles of intersection.

Ore pass Finger raise Inclination angle (P) (°) Inclination (a) (°) 30 35 40 45 50 55 60 65 70 75 80

Angle of Intersection

(Y)(°)

70 100 105 110 115 120 125 130 135 140 145 150 Angle of Intersection

(Y)(°) 80 110 115 120 125 130 135 140 145 150 155 160

Angle of Intersection

(Y)(°) 90 120 125 130 135 140 145 150 155 160 165 170

In the simulations undertaken, the ore pass dimension was 3.5 m and the maximum size of rock fragments was 0.8 m. In order to avoid hang-ups in the ore pass, a 3 to 5 ratio of ore pass dimension per maximum rock block size (D/d) was necessary, Hadjigeorgiou & Lessard (2007). The finger raise was assigned a width of 2 m, which is smaller than the ore pass width, Figure 8-3.

Particle Generation Zone

50 m

Figure 8-3. Ore pass and finger raise configurations used in the numerical modeling.

The ore pass was 50 m long and the finger raise was 20 m long. In order to simulate the particles in the finger raise a particle generation zone was constructed. In each numerical

171


experiment only one batch of rock fragments were generated in the particle generation zone. Each batch contained 80 circular particles of varying size ranging from 0.12 to 0.4 m of radius. Assuming that the particle shape can be extrapolated as a sphere the total volume for each generated batch of particles is approximately 5.5 m3, which corresponds to a typical bucket size of load-haul-dump (LHD) machine, used in Quebec underground mines.

8.5 Impact load on ore pass wall

Ten numerical experiments were performed for the 33 ore pass - finger raise configurations. In all 330 numerical experiments were performed to investigate material flow and the resulting impact loads on the ore pass wall.

The first step in the simulations was the generation of the particles which were introduced into the finger raise. Particles flowing through the finger raise collide with other particles and with the walls of the finger raise before entering the ore pass and hit the facing ore pass wall. During the simulations the following parameters were monitored: velocity and kinetic energy of particles, impact duration, average normal and shear impact force and peak impact load.

8.5.1 Influence of finger raise inclination on particle velocity and kinetic energy

The velocity of particles hitting a rock mass influences the extent of inflicted rock mass damage. Hutchings (1992) reported that the extent of impact damage depends on the number and mass of individual boulders striking the surface, and on their impact velocity. More specific to ore passes, Goodwill et al. (1999) suggest that erosion wear in ore passes is roughly proportional to the impact velocity raised to the power of 2.5.

Calculation of dynamic characteristics (i.e., velocity and kinetic energy) of flowing rock fragments in an ore pass is not an easy task. Hambley et al. (1983) used the trajectory method to calculate the dynamic characteristics of an individual rock fragment including its acceleration, velocity, position, linear momentum and translational kinetic energy as it dumps in an ore pass. Although this approach provides a simple tool to estimate the dynamic characteristics of a flowing rock fragment, it requires some assumptions and simplifications which may not be necessarily valid. It assumes a single particle and it does not consider the inter particles collision which can change the dynamic characteristics of flowing materials. It is also required that the initial velocity of the particle is known. Another critical assumption is that the angle of rebound is equal to the angle of impact.

In the undertaken PFC analysis the motion of selected modeled particles was defined by the resultant force and moment vectors acting upon it. This can be described by the

172


translational and rotational motion of a particle. Consequently the translational velocity of the particle, at any given time, can be expressed as, Itasca (2008a):

/ At\ ( At\ (F i t ) + g )At ( 8 - 4 )

Where:

V is the particle velocity, At is a time-step, F is the sum of all externally applied forces acting on the particle, m is the mass of the particle and g is the gravitational acceleration.

Figure 8-4 presents a time history of velocity for two discrete particles flowing in a 60° inclined finger raise. Particles generated in the generation zone are allowed to drop into the finger raise (time elapsed between point A and B in Figure 8-4). Once the particles enter the inclined finger, they begin to collide amongst themselves and the excavation floor (time elapsed between point B and C in Figure 8-4). The velocity of a particle travelling in the finger raise depends on the mass of the particle and the external forces acting on the particle via particle-particle or particle-wall collisions. The velocity of a particle increases with an increase in finger raise inclination and reaches a maximum at the impact zone on the ore pass wall.

oo 0.5 1 0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Time (Sec) — Particle size (R=0.35 m) Particle size (R = 0.28 m)

Figure 8-4. Temporal responses of two particles velocity, flow through a finger raise inclined at 60°.

173


In each numerical experiment, the velocities of two random particles were monitored. In addition for every finger raise and ore pass configuration the average impact velocity was calculated based on the results of ten numerical experiments. Figure 8-5 presents the effect of finger raise inclination on the impact velocity of particles at the ore pass wall for 70°, 80° and 90° ore pass inclinations. For finger raises with shallower inclinations, between 30° to 40°, there is only a slight variation in the impact velocity, irrespective of ore pass inclination. However, for steeper finger raise inclinations, the impact velocity in vertical ore passes is slightly higher. This is due to the longer distance that particles exiting the finger have to travel before they hit the ore pass wall.

u _o * - i u Q.

50 60

Finger Raise Inclination (°)

r

70 80

Figure 8-5. The influences of finger raise inclination on the particle impact velocity; for 70°, 80°, and 90° ore pass inclination.

The energy of rock fragments that enter a finger raise is dissipated through inter-particle collisions and through collisions with the ore pass walls. The amount of energy transferred to the walls depends on the deformation characteristics of the flowing rock particles and those of the wall surface. It is also recognized that particle impact generates heat and sound. However, these considerations were outside the capabilities of the designed numerical experiment.

In the undertaken numerical models the rock fragments were assumed as rigid particles, however, in the reality they are deformable and they may break when they hit the ore pass wall. The total kinetic energy of all particles accounting for both translational and rotational motion was traced. The kinetic energy in the PFC is defined by:

174


E k = - £ ( m f t 2 + /£a>iu>i) ( 8 - 5 )

Where Nb is the number of particles, mi, inertial mass, Ij, inertia tensor, Vj is the translational and ©j is the rotational velocity of particle i.

Figure 8-6 illustrates the influence of finger raise inclination on the kinetic energy of the whole batch of particles at the zone of impact. The kinetic energy of particles increases with an increase in finger raise inclination. This is more pronounced for vertical ore passes fed by steep finger raises. This is due to the higher impact velocity of particles for this type of configurations.

- i —

40 50 60 70


Ore Pass Inclination

A 90° A 80° A 70°

90

Figure 8-6. Influence of the finger raise inclination on the kinetic energy of particles.

8.5.2 Effect of finger raise inclination on impact load of particles on ore pass wall

The average force on an ore pass wall due to a single particle impact is directly proportional to the particle velocity and particle weight, and is inversely proportional to impact duration, Hambley et al. (1983). The monitoring of particles in the PFC models indicates that impact duration for a single particle is in the range of 1/100 of a second.

For flow of a stream of material, another parameter which can be monitored is the period of time between the first and the last particle impact (impact duration for one batch of

175


particles). The length of this period depends on the weight flow rate of particles in the finger raise. According to Goodwill et al. (1999) an increase in the flow rate of particles (and thus a decrease in the duration of the impact) can increase the impact pressure on the ore pass wall inflicted by a stream of rock fragments. Figure 8-7 illustrates that the impact duration for a batch of particles decreases as finger raise inclination increases. For a given finger raise inclination the impact duration for vertical ore passes is shorter, particularly when ore passes are fed by shallower finger raises.

T

30 40 50 60 70


Figure 8-7. Influence of the finger raise inclination on the impact duration.

The normal and shear impact forces on the ore pass wall were measured for 33 ore pass-finger raise configurations. Figure 8-8 presents the resulting normal and shear impact forces acting on ore pass walls inclined at 90°, 80° and 70° while the finger raise inclination was kept at 60°. The normal impact force is represented in blue, while the red indicates the shear impact force. The shear impact force on the ore pass wall is smaller than the normal impact force. However, decreasing of the ore pass inclination, increase the shear impact force on the ore pass wall. An increase in the shear impact force can potentially accelerate the rate of degradation at the ore pass wall.

176


a) Time (Sec)

NormalForce Shear Force

RM 9000 -8000 -7000 -6000 •

z *

9000 -8000 -7000 -6000 •

E o SOOO - t -i E

4000 • 3000 ■ 2000 -1000 -

0 — t u u t—b)

Tim» (Sec) NormalForce Shear Force

Impact

Forc

e (

KN

)

72 63

Impact

Forc

e (

KN

) Im

pact

Forc

e (

KN

) Im

pact

Forc

e (

KN

) Im

pact

Forc

e (

KN

) Im

pact

Forc

e (

KN

)

1000

o ■

c)

3 4 5 6 7 8

Time (Sec)

NormalForce ^—Shear Force

Figure 8-8. Normal and shear impact forces acting on inclined ore pass walls with the finger raise inclination of 60°; a) Ore pass inclination 90°, b) Ore pass inclination 80° and c) Ore pass

inclination 70°.

177


For each ore pass-finger raise configuration, the average normal and shear impact forces were calculated for ten repetitions of each experiment. Figure 8-9 illustrates the influence of finger raise inclination on the average shear and normal impact forces acting on the ore pass walls. The average normal impact force for all investigated ore pass inclinations is between 1200 to 1800 kN. The average shear impact force increases with a decrease in ore pass inclination. The ratio of the average shear impact force to the average normal impact force for vertical, 80° and 70° inclined ore passes is approximately 0.25, 0.42 and 0.55 respectively. The coefficient of variation for the average normal and shear impact forces recorded on the ore pass walls vary from 0.05 to 0.14, which indicates that there is only a small degree of variation among the average impact forces.

The results of average impact forces on the ore pass wall, however, should be interpreted with reference to the angle of intersection (y) between the raise finger and the ore pass, as defined in Figure 8-2. The highest average forces are recorded for angles of intersection between 140° to 145°. For a vertical ore pass, maximum impact loads were recorded when the raise finger was close to a 55° incline. This implies that when the finger raise inclination exceeds this critical value (55°), the mechanism of collision between particles and the ore pass wall changes from impact to frictional sliding.

178


1600 1400 1200 1000 S00

600

400

200

0

a)

zs:

20 30 40 SO 60 70

Finger Raise Incl inat ion (*)

♦ Avg. Normal Force90 » Avg. Shear Force90

80 90

b)

1800 1600 1400 1200 1000 800 600 ■100 200

0

20 30 40 SO 60 70

Finger Raise Inc l inat ion (*)

♦ Avg. Normal Force80 * Avg. Shear Force80

80 yo

2200

z 1É

2000 1800

HJ u 1

O u

1600 1400

tt 1200 a. E u 00 m OJ

1000 a. E u 00 m OJ

800 600 400 200

0

O

i r~

20 30 40 50 60 70 80 90

Finger Raise Inclination (*)

♦ Avg. Normal Force70 * Avg. Shear Force70

Figure 8-9. The influence of finger raise inclination on the average normal and shear impact forces; a) Vertical ore pass, b) Ore pass inclination 80°, c) Ore pass inclination 70°.

179


Beus et al. (1999) measured the magnitude of dynamic impact loads from rock fragments on the gate of an ore pass in an underground mine in Idaho. These were used in a comparison with PFC3D simulations. The results of these numerical tests were five times higher than the field data. This overestimation is probably due to the difficulties in modeling material properties such as particle shape, stiffness, coefficient of restitution and etc. Nevertheless, this approach can provide valuable information in comparing different design options.

The peak impact force of particles acting on the ore pass wall depends on the number of collisions that a particle incurs prior to striking the ore pass wall. A particle which incurs a higher number of collisions with other particles or with the finger raise will have a lower peak impact force on the ore pass wall. The distribution of large and small particles flowing in the finger raise can also influence the resulting peak impact force on the ore pass wall. If the larger particles are in front of flowing material, the peak impact force on the ore pass wall will be greater. However, if front part of material flow is composed of smaller particles, these particles may act as a cushion between the larger particles and the ore pass wall, and thus prevent the larger particles from colliding directly with the ore pass wall. For these reasons, the results of peak impact load on the ore pass wall indicate significant variance during the ten repetitions of each experiment. The coefficient of variation of peak impact load recorded on the ore pass walls varies between 0.25 and 0.4, a variation which is greater than the coefficient of variation determined for average impact forces. The relation between finger raise inclination and the peak impact loads acting on the ore pass walls is shown in Figure 8-10. The scatter lines in the same graph illustrates the observed variation in the peak impact load for each configuration.

180


no

OJ u

u ro CL

E CO V

CL

a)

120

Angle of Intersection (°)

130 140 150 160 170

60 70


180

♦ Ore Pass Inclination 90°

90

100

u re Q.

ro OJ Q.

16000 14000 12000 10000 8000 6000 4000 2000

0

b)

no Angle of Intersection (°)

120 130 140 150 160

20 30 40 50 60 70 Finger Raise Inclination (°)

80

170

I : . I T r i > ♦ Ore Pass

Inclination 80°

90

181


90 Angle of Intersection (°)

100 110 120 130 140 150 160

0J u

u ro Q_

E

ro

16000

14000

12000

10000

8000

6000

4000

2000

0

1

< •

r " i ' i • « ■

'

♦ Ore Pass Inclination 70°

c)

20 30 40 50 60 70


80 90

Figure 8-10. Influence of finger raise inclination on the peak impact load of particles on the ore pass wall inclined at a) 90° b) 80° and c) 70°.

The results of 330 simulations of material entering the ore pass through inclined finger raises were analyzed with reference to the peak impact force acting on the ore pass wall. The worst case scenario, which generates high impact loads and is, consequently, likely to result in a greater damage on the wall was for intersection angles of 140° and 145°. The maximum impact load was generated for the vertical ore pass at a 55° finger inclination (intersection angle of 145°). Maximum impact loads for the ore pass inclined at 80° was occurred when the finger inclination was 60° (intersection angle 140°) and for an ore pass inclined at 70° and a finger at 70° at an intersection angle of 140°. These results identify an interesting trend that, if it is validated by future investigations, could have significant implications on the design of raise fingers and ore pass systems. This is of particular concern as several ore passes use this configuration. This would explain the number of reported degradation problems that have been associated with ore pass-finger raise configuration. For example, the observation of the ore pass systems at Brunswick Mine indicates that more than 46% of the finger raises intersect their associated ore pass systems at an angle of intersection of 140°-150°. The results of different configurations have been provided in appendix-A.

8.6 Effect of particle impact on a deforniable ore pass wall

In the preceding numerical experiments, rock mass along the ore pass wall was simulated using a rigid wall in the PFC2D models. Although this is representative of good rock mass quality along the ore pass wall, ore passes in underground mines often intersect rock masses of varied quality, including weak or foliated rock masses. Two examples of

182


the ore passes constructed in bad ground conditions in the Brunswick Mine (OP. # 15, OP. # 1000SFR) were discussed in chapter 3.

In order to investigate the response of an ore pass wall to a rock fragment impact, another series of numerical experiments were designed using PFC2D. In these numerical experiments, the rock mass along an ore pass wall was simulated by a bonded particle model. This included the construction of a block of rock mass of 5 m x 5 m along an ore pass wall, opposite a finger raise junction, which was represented by an assembly of circular particles bonded together at their contact points, Figure 8-11. The model was composed of 55960 circular particles with a minimum diameter of 0.86 cm. The ratio of maximum versus minimum particle size was fixed at 1.66 in order to generate a uniform particle size distribution in the model. The micro-mechanical properties listed in Table 8-3 were assigned to the particles and bonds in the model. These micro-mechanical properties resulted in the macro-mechanical properties of the intact rock summarized in Table 8-4. In order to depict the damage caused by impact on the ore pass wall, micro mechanical properties were selected which would result in intact rock with a relatively low compressive strength, (UCS = 103 MPa), for the rock block along the ore pass wall.

Table 8-3. Micro-mechanical properties employed for generation of the 2D bonded particle model.

Particle micro-mechanical parameter Value Bond micro-mechanical parameter Value

Particle density (kg/m3) 2800 Parallel bond modulus (GPa) 53

Particle contact modulus (GPa) 53 Bond normal/shear stiffness 2

Particle normal/shear stiffness 2 Bond normal and shear strength (MPa) 75

Particle friction coefficient 0.4 Bond strength standard deviation (MPa) 20

Table 8-4. Macro-mechanical properties of the 2D bonded particle model.

Property Value

UCS (MPa) 103

Elastic Modulus (GPa) 65

Poisson Ratio 0.26

183


5 m

5 m

Figure 8-11. A 5m x 5 m bonded particle model represents a block of rock along an ore pass wall.

Once the 5 m x 5 m bonded particle model was generated, three ore pass inclinations (90°, 80° and 70°) were simulated in the model by removing the particles along the ore pass wall.

A projectile rock fragment of 0.8 m in diameter was generated to the right of the ore pass walls. The mechanical properties listed in Table 8-1 were assigned to the projectile particle. This particle was thrown against the ore pass wall at different angles of impact ranging from 30° to 80° in 10° increments. In all, 18 configurations were simulated. Figure 8-12 illustrates the model of an ore pass wall inclined at 80° in which a projectile particle has been thrown at 60° toward the ore pass wall. Based on the angle of impact, various impact velocities were chosen for the projectile, using the impact velocity analysis illustrated in Figure 8-5, and summarized in Table 8-5. Finally, impact induced stresses, impact induced damage and energy balance between the projectile rock fragment and the ore pass wall were measured for all configurations.

184


5m

0.8 m

80

5 m r

Figure 8-12. A projectile particle impacting the 80° inclined ore pass.

Table 8-5. Impact velocities employed for different numerical experiments.

Ore pass inclination (°)

Impact Angle (c) Ore pass inclination (°) 30 40 50 60 70 80

70 13.0 16.2 19.0 21.7 23.4 24.9

80 12.7 16.7 19.6 22.5 24.3 26.8

90 13.2 17.1 20.1 22.9 25.1 28.0

8.6.1 Impact induced stresses on the ore pass wall

The impact of a projectile particle hitting an ore pass wall results in induced stresses on the wall. The magnitudes of impact induced stresses on the ore pass walls were recorded using a measuring circle with a 0.56 m radius. The center of the measuring circle was placed at the impact point on the ore pass wall. This has resulted in the calculation of the magnitude of stresses in 0.5 m2 of the rock mass. Figure 8-13 presents the location of the measuring circles along the vertical, 80° and 70° inclined ore passes.

185


a) " r b)

c)

Figure 8-13. Impact of projectile particle with ore pass wall inclined at a) 90° b) 80° and c) 70°. The measuring circles are identified in black on each ore pass wall.

In PFC models, inter-particle contact forces and particle displacements are the only parameters which are computed. Stress, which is a continuum quantity, can be measured in the PFC models by averaging the contact forces developed in the area specified by a measurement circle. Figure 8-14a indicates the contact force distribution at a collision point for an ore pass wall inclined at 80° when the rock fragment impact is at an angle of 60°. Figure 8-14b represents a close-up view of Figure 8-14a, which displays the inter-particle force distribution caused by the impact of the projectile particle within the measurement circle. The thickness of the lines is proportional to the magnitude of the force.

For all configurations, the magnitude of horizontal (axx) and vertical (a^) stresses inflicted on the ore pass walls by the collision of the projectile particle were calculated

186


within the measuring circle zone. Figure 8-15 illustrates the relationship between the impact angle and the magnitude of horizontal and vertical stresses induced on the ore pass walls. The figure shows that for all configurations the horizontal stresses induced on the ore pass walls are greater than the vertical stresses. However, for an ore pass inclined at 70° the magnitude of vertical stresses is much higher than the vertical stresses recorded for ore passes inclined at 90° and 80°.

The highest magnitude of horizontal stresses for 90°, 80° and 70° ore passes were recorded when the impact angles were 50°, 60° and 70°. This result indicates that the maximum impact induced stress on the ore pass wall occurs when the angle of intersection is 140°. This is the same as the angle that was recorded in the analysis discussed in section 8.5.2.

5 m

a)

b)

Figure 8-14. a) Contact force distribution in the ore pass wall inclined at 80° due to the impact of a rock fragment, b) A close-up view of the model.

187


a)

b)

i i

in

25

2 I. 20

S is Cn

10 S

o i»

C)

SO 60

Impact Angle (°)

SO 60

Impact Angle ('■')

SO 60

Impact Angle (")

• 90Sig(XX)

90Sig(YY|

« 80 Slg(XX)

. 8asig(YY)

^ ^ - — ' * " ^ ^ V

• 70 Slg(XX|

. 70 Sig(YY)

KM 90

Figure 8-15. The relationship between impact angles and the horizontal and vertical stresses induced on the ore pass walls inclined at a) 90°, b) 80° and c) 70°.

The magnitude of stresses at the impact region on the ore pass wall is high and decreases with increase of distance from the collision point. In an attempt to measure the magnitude of stresses around the impact point, four measuring circles were placed concentric with

188


the initial measuring circle on the wall of ore pass inclined at 80°. The radiuses of the additional measuring circles were considered to be two to five times that of the initial circle, Figure 8-16. The magnitudes of horizontal and vertical stresses at different distances from the impact point are illustrated in Figure 8-17. The effect of impact induced stresses become negligible beyond 3 m of distance from the impact point.

Figure 8-16. Measuring circles of different radii used to compute the impact induced stresses around the impact region on the ore pass wall.

♦ Sig(XX)

ASig(YY)

0.5 1 1.5 2

Distance from the impact point (m)

2 5

Figure 8-17. Impact induced stresses at different distances from the impact point.

189


8.6.2 Impact induced damage on the ore pass wall

Collision of the projectile particle with the ore pass walls creates damage in the form of micro cracks. There are a number of definitions of damage, which are discussed in detail by Kachanov (1986). In this work, damage is defined as the appearance and growth of cracks induced by particle impact loading. This is one of the definitions used by Kachanov (1986). The shear and tension cracks are generated on the ore pass walls and inside of the rock mass. The density of micro crack growth in the vicinity of the impact point is high. Around the impact region, a cluster of micro cracks are generated, while at a distance from this region discrete micro cracks are initiated. The majority of the inter-particle bonds fail in tension. Figure 8-18 demonstrates the damage inflicted by the projectile particle on the ore pass walls with inclinations of 90°, 80° and 70°. A close-up view of the damage zone for each model is presented to the right of the model. In Figure 8-18 the tension cracks are indicated in blue, while the shear cracks are identified in red.

The number of micro-cracks developed on the ore pass wall depends not only on the inclination angle of the ore pass wall but also on the impact angle of the projectile particle. The number of impact induced micro-cracks on the 70° inclined ore pass is significantly greater than that of the vertical and 80° inclined ore passes. This is due to the greater vertical stresses which were recorded for the 70° inclined ore passes.

Figure 8-19 illustrates the relationship between the impact angle of the projectile particle and the number of micro-cracks that are produced on the ore pass walls with inclinations of 90°, 80° and 70°. The maximum number of impact induced micro-cracks on the 90°, 80° and 70° inclined ore passes were recorded when the impact angle was 50°, 60° and 70° respectively. These results further support the conclusion that the intersection angle of 140° is the most critical angle for impact induced damage on the ore pass wall.

190


b)

/

m p i ? '

' I V

C) • v

Figure 8-18. Impact induced damage on the ore pass wall inclined at a) 90°, b) 80° and c) 70°, with a close-up view of the damage zones.

191


a)

20 30 40 50 60

Impact Angle (°)

70

90

80 90

b)

20 30 40 50 60

Impact Angle (°)

70

■ 80

80 90

20

o

30 40 50 60

Impact Angle (°)

70

♦ 70

80 90

Figure 8-19. Influence of impact angle on the number of micro-cracks initiate on the ore pass wall inclined at a) 90°, b) 80° and c) 70°.

192


8.6.3 Balance and dispatching of energy in an ore pass

The impact of a single particle or of multiple particles on an ore pass wall can be evaluated using an energy approach. The potential energy of particles tunneled into a finger raise is transformed into kinetic energy during the flow stage. Once the particles impact the ore pass wall, they lose part of their kinetic energy. Thus, there is a balance between the magnitude of kinetic energy that the projectile particles lose during the impact and the magnitude of energy that is absorbed by the ore pass wall. Nazeri (2001) employed an analytical method to take into account the energy transferred by the impact of a single particle on an ore pass gate. It was assumed that all of the kinetic energy of a particle is absorbed by the chute gate. It should be noted that this assumption is true if the hitting particle does not break.

Based on the law of conservation of energy, an energy balance can be written for a single particle sliding along a finger raise and hitting an ore pass wall, Figure 8-20. The first energy balance is written for the particle moving from point A to the point B, a moment before hitting the ore pass wall. If the air resistance force is ignored and it is assumed that the initial velocity of the particle at point A is equal to zero (V0 = 0), the balance of energy for the particle the instant before impact with the ore pass wall is as follows:

B I \ V,

S Vo_ A

H,

I \ V,

>

H;

Figure 8-20. Impact of a single particle.

KEA + PEA + Wext = KEB + PEB (8 - 6)

0 + PEA + Wfric = KEB + PEB (8 - 7)

mg(H1 + H2) - pmgcos(a)L = KEB + mgH2 (8 - 8)

193


m^(//1 - pcos(a)L) = KEB (8 - 9)

Where: KE is the kinetic energy (Joules), PE is the potential energy (Joules), Wext is the work done by external force (Joules), Wfnc is the friction work along the finger raise (Joules), m is the particle mass (kg), p is the friction coefficient, Vi is the impact velocity of particle (m/s), Vr is the rebound velocity of particle (m/s), g is the gravitational acceleration of the earth (9.8 m/s2), Hi is the elevation difference between point A and B (m) and H2 is the elevation difference between point B and the bottom of the ore pass.

The equation 8-9 implies that all the potential energy of the particle in point A is either transformed into kinetic energy at point B or dissipated by the frictional work along the finger raise.

When the particle hits the ore pass wall at point B, it loses part of its kinetic energy. The majority of the dissipated energy is absorbed by the ore pass wall, while a small part of it is converted into other forms of energy, such as fragmentation of the projectile particle, heat and sound. The amount of energy which is absorbed by the ore pass wall depends on wall's properties and conditions of impact. Based on the diagram in Figure 8-20, the energy balance for the particle before and after hitting the ore pass wall can be expressed as follows:

KEbefore + PEbefore + Wext = KEaf ter + PEa f t e r (8 - 10)

By assuming that the potential energy of the particle a moment before and after impact remains constant, equation 8-10 can be summarized as:

ImVf + Wex^^mVr2 (8-11)

1 Energy loss = -m(Vr

2 - V?) (8 - 12)

Although it is possible, for a single particle impact, to calculate analytically the balance of energy and to determine the kinetic energy of a particle in the target point, for multi particle impact, this balance is difficult to calculate. The difficulty of this calculation is due to the collision, friction and breakage that occurs when particles come in contact with each other while flowing through the finger raise. Figure 8-21 shows a diagram of multi particle impact in an ore pass system. It is difficult to estimate the velocity of particles at point B, as the particles collide with each other and with the finger raise wall when they move in the finger raise. The distinct element methods can be used to determine the velocity and kinetic energy of multi particles at the point of impact. This was shown in section 8.5.1, where the impact velocity and the kinetic energy of a batch of circular particles were measured for different ore pass and finger raise configurations.

194


B

Vi

V,

Figure 8-21. Impact of multiple particles in an ore pass system.

For the numerical models generated in section 8.6.1, the energy balance was investigated for the projectile particle as it hit the ore pass wall. In these models, the kinetic energy of the projectile particle was monitored for all the configurations discussed in section 8.6.1. The results of the energy balance for different ore pass configurations are summarized in Table 8-6. In this table, the dissipated energy represents the difference in the kinetic energy of the projectile particle before and after impact (equation 8-12). The dissipated energy is stored in the synthetic rock block as strain energy either in the contact points between particles or in the bonds that exist between the particles. The dissipated energy can also result in the sliding of the particles against each other at their contact point. This can be traced as friction energy. Next, in order to calculate the amount of energy absorbed by the rock mass along the ore pass wall, the strain energy stored at contact points between particles (particle strain energy) and in the bonds (bond strain energy) was measured. Finally, the total energy dissipated by frictional sliding (friction energy) at the contact points between particles was calculated. Subsequently, the energy balance was obtained by the subtraction of the dissipated energy from the sum of the absorbed energies. The results of energy balance presented in Table 8-6 show the values greater or smaller than zero. This can be justified by the fact that the calculations were not uniformly precise in the calculation of the energy values in the PFC models.

The result of the energy dissipated by the projectile particle on ore pass walls inclined at 90°, 80° and 70° indicates that the maximum energy dissipation happens when the impact angle is about 50°, 60° and 70° respectively. Energy dissipation increases as the ore pass inclination decreases. The majority of the dissipated energy (more than 80%) is stored as strain energy at contact points between the particles.

195


Table 8-6. The results of energy balance in different ore pass configurations.

OP. Inclination

Particle Impact Angle

Dissipated Energy (DE) Absorbed Energy (AE) Energy

Balance (DE-AE)

(J)

OP. Inclination

Particle Impact Angle

AKE(J) SE1 (J) BSE2

(J) FE3(J)

Energy Balance (DE-AE)

(J)

90

30 1.28e5 1.07e5 0.14e5 0.066e5 0.004e5

90

40 1.59e5 1.31e5 0.19e5 0.062e5 0.028e5

90 50 1.75e5 1.43e5 0.22e5 0.067e5 0.033e5

90 60 1.29e5 1.05e5 0.16e5 0.059e5 0.02 le5

90

70 0.75e5 0.62e5 0.077e5 0.038e5 0.015e5

90

80 0.24e5 0.2e5 0.02e5 0.025e5 -0.005e5

80

30 1.35e5 1.2e5 0.1 e5 0.038e5 0.012e5

80

40 1.92e5 1.7e5 0.16e5 0.047e5 0.013e5

80 50 1.95e5 1.67e5 0.19e4 0.058e5 0.032e5

80 60 2.1e5 1.8e5 0.2e5 0.065e5 0.035e5

80

70 1.4e5 1.25e5 0.12e5 0.029e5 0.001e5

80

80 0.92e5 0.84e5 0.063e5 0.014e5 0.003e5

70

30 1.62e5 1.4e5 0.15e5 0.05e5 0.02e5

70

40 2.18e5 1.8e5 0.23e5 0.14e5 0.01e5

70 50 2.62e5 1.84e5 0.25e5 0.52e5 0.01e5

70 60 3.07e5 1.9e5 0.3e5 0.84e5 0.03e5

70

70 3.18e5 2e5 0.27e5 0.86e5 0.05e5

70

80 1.87e5 0.99e5 0.1e5 0.77e5 0.01e5

Strain Energy

Bond Strain Energy

Friction Energy

196


In the numerical models, the projectile particle that was thrown against the ore pass wall had a linear trajectory, however, in reality a particle which exits a finger raise follows a curve-linear trajectory as it hits the ore pass wall. The curve-linear movement of a projectile particle can result in slightly greater impact angle and consequently can influence the amount of energy dissipated. The curvature of the particle trajectory depends on its velocity and mass, as well as the diameter of the ore pass.

8.7 Impact induced damage on a foliated rock mass along ore pass wall

The presence of geological structures along ore pass walls may result in collapse. The importance of driving the ore passes against the foliation or strata dip has been recognized by Stacey & Swart (1997). They have recommended that ore passes be oriented so as to intersect the strata as close to perpendicular as possible. Based on a database of South African mines, Joughin & Stacey (2005) demonstrated a relationship between the extent of ore pass degradation and the angle between the ore pass and strata for various qualities of rock mass. According to Joughin & Stacey (2005), when the angle between the ore pass and rock mass strata approaches 90°, in moderate to good quality rock mass, the ore passes become more stable. Based on a data base developed for ore pass systems in Quebec underground mines, Hadjigeorgiou et al. (2005) confirmed this observation. Figure 8-22 shows two ore passes in Quebec, which were constructed in the same structural regime characterized by steep bedding. It can be seen that in case (a), in which the ore pass has been constructed against the bedding planes, the ore pass is stable and there is no sign of degradation. In case (b), where the ore pass has been driven sub-parallel to the bedding planes the ore pass has been expanded in varying degrees along its length.

197



The state of ore pass systems at the Brunswick Mine was reviewed in chapter 3 and in appendix A. The influence of structure in the degradation of the #15 and # 1000SFR ore passes was investigated. The upper section of the 1000SFR ore pass, which was constructed sub-parallel to the host rock foliation (Quartz Augen Schist), has been significantly expanded along the upper section, Figure 8-23a. The lower part of the ore pass was developed in a favorable orientation that allowed the ore pass to maintain its integrity, Figure 8-23b. The two upper sections of the # 15 ore pass were constructed sub-parallel with the dip angle of a laminated metasediment rock mass. In addition, ten finger raises dump material into this ore pass. Most of these dumped materials directly impact the footwall of the ore pass. The interaction between unfavorable rock structures with material impact in this ore pass caused considerable degradation of the #15 ore pass, in particular along the footwall side.

198


upper section

Foliation

(QAS) Lower section

Figure 8-23. The upper and lower sections of the 1000 SFR ore pass with respect to the rock mass foliation.

The influence of structural defects on the stability of ore pass systems can be simulated using boundary element models in conjunction with a ubiquitous joint approach. However, what distinguishes ore or waste passes from other vertical and horizontal excavations is that they involve material transport. The flow of material within an ore pass constructed in a foliated rock mass can compromise the structural integrity of the ore pass, resulting in more pronounced degradation. This cannot be modeled in boundary element models.

Three foliated rock mass models with foliation angles of 60°, 90° and 120° were generated using the bonded particle model constructed in section 8.6. The normal spacing between the foliation planes was fixed at 0.2 m. The models are presented in Figure 8-24. To simulate the mechanical properties of foliation planes, the smooth joint function was employed. Parallel bonds along the foliation planes were deleted and both the normal and shear stiffness of particles located along the foliation planes were modified to 1.3 x 1013

N/m3. A coefficient of friction of 0.3 was assigned to the particles adjacent to the foliation planes. The applied micro-mechanical properties have resulted in macro-properties of zero cohesion and an angle of friction of 30° for the foliation planes.

199


5 m

b)

*

Y

<

-I 2 rr

*

<

-I 2 rr

<

-I 2 rr

5 m

O

5 m

Figure 8-24. Foliated rock mass models with foliation angles of a) 60°, b) 90° and c) 120°.

200


Once the foliated rock mass blocks were generated, the desired ore pass inclinations could be simulated by deleting the particles along the ore pass wall. In order to investigate the influence of foliation planes on the impact induced damage of a projectile particle, an ore pass inclination of 80° was simulated in the three foliated rock mass models. A projectile particle with the same properties discussed in section 8.6 was thrown against the foliated ore pass walls at an impact angle of 60° and an impact velocity of 22.5 m/s. Damage inflicted by the rock fragment on the foliated rock masses are presented in Figure 8-25, where they are compared with the damage inflicted on a solid rock block model with a similar configuration, discussed and presented in section 8.6.1. A close-up view of the damaged zone for each model has been presented to the right of the models.

The ore pass wall foliation had a significant influence on the size of the damage zone. The presence of foliation planes in the rock mass along the ore pass wall increased the extent of impact induced damage. The number of micro-cracks developed in the foliated rock models due to the rock fragment impact was raised significantly. This implies that the foliations acted as weakness planes which accelerated the impact-induced crack formation. Furthermore, the interaction of impact induced micro-cracks with the foliation planes in the rock mass created small rock slabs which were detached from the rock mass and fall or slide to the ore pass.

201


a)

c)

m

\

-

/

V ' t ie V-V ■ <

v^^ \ \

v

! U,

i

Figure 8-25. Impact induced damage on the ore pass walls inclined at 80° and comprising, a) a solid rock mass, b) a 60° foliated rock mass, c) a 90° foliated rock mass and d) a 120° foliated rock

mass.

202


The observation of the patterns of impact induced damage zones in various foliated rock mass models indicates that the presence of foliation planes results in wider and deeper damage zones. The damage pattern generally forms as a layer of heavily crushed area around the impact point from which several radial (wing) cracks extend. The radial cracks generally develop perpendicular to the foliation planes. The pattern of impact induced damage zones on the ore pass wall which contained the 60° foliation planes and was inclined at angles of 90°, 80° and 70° are presented in Figure 8-26. A zoom view of the damage zone has been presented to the right of each model.

v\\\

Figure 8-26. The patterns of impact induced damage on the ore pass wall inclined at a) 90° b) 80° and c) 70° which contains foliation planes of 60°.

203


The effect of the intersection angle between ore pass wall and rock mass foliation was further investigated by repeating the numerical experience demonstrated in Figure 8-25. The projectile particle was thrown three times against each foliated ore pass wall. After each impact, the particles along the ore pass wall which were fully detached from the foliated rock mass were deleted. Figure 8-27 displays the evolution of impact induced damage on ore pass walls of varying foliation angles. The zoom views of the models which were presented in figure 8-27 are demonstrated in Figure 8-28. Both figures clearly show that the extent of damage to the ore pass wall with a foliation of 60° is greater than of the ore pass wall with foliation planes of 120°. In the first configuration, the ore pass wall is semi-parallel to the foliation planes, while in the last configuration, the ore pass intersected the foliation planes at an oblique angle.

b)

» "

1 i

i

X

\

i

Figure 8-27. Evolution of an impact induced damage zone in an ore pass wall with an inclination of 80° and containing foliation planes of a) 60°, b) 90° and c) 120°.

204


S Y

Y

b) i

S \

\ \

! ( r

> • &

-' x Y „

i I '

v \ \ s

! ii ; ■ *

)

' II

/

Figure 8-28. A zoom view of the evolution of a damage zone along the ore pass wall with an inclination of 80° and containing foliation planes of a) 60°, b) 90° and c) 120°.

The impact of rock fragments on foliated rock masses can also create rock wedges along the ore pass walls. The wedges can slide or collapse into the ore pass if the foliation angle is favorable with respect to the ore pass wall. This type of collapse was simulated in a model with foliation planes of 60°, in which a large rock wedge that was created on the

205


ore pass wall slide into the ore pass and caused an expansion of the degradation zone, Figure 8-29.

b)

Figure 8-29. a) Impact induced damage on the ore pass wall with foliation angle of 60°, b) The same model after sliding of a rock wedge into the ore pass.

8.8 Conclusions Two different strategies were used to investigate ore pass degradation by material impact. Both strategies used the available data from field to construct a series of parametric investigations using the particle flow code. In the first approach, several ore pass and finger raise configurations were simulated. In the numerical investigation, material properties were kept constant during 330 runs, while the ore pass and finger inclination was allowed to vary. The influence of different configurations on the impact that occurs on the ore pass wall has been quantified and presented in the form of charts.

It has been demonstrated that particle impact velocity and kinetic energy increases with an increase in finger raise inclination. The impact duration decreases with an increase in finger inclination. These observations can be used to evaluate different options of finger inclination for any particular ore pass inclination. In order to compare the influence of both ore pass and finger inclination, it is necessary to account for the resulting intersection angle. This consideration does not appear to have been taken into account in current design practice, where empirical rules ignore the ore pass-finger raise intersection. The results of the analysis that was undertaken clearly demonstrate, however, that the choice of intersection angle can have a significant influence on the resulting impact loads on the ore pass wall and the location and magnitude of damage to the ore pass. The highest impact loads were reported for intersection angles of 140° and 145°. These results can explain some of the observed ore pass degradation problems at Brunswick Mine, where more than 46% of the finger raises intersect the ore pass systems at angles of intersection of 140°-150°.

206


In this numerical approach, the ore pass walls were simulated as rigid components representing good rock mass quality. To estimate the impact induced damage along the ore pass walls, the second numerical approach was designed. In this approach the ore pass wall was considered to be deformable. For this purpose a block of rock mass along an ore pass wall was simulated using the bonded particle model. It was assumed that the block was located opposite a junction point between a finger raise and the ore pass. Three ore pass inclinations of 70°, 80° and 90° were simulated in the rock block model. A single projectile particle was then thrown against the ore pass walls at different impact angles.

The results of particle impact on the ore pass walls demonstrated that the magnitude of impact induced stresses, and consequently the impact induced damage, on an ore pass wall with an inclination of 70° is greater than that of the other ore pass inclinations. The maximum impact induced stresses and impact induced damage on the ore pass walls with inclinations of 90°, 80° and 70° was recorded when the impact angles were 50°, 60° and 70° respectively. This observation reconfirmed that 140° is the critical intersection angle between a finger raise and an ore pass.

The following recommendations, Table 8-7, can be made for design of ore pass and finger raise configurations based on the results of the numerical experiments obtained. It should be noted that other factors such as the free flow of rock fragments in the fingers and the stability condition of the pillar between ore pass and finger raise should be also taken into consideration. Best practice would suggest that there would be minimum problems during the operation of the ore pass. Acceptable implies that the ore pass will be subjected to significant loads which may result to some problems. It is recognized that these configurations may be dictated by other considerations such as uninterrupted material flow, operational constrains, access, etc. Problematic configurations will result in situation where there will be problems due to the impact. These configurations should be avoided if possible. Should a mine decide to go forward with this problematic ore pass-finger raise configurations, measure should be taken to consider rehabilitation or the use of liner.

Table 8-7. Recommendations for the design of different ore pass and finger raise configurations.

Ore Pass Inclination (°) Finger Raise Inclinations (°)

Ore Pass Inclination (°) Best Acceptable Problematic

90 75,80 30, 35, 40,45, 60, 65, 70 50,55

80 30,80 35,40, 45, 50, 65, 70, 75 55,60

70 30,35 40, 45, 50, 55, 60, 65, 80 70,75

Finally, the effect of foliation planes on the degradation of ore pass walls subjected to impact loads was investigated. The results demonstrated that the extent of a damage zone

207


in an ore pass wall constructed sub-parallel to the rock mass foliation planes is greater than the damage to an ore pass wall that intersects the foliation planes at an oblique angle. This conclusion validates the observation first suggested by Stacey & Swart (1997) and subsequently corroborated by Hadjigeorgiou et al. (2005) in Quebec underground mines.

208

Chapter 9: Conclusions and Future Work

9 Conclusions and Future Work

9.1 Introduction

This final chapter serves as a summary and conclusion, and provides recommendations for further work. An overview of the work undertaken within the framework of the present thesis is presented. This is followed by a highlighting of the original aspects of the work and identification of contributions made to the analysis of ore pass degradation. In addition, the important conclusions drawn from the work are recapitulated. Finally, perceived limitations of the conclusions of this study are outlined and suggestions for the refinement of the present work in future research and several other general recommendations are provided.

9.2 Summary of the research work

In most underground mines, minerals and waste materials are handled by vertical or inclined excavations call ore passes. In these mines, production relies on the smooth and efficient operation of these ore pass systems. This implies an ore pass design to ensure material flow as required, and should remain stable for the duration of its operating life.

Observation of mining operations indicated that two types of problems are frequently encountered in ore pass systems. The first type is related to degradation of the structural integrity of the pass walls, while the second is linked to material hang-ups. The ore pass degradation is caused by induced stresses imposed on the ore pass wall and by structural defects in the rock mass. The movement of materials through the ore pass also causes degradation in the form of impact-induced damage or wear caused by wall abrasion. In addition, hang-ups can interrupt the operation of the ore pass, which can decrease mine production. Blasting for hang-up release can further impair the integrity of the ore pass walls. These problems are an important economic consideration, as they have the potential to interrupt the mine's operation.

This thesis focused solely on the problem of ore pass degradation. To address the problem of ore pass degradation, the present work has developed a methodology of ore pass stability analysis. This analysis not only serves as a means of understanding the ore pass degradation problem, but would also be a useful reference for engineers who design ore passes or for consultants confronted with issues of ore pass stability.

209


The principal objective of this work has been to get better understanding of ore pass degradation problems. In a broad sense the main objective was divided into three specific objectives: 1) investigate ore pass stability problems at Brunswick Mine, 2) develop a numerical method that can quantitatively evaluate the interaction of different ore pass failure mechanisms, 3) quantification of failure initiation due to material impact on the ore pass.

The objectives were achieved by carrying out the following tasks:

• An ore pass observation and data collection campaign was undertaken at the Brunswick Mine. This was followed by statistical analysis of the ore pass design parameters. This process allowed for better understanding and identification of the relationship between the design parameters and the ore pass performance. Consequently, a detailed investigation of degradation problem in three ore pass case studies in the mine demonstrated that instability of the ore passes is the result of the simultaneous interaction of stress, structure and material flow impact.

•

•

A numerical approach was proposed for stability analysis of ore pass systems. This approach was based on a hybrid numerical model called the synthetic rock mass. The #19A ore pass located in a massive sulphide rock mass, in a lower block of the Brunswick Mine was selected as a case study for the proposed approach. The #19A ore pass is a part of 18-21 ore pass complex located in the 20 & 21 mining zones in a high stress region. The following steps were undertaken in the analysis:

Scanline mapping was undertaken in the mine in order to collect fracture data from the massive sulphide rock mass around the #19A ore pass. A statistical analysis of the fracture data collected from the region studied served to identify the structural characteristics of the rock mass. Based on the characteristics of the fracture sets derived from the field data, a fracture system was generated for the site at Brunswick Mine, using the Fracture-SG code. The fracture system generated was validated through the use of a calibration process that compared the distribution of fracture sets characteristics obtained from the model with those obtained from the field data.

A synthetic rock mass that represented the massive sulphide rock mass was then constructed. The intact rock properties of the massive sulphide rock were assigned to a bonded particle model through an inverse calibration process.

Mechanical and structural properties of the generated synthetic rock mass were determined by sampling the synthetic rock mass in varying size ranging from (0.05 m x 0.05 m x 0.1 m) to (10 m x 10 m x 20 m). The strength and elastic modulus of the synthetic rock mass samples were determined. This allowed the identification of the REV size for the synthetic rock mass model based on the measured structural and mechanical properties.

210


Variation of the induced stresses around the #19A ore pass resulting from mining sequences undertaken between 2002 and 2007 in the zones 20 & 21 were simulated by a linear elastic Map3D model. This analysis provided necessary boundary conditions for the synthetic rock mass. The dimensions of the #19A ore pass were simulated in the synthetic rock mass by deleting particles in the center of the model.

Finally, the Particle Flow Code was used to assess the response of ore pass walls to material impact. Two numerical approaches were employed for this analysis. In the first approach, different ore pass and finger raise configurations were simulated using the rigid wall tool in the PFC2D. The impact load on the ore pass wall was measured for a batch of rock fragments. In the second approach, a portion of an ore pass wall opposite a junction point between a finger raise and the ore pass was simulated through a bonded particle model. The damage inflicted by rock fragments impacting on the ore pass wall was qualified for different ore pass configurations. Ultimately, the influence of foliation planes on the degradation of ore pass walls subjected to impact loads was simulated.

9.3 Conclusions

The major contribution of this thesis is the development of a comprehensive approach to the analysis of the stability of ore pass systems with reference to the #19A ore pass at Brunswick Mine which is situated in an area of high stress and recorded seismic activity. The results of this case study presented in chapter 7 led to the following conclusions:

• The proposed approach was appropriately used to interpret the influence of stress and structure in the stability analysis of the ore pass systems. Modeling the fracture geometries of the massive sulphide rock mass with a fracture system provided a realistic means of representing the pre-existing fractures. The influence of the pre-existent fractures on deformation and failure of the rock mass around the #19A ore pass was determined by linking the fracture system with bonded particle model. The adopted approach modeled failure of rock mass around the excavation as a combination of the interaction of pre-existing fractures and the failure of the intact rock bridges between the pre-existent fractures.

• Although the proposed approach was used as a back analysis for the #19A ore pass case study, it can also be employed in the design of new ore passes. With this method, a variety of ore pass configurations can be simulated in a 3D synthetic rock mass and the damage zones associated with each configuration can be compared. The configuration which results in smaller extent of rock mass failure zone around the ore pass can then be selected.

• It was demonstrated that, for the 2D model of the #19A ore pass, failure zones developed in the North and South walls of the ore pass where stresses were

211


concentrated. The damage zones were extended deep into the rock mass, extending inward from the surface of the ore pass wall to a depth roughly equal to the radius of the ore pass (1.5 m), Figure 9-1. Moreover, the presence of pre-existing fractures has resulted in an asymmetric failure pattern around the ore pass. This pattern is different than the symmetric notch-shape failure generally observed around underground excavations develop in solid rock masses. This type of failure reflects the anisotropic and inhomogeneous distribution of the pre-existent fractures.

Figure 9-1. Extent and pattern of failure zone around the #19A ore pass using the 2D synthetic rock mass model.

• Tracing of failure mechanisms for the 2D synthetic rock mass model of the #19A ore pass demonstrated that, in stress concentration zones, failure is initiated by intact rock matrix fracturing (tension cracks) between pre-existing fractures, parallel to the ore pass wall boundaries. This type of failure results in the creation of small and large rock slabs around the excavation, which then slide or fall into the ore pass.

• Interpreting the results of the 3D synthetic rock mass model indicated that, in North and South walls (where induced stresses were concentrated), the extent of the damage varies according to its elevation within the ore pass, Figure 9-2. This variation indicates anisotropic and inhomogeneous distribution of the pre-existing fractures and their effect at different elevations. At some elevations, the damage penetrated the ore pass wall to a depth of more than 2 m. However, on the East and West sides of the ore pass, the only structural failures observed were in the form of rock wedge sliding and falling, Figure 9-3.

212


m JL' •

fllPRfl * - » •

*ar» • • 2^H#> ' 0

« < »»» « II','

J H ' i>%-« •' « f t . w

Élfr» * iSftfaOCW/Jty.

w\ * °#- o L*!»r

Figure 9-2. Extent of damage zone along the North and South wall sides of the #19 A ore pass using the 3D synthetic rock mass model.

Figure 9-3. Structural failure (wedge falling) along the East and West wall sides of the #19A ore pass using the 3D synthetic rock mass model.

The 2D model was used as a preliminary analysis to evaluate the interaction of stress and rock structural defects on the failure zone initiated around the ore pass wall. The 2D model is easier to construct and faster to run. In addition, the 2D model allows the tracing of the deformation process and failure mechanisms taking place around the ore pass wall. This is not possible in the 3D model due to the complexity of the 3D model and the lack of a proper visualization tool that permit a comprehensive observation and monitoring of the damage initiation and propagation while the model is running. On the other hand the 2D model is not fully representative of a fractured rock mass. It only presents a cross section of it.

213


The geometry of fractures is better represented in the 3D model. Moreover, the gravitational acceleration can be assigned to the particles in the 3D model while this is ignored in the 2D cross section model. Finally, the impact of flowing particles in the ore pass can be integrated in the 3D. Nevertheless, running of the 3D model is longer and interpretation of the results is more complicated.

The influence of material impact on the ore pass walls, represented by the 3D synthetic rock mass, was quantified and discussed in chapter 7. This model presented the interaction of all failure mechanisms on the integrity of the #19A ore pass. Based on the results of this analysis, as well as the observations of the #21 ore pass presented in chapter 3, Figure 9-4, the general conclusion can be made that material impact causes a greater degree of failure on the ore pass wall, where higher induced stresses have already compromised the integrity of this ore pass wall.

1125-4 sub

Dump Point

Stress Concentration Zones

•/-

1 125 SC. 4 5 J j n u t v*w

Figure 9-4. Interaction of stress, structure and particle impact on degradation of the # 21 ore pass.

Numerical simulations were undertaken in chapter 8, in order to evaluate the response of ore pass walls to material impact. Overall, the results of these simulations showed that the developed methodology is a suitable tool for quantifying the influence of different ore pass and finger raise configurations on the impact loading of ore pass walls. Based upon this analysis, it was concluded that:

214


• The magnitude of particle impact load on the ore pass walls depends on the angle of intersection between the ore pass and its associated finger raise. The highest impact load and, consequently, greatest damage were reported for intersection angles of 140° and 145°. This could be taken into consideration in the design of ore pass and finger raises in underground mines.

• Although it has already been suggested by Stacey & Swart (1997) and by Hadjigeorgiou et al. (2005) to that an ore pass section should be driven so as to intersect the foliation planes of a rock mass perpendicularly, the current work corroborated their recommendation by simulating and quantitatively demonstrating this ore pass design fact.

The findings and discussions derived from synthetic rock mass characterization, presented in chapter 6, clearly highlighted the potential contribution of the present research to the understanding and modeling of rock mass behavior. The results of this chapter provide interesting conclusions. These include:

• The approach adopted does not rely on a user specified constitutive model assigned to intact rock and fractures. The properties of the generated synthetic rock mass were the result of the combination of the behavior of the solid rock matrix (represented by a bonded particle model) and the integrated fracture fabric (smooth-joints).

• Although the influence of scale on the strength and elastic modulus of rock mass has been recognized for a long time, it has been extremely difficult to quantify. Previous work has often relied on extrapolation, as it has been based on limited or incomplete data. The synthetic rock mass was successfully used to determine the effect of sampling size on the uniaxial compressive strength and the elastic modulus of the massive sulphide rock mass.

• The Representative Elemental Volume (REV) size of the selected rock mass properties was determined using statistical tests. For this particular site at Brunswick Mine, the structural REV size was 3.5 m x 3.5 m x 7 m and the mechanical REV size was 7 ra x 7 ra x 14 m. It was interesting to show that the REV size of a rock mass for two different properties is different. Consequently, the largest REV constitutes the global REV beyond which both the mechanical and structural properties of the rock mass are reasonably consistent with repeated testing. In chapter 7 a synthetic rock mass, larger than the REV size of the massive sulphide rock mass was generated to evaluate the stability analysis of the #19A ore pass.

As part of this research a database of ore pass design and performance was developed for the Brunswick Mine. This included identifying 25 ore pass systems with a total length of 7200 m which were classified in 98 ore pass sections. The database was used to assess the performance of ore pass systems in the mine and to determine the success or failure of various ore pass designs. Finally, applied guidelines for the design of ore pass systems at

215


the Brunswick Mine were developed. The proposed guidelines are general and can be used in all underground metal mines for design of ore pass systems.

9.4 Limitation of the employed methodology

Some of the limitations of the employed methodology for stability analysis of ore pass systems were well known. These limitations are summarized as follow:

• Comparison of the degradation observed by the CMS monitoring, in the #19A ore pass with the results of the synthetic rock mass models indicated that the numerical models could not accurately predict the dimensions of the degraded ore pass observed in the field. The numerical models only simulate the failure initiation around the ore pass and not propagation.

• In the present work, only circular and spherical particles were used in the investigation of impact-induced damage on ore pass walls. In reality, the blasting of a rock mass produces fragments of various shapes. Movement of rock fragments in an ore pass or a finger raise can be significantly influenced by the shape of the fragments. This can consequently affect the impact velocity of rock fragments on the ore pass wall.

• The development of the fracture system model for the massive sulphide rock mass at Brunswick Mine highlighted some of the limitations of field mapping with standard mapping techniques such as scanlines. This includes limitations in the number of fractures that can be mapped on the walls of the accessible drifts, and the limitations in the accuracy of mapping data due to the presence of support systems and other magnetic sources.

• Although the generation of 2D and 3D synthetic rock mass model with PFC-V4 is much faster and a very big model occupied by several hundred thousand particles can be generated in less than 24 hours, running the models for stability analysis of the underground excavations, as performed in chapter 7 and appendix-E required long execution time. The 3D models presented in chapter 7 and appendix-E were run in approximately 514 hours and 375 hours, respectively. The numerical models were run on a computer using a Pentium IV processor with a speed of 3.3 GHz.

9.5 Future work

The numerical approach employed for ore pass stability analysis needs further refinement and calibration in order to precisely predict the extent of damage zone around the ore pass. In this work, the intact rock properties on which the bonded particle models were constructed are: uniaxial compressive strength, elastic modulus and Poisson's ratio.

216


However, the result of tensile strength for the simulated model was much higher than that of the intact rock. Cho et al. (2007), working in PFC2D, suggested the use of clumped particles in order to generate a bonded particle model with irregular shape of clumped particles. This idea has yet to be successfully implemented in PFC3D. Other issues, such as the compatibility of this approach with synthetic rock mass and smooth-joint contact model, must also be reviewed. Another way to refine the bonded particle model would be to take into account the influence of stress corrosion, as suggested by Potyondy (2006). For this mean the bonded particle model must be further calibrated for long term mechanical properties using a static fatigue test. Moreover, the influence of material flow as a precursor to ore pass wall degradation should be further investigated.

The approach can also be used to model failure propagation from the removal of unstable/damaged wall zones by the flow of material in ore passes. This can be linked to the expansion of the ore pass size that is frequently observed in mines.

Ore passes can be operated as flow through or kept full. Keeping the muck inside an ore pass provide a confinement pressure to the rock mass around the ore pass which can have an important influence on the stability of the ore pass. In the case study investigated in chapter 7, the ore pass volume was deleted from the 3D synthetic rock mass model and the ore pass was kept empty to represent a flow through ore pass system. The confining effect of the filled materials on the stability of the ore pass walls can be simulated by blocking the bottom of the ore pass in the model and filling the ore pass with an assembly of discrete particles (not bonded together), representing ore material. The results of stress-induced damage for the ore pass model filled with the ore particles can be compared with the results of the empty ore pass model presented in chapter 7. This allows measurement the confining effect of the filled material on the stability of ore pass systems.

The analysis of rock mass damage inflicted by impact of a single rock fragment, presented in chapter 8, can be further extended to consider the impact of a stream of flowing particles on the rock mass. A parametric analysis can be followed to investigate the influence of different parameters on the response of the rock mass to the rock fragments impact. For the rock mass, the parameters that can be investigated are, the influence of rock mass deformation, strength, structural characteristics (number of fracture sets, orientation, persistence, etc.), and stress states. For the projectile particles, the parameters that can be evaluated are: the influence of particle size distribution, particle shape and density.

Recent developments in the application of digital photography and laser scanning methods for fracture characterization can overcome the limitations of the traditional fracture mapping methods. These techniques can be used to record the geometry of rock mass faces more quickly in three dimensions. The recorded data can be then converted into useful information that can be directly used for a fracture system generation.

The synthetic rock mass approach that was employed for the stability analysis of ore pass systems can be used for stability analysis of any other underground or surface

217


excavations in fractured rock masses. Moreover, this approach can be potentially used for estimation of pre-peak failure behavior and post-peak failure behavior of fractured rock masses.

218

Appendix A

Appendix A: Ore pass configurations at Brunswick Mine

This appendix illustrates the statistical analysis of ore pass and finger raise configurations at Brunswick Mine, discussed in chapter 3.

Square; 44 Eliptical; 0

Circular; 13

Rectangular; 41

Figure A-1. Shapes of ore pass cross sections at Brunswick Mine (98 sections).

219

Appendix A

50

40

c o Ë 30 <D l/>

Q E

20

10

H B G

■ V

# 4°

Section Minimum Dimension (m)

Figure A-2. Minimum dimension of ore pass sections at Brunswick Mine (98 sections).

c o 4» U

S o k

01

E

35

30

25

20

15

10

29

7 7 8

6 6 ^

■ ■ ■ I I <? / JP

♦ ^

# Trends of Sections(°)

Figure A-3. Trends of the ore pass sections at Brunswick Mine (98 sections).

220

Appendix A

60 -,

Number of Fingers in Section

Figure A-4. Numbers of finger raises per each ore pass section at Brunswick Mine (98 sections).

Finger raise cause side wall Finger raise cause footwall impact impact

Figure A-5. Orientation of finger raises with respect to the orientation of ore pass sections at Brunswick Mine.

221

Appendix A

60

50

40

<u 00 •E 30 LL.

| 20

10

51

<170-180 <160-170 <150-160 <140-150 <130-140 <120-130 <110-120

Intersection angle between Finger and OrePass (°)

Figure A-6. Intersection angle between ore pass and finger raise at Brunswick Mine.

Figure A-7. The scalpers of 0.91 m x 1.4 m employed at Brunswick Mine.

222

Appendix A

Unknown; 69

Significant; 8

Abandoned; 5 Notsignificant; 16

Figure A-8. Status of degradation in ore passes of Brunswick Mine database, (98 sections)

Material Name

Material tt

Material Type

2005 02 22

P |MohiCoiJomb ~B

Peak Residual 8

130

21 52

Tension Cutoff

UCS

Cohesion

Friction Angle

DilabonAnoje (ô

Young's Modulus 156000

Poisson"» Rat» |o 290

Viscous Mod |Gn) [Ô

~3W ~B\* HI»"

~3 ~3 "3

~3I~ ~B "3|60000 ^ "311°250 ~~]

"31 ff Elasbconly

■71 r ElastoPlasitc d f * Inactive

"3 Viscous Mod (Gs)|0

Expansion Cœ) fï

Conductivity H

? j Copy horn matl j Stress State | Close '

Inactive

UsetDehned Paiameters

o constant

o b constant

o constant

AOj variation

Ao. vanation

Ao t variation

o t iend

a , plunge

o t lu-mi

t /h constant

At/h vdilation

21 isa d 1° d jo d jo d 10055000 d j 0044000 d j0 028000 z\ j 90 000000 d jo d jo d jo d jo d

7 Copy horn Cartesian II Pose |

Figure A-9. The input data for the Map3D model of the Brunswick Mine.

223

Appendix B

Appendix B: Statistical Analysis of Fracture Data

B.l Fracture orientation

The orientation of fractures in rock mass is not random. A set of fractures is comprised of several fractures of same or similar orientations (dip and dip direction). The dip and dip direction of a fracture can be represented by a single pole on a projection stereonet. Sets of fractures are identified based on the clusters of their poles. Once the fractures are grouped into their corresponding sets, the mean values for dip angle and dip direction and their probabilistic density functions can be determined.

A Fisher univariate distribution is commonly used for modeling the distribution of three dimensional orientation vectors, such as the distribution of fracture set orientations (pole vectors) on a sphere, Mardia (1972). The Fisher distribution describes the angular distribution of orientations about the mean orientation vector, and is symmetric about the mean. A Fisher constant (or coefficient of dispersion) is a measure of the degree of the clustering or divergence of a fracture set around the mean orientation pole. The probability density function can be expressed as:

K s i n 6 e K c o s e

W = e K _ e - K (B - 1)

Where 0 is the angular deviation from the mean pole, in degrees, and K is the "Fisher constant" or dispersion factor. A bigger value of K indicates a more clustered fracture set, Priest (1993). The K value can be estimated using the following equation, Fisher (1953):

J V - 1

Where N is the number of poles, and R is the magnitude of the resultant vector for the fracture set in question.

Mardia (1972) proposed an approach to verify the Fisher distribution of a fracture set orientation. The approach is based on the evaluation of distribution of the angles which characterize the difference between the orientation of the poles of fractures for each set and the orientation of the mean pole of the same fracture set. The angle for each fracture can be measured graphically either by hemispherical projection using vector algebra methods or via the following expression, Priest (1993):

cos ô = cos(an — a s) cos Bn cos Rs + sin Bn sin Bs (B — 3)

Where as and ps are the trend and plunge of a fracture pole and ctn and pn are the trend and plunge of the mean pole.

224

Appendix B

Once the angles were measured for all the fractures in each set, a statistical goodness of fit test is performed to verify if the distribution of these values (angles) follows a normal distribution.

Several statistical goodness of fit tests are available to compare the distribution of a sampled data with a reference theoretical distribution or with distribution of another sample. The chi-square %2 and Kolmogorov-Smirnov are two popular goodness of fit test which are used to investigate the distribution of fracture characteristics. The main problem with chi-square test is the choice of number and size of the intervals. Although rules of thumb can help produce good results, there is no panacea for all kinds of applications. The Kolmogorov-Smirnov (K-S) test is designed to address this issue. The test is based on cumulative distribution function (CDF). The maximum distance between the curves of cumulative distribution functions for sampled data and the theoretical distribution is measured. If the distance is greater than a critical value, the hypothesis regarding the distributional form is rejected.

B.2 Fracture spacing and frequency

Fracture spacing is a measurement of the distance between two adjacent fractures along a line of specified location and orientation. The spacing of fractures can determine the size of individual blocks of intact rock that make up the rock mass. Measurements of fracture spacing can be done in the three forms of total spacing, set spacing and normal set spacing, Priest (1993). Total spacing is the distance between two immediately adjacent fractures, measured along a line of general, but specified, location and orientation. Set spacing is the distance between a pair of immediately adjacent fractures from a particular fracture set, measured along a line of any specified location and orientation. Finally, normal set spacing is the set spacing when measured along a line that is normal to the mean orientation of the set. The first two parameters can be directly measured from a scanline mapping. The normal set spacing can be determined using the following equation:

Xn = Xd cos ô (B - 4)

Where Xn and Xd are the normal set spacing and set spacing, respectively. The ô represents an acute angle between the sampling line and the normal of a fracture plane. The acute angle 8 can be determined using the equation B-3, where as and ps are the trend and plunge of the sampling line and an and pn are the trend and plunge of the normal to fracture plane intersected by the sampling line.

The most commonly adopted distribution function for the total fracture spacing is the negative exponential, Priest & Hudson (1976), Baecher et al. (1977), Kulatilake et al. (1993). For distribution of set spacing and normal set spacing Priest (1993) stated that it is yet to be determine whether they are exponentially distributed. The exponential distribution of total fracture spacing will not guaranty the same distribution for set

225

Appendix B

spacing and normal set spacing. Priest & Hudson (1976) demonstrated that the combination of non-exponential distributions of fracture set spacing can produce exponential distribution of total spacing.

Sen & Kazi (1984) and Park et al. (2005) suggested the log-normal distribution function for the total fracture spacing. Sen & Kazi (1984) reported that this form of distribution provides more flexibility as it considers both the mean and the standard deviation of fracture spacing.

In order to verify which theoretical probability distribution form better represents the fracture spacing values, the Kolmogorov-Smirnov test can be employed.

The fracture frequency is defined as the mean number of fractures per unit length (ID), per unit area (2D) or per unit volume (3D). The linear (ID) fracture frequency can be measured by scanline mapping. It can be applied to all fractures or to a specified fracture set. The fracture frequency can be obtained by inversing the value of fracture spacing. Using this approach, it is possible to evaluate the total, apparent and normal fracture frequency based on the total, set and normal set spacing values.

B.3 Fracture size Fracture size can lead the formation of rock wedges in a fractured rock mass and associated block size distributions that determine the deformability and stability of the block system.

The most widely used method of estimating fracture size is based on measurements of its trace lengths exposures in natural outcrops and underground or open cut excavation walls. Fracture trace is a line formed by the intersection of a planar fracture with a planar rock face. The end of a trace occurs either at another fracture or within the rock. The end may not be visible at a given rock face due to excavation, erosion or presence of vegetation.

The fracture trace lengths may have different distributional forms. The distribution of trace lengths of a fracture set can be modeled by lognormal distribution, Baecher et al. (1977), or by negative exponential distribution, Priest & Hudson (1981) and Kulatilake et al. (1993).

B.4 Sampling errors from biases

Several authors including Baecher & Lanney (1978), Priest & Hudson (1981), Kulatilake & Wu (1984) reported that the scanline mapping of rock fractures in the field introduces several biases into the sampled data. These include:

226

Appendix B

• Orientation bias: fractures striking parallel to the scanline are sampled in a lesser degree than fractures striking normal to the scanline.

• Truncation bias: Observations below or above a certain value are disregarded during field mapping.

• Censoring bias: fracture traces are shortened due to edge effects of the observation window (e.g. one or both ends of a fracture are concealed).

• Size bias: large fractures have a greater probability a) to intersect the rock surface and b) of being sampled than small fractures.

Carefully defined sampling or correction procedures are essential in order to eliminate or minimize these effects. The orientation bias can be corrected or at least reduced by sampling in different directions, Villaescusa & Brown (1992). The truncation bias can be reduced by decreasing the truncation level in fracture mapping or it can be corrected by applying an analytical method proposed by Warburton (1980).

There are several analytical approaches to account for the censoring bias. Mauldon et al. (2001) classified the approaches of sampling biases correction and estimate real mean trace length of fractures to two groups, 1) approaches which assumes a particular form of distribution for the trace length of the sampled data (e.g. Villaescusa & Brown (1992) assuming lognormal distribution and Priest & Hudson (1981) assuming negative exponential distribution), 2) methods that are distribution free (e.g. Zhang & Einstein (2000) using a circular window).

The approach reported by Villaescusa & Brown (1992) which originally due to Laslett (1982), is generally used to correct the censoring bias of fracture trace length data collected by a scanline mapping. In this method the ends status of each fracture intersected by the scan-line is observed. Based on the types of their ends, the intersected fractures are classified into three groups including (a) both ends of fracture trace visible, (b) one end of fracture trace censored and (c) both ends of the fracture trace censored. The following equation was proposed for estimation of unbiased mean trace length of sampled data.

ML = TT (B - 5) ^L 2n + m K J

Where:

PL is real mean trace length

• Xj, , Xn is observed trace lengths with both ends exposed

• Yi, , Ym is observed trace lengths with one end exposed

227

Appendix B

• Zi, , Zp is observed trace lengths with no ends exposed

Another analytical method for correction of censoring bias is based on a circular window mapping. This method was firstly developed by Mauldon (1998) and later revised by Zhang & Einstein (1998). In this method a finite circular window is considered on the rock mass exposure and the fracture traces enclosed inside the window or intersected by the boundary of the window are counted, Figure B-l. The unbiased mean trace length then can be expressed as, Zhang & Einstein (2000):

7r(JV + N 0 - JV2) pL = - ^ — J — ^ - XC (B - 6) 2 ( N - N 0 + N2)

Where C is the radius of sampling window; N is the total number of traces intersecting the sampling window; NQ is the numbers of fractures on the window with both ends censored and finally N2 is the numbers of fractures on the window with both ends observable.

Both ends are observable ( jV, ) Circular sampling window /

Traces

One end is censored ( iV, I Boih ends are censored ( JV„

Figure B-l. Three types of intersection between fractures and a circular mapping window, after Zhang & Einstein (2000).

228

Appendix C

Appendix C: Bonded Particle Model

This appendix provides a summary of the procedure for the bonded particle model of rock, described by Potyondy & Cundall (2004).

C.l Introduction

The Bonded Particle Model (BPM) proposed by Potyondy «fe Cundall (2004) is implemented in the two and three dimensional discontinuum programs PFC2D and PFC3D. The BPM represents an intact rock as a dense packing of non uniform sized circular or spherical rigid particles (grains) that are connected together at their contact points with parallel bonds (cements). The rigid particles interact only at the soft contacts, which have normal and shear stiffnesses. The mechanical behavior of this system is described by the movement of each particle and the force and moment acting at each contact. Newton's law of motion provides the fundamental relation between particle motion and the resultant forces and moments that cause that motion.

Application of a load to a bonded particle model is carried by the particles and bonds skeleton in the form of force chains which are propagated from one particle to the neighbor particles with which it is in contact. The force is also developed through the bond between particles at contact points. The bonded contacts experience compressive, tensile and shear loading, and can also transmit a bending moment between the particles, while the empty contacts (not bonded) experience only compressive and shear loading. In this way a heterogeneous force transmission is produced by the applied loading, which can induce, in several locations, tension/compression forces perpendicular to the direction of applied force. The force chains are not uniform. They may be much higher than the applied forces, causes a few particles to be highly loaded, while other particles carry smaller loads or no load at all because the forces arch around them. The concentration of these micro forces and micro moments in a localized area of a BPM model can produce bond breakage. This breakage can result in the redistribution of induced global force (because damaged material is softer and sheds load to stiffer, undamaged material) and, ultimately, can result in the formation of macroscopic fractures and/or rupture zones. With the increased breakage of bonds, the material behaves more like a granular material with highly unstable force chains, Potyondy & Cundall (2004).

The following assumptions are inherent in the BPM: the particles are circular or spherical rigid bodies with finite masses, the particles move independently of one another and can both translate and rotate, the particles interact only at contacts; because the particles are circular or spherical, a contact is comprised of exactly two particles, the particles are allowed to overlap one another, and all overlaps are small in relation to particle size such that contacts occur over a small region (i.e., at a point), bonds of finite stiffness can exist at contacts and these bonds carry load and can break (the particles at a bonded contact

229

Appendix C

need not overlap), generalized force-displacement laws at each contact relate relative particle motion to force and moment at the contact, Potyondy & Cundall (2004).

The rigid circular or spherical particles in a BPM cannot break. However, if an individual particle is represented as a cluster of bonded particles, the particle crushing can also be accommodated by this model."

C.2 Behavior of particles and bonds

The bonded particle model considers a grain to be a circular or spherical shape particle and the cement between two particles to be a parallel bond. The behavior of the particle-bond (grain-cement) system and the formulation of the BPM have been presented by Potyondy «fe Cundall (2004) and can be quoted as follow:

"The total force and moment acting at each bonded contact is comprised of a force, F, , arising from particle-particle overlap, and is denoted as the grain behavior in Figure C-l. In addition, a force and moment, F, and Mi , carried by the parallel bond and are denoted as the cement behavior, Figure C-l. These quantities contribute to the resultant force and moment acting on the two contacting particles." The force-displacement law for this system can be described for both grain and cement behavior.

r=icw deformability

AF"=k'AAU" AF '= -k 'AAU'

AM"=-k ' JAB ' AM'=-k a IAQ'

F ' <, \iF"

grain behavior

strength

-F* ô ° " = — +

A M'

< o<

\F'\ \M"\R x - = l _ l + L _ l _ < ? A J

cement behavior

Figure C-l. Force-displacement behavior of grain-cement system, after Potyondy & Cundall (2004).

230

Appendix C

At each contact point between two particles the micro properties of the two particles which are used to describe the force-displacement behavior are the normal and shear stiffnesses of the two particles (kn and ks), and their coefficient of frictions (p). These six parameters are called particle-based micro-properties, Figure C-l.

Parallel bonds establish an elastic interaction between particles that acts in parallel with the particle-based force-displacement behavior. Thus, the existence of a parallel bond does not prevent slip. Particles can only transmit force, while parallel bonds can transmit both force and moment between particles. A parallel bond can be envisioned as a set of elastic springs uniformly distributed over a rectangular cross section in PFC2D and a circular cross-section in PFC3D lying on the contact plane and centered at the contact point. At each cemented contact the force-displacement behavior is described by the following five parameters that defined a parallel bond: normal and shear stiffness per unit area, kn and ks, tensile and shear strengths, dc and fc; and bond radius multiplier, X, Figure C-l. These parameters are cement-based micro-properties".

The particle-based and bond-based micro properties are used to characterize a bonded particle model. Moreover, the particle density, shape and size distribution are the parameters which are considered in a BPM characterization. All of these parameters can influence the macro mechanical behavior of the BPM and the nature of failure that occur during loading.

C.3 BPM genesis procedure

The BPM material-genesis procedure produces a dense packing of non-uniform sized circular or spherical particles that are bonded at their contact points with parallel bonds. This procedure ensures that the particles are well connected and that the locked-in forces are low. Potyondy & Cundall (2004) summarized the material-genesis procedure which is quoted as follows:

1. Compact initial assembly: A material vessel consisting of planar frictionless walls is created, and an assembly of arbitrarily placed particles is generated to fill the vessel. The vessel is a rectangle bounded by four walls for PFC2D and a rectangular parallelepiped bounded by six walls for PFC3D. The normal stiffness of the wall is made slightly greater than the average particle normal stiffness to ensure that the particle-wall overlap remains small. The particle diameters satisfy a uniform particle size distribution bounded by Dmjn and Dmax (minimum and maximum diameter of the particles). To ensure a reasonably tight initial packing, the number of particles is determined in such a way that the overall desired porosity in the vessel is achieved. The particles, at half their final size, are placed randomly such that no two particles overlap.

231

Appendix C

Then, the particle radii are increased to their final values, and the system is allowed to rearrange under zero friction, Figure C-2a.

2. Install specified isotropic stress: The radii of all particles are reduced uniformly to achieve a specified isotropic stress, cr0 defined as the average of the direct stresses. These stresses are measured by dividing the average of the total force acting on opposing walls by the area of the corresponding specimen cross-section. Stresses in the PFC2D models are computed assuming that each particle is a disk of unit thickness. The magnitude of the locked-in forces (both tensile and compressive) is comparable to the magnitude of the compressive forces at the time of bond installation, Figure C-2b.

3. Reduce the number of "floating" particles: An assembly of non-uniform-sized circular or spherical particles, placed randomly and compacted mechanically, can contain a large number (perhaps as high as 15%) of "floating" particles that have less than Nf contacts, as shown in Figure C-2c. Nf is the minimum acceptable number of contacts for each particle in the assembly which is generally considered as Nf =3. It is desirable to reduce the number of floating particles so that a denser bond network can be obtained in step 4.

4. Install parallel bonds: Parallel bonds are installed throughout the assembly between all particles that are in close proximity (with a separation of less than 10"6 time the mean radius of the two particles), as shown in Figure C-2d.

5. Remove from material vessel: The material-genesis procedure is completed by removing the specimen from the material vessel and allowing the assembly to relax. This is done by deleting the vessel walls and stepping until static equilibrium is achieved. "

232

Appendix C

Figure C-2. Material-genesis procedure for a PFC2D model, after Potyondy & Cundall (2004).

233

Appendix D

Appendix D: Stress-Strain Curves for Synthetic Rock Mass Samples

This appendix illustrates the stress-strain curves recorded for the synthetic rock mass samples, presented in chapter 6, under uniaxial compressive test.

>• I I 20 13

a) b) Strain X10"-3

c) Strain X 10"-»

Figure D-l. Stress-Strain curve for synthetic rock sample of 0.05 m x 0.05 m x 0.1 m, a) Sample #1, b) Sample #2, #4 and #5, c) Sample #3.

234

Appendix D

Figure D-2. Stress-Strain curve for synthetic rock sample of 0.1 m x 0.1 m x 0.2 m, a) Sample #1 and #3, b) Sample #2, c) Sample #4, d) Sample #5.

235

Appendix D

a) a TI jo

b) Strain X10A-3

c) Strain X 10A-3 d) Strain X 1 iy-A

I 7,

e) Slram X 10A-3

Figure D-3. Stress-Strain curve for synthetic rock sample of 0.2 m x 0.2 m x 0.4 m, a) Sample #1, b) Sample #2, c) Sample #3, d) Sample #4, e) Sample #5.

236

Appendix D

a) Strain X 10A-3 b) Strain X 10*-3

o Strain X 10"-3 d) Strain X 10*-3

e) Strain X 10A-3


237

Appendix D

e) Strain X 10" 3


238

Appendix D

e) Strain X 10* 3


239

Appendix D

00 02

a) Strain X 10"-3 b) Strain X 1C/-3

00 01

c) Strain X10 " -3 d) Strain X 10" 3

e) Sl ramX10»-3

Figure D-7. Stress-Strain curve for synthetic rock sample of 7.0 m x 7.0 m x 14 m, a) Sample #1, b) Sample #2, c) Sample #3, d) Sample #4, e) Sample #5.

240

Appendix D

a) SlnanX10"-3 b) 08 oa to 13

Strain X 10A-3

c) Strain X10"-3 d)

e)


241

Appendix E

Appendix E: Stability Analysis of Vertical Raises in Hard Rock by Integrating a Fracture System into a

PFC Model

E.l Introduction

This appendix illustrates a preliminary work on how the synthetic rock mass approach, developed as part of this thesis, can be used to investigate the stability of a vertical raise. The #19A ore pass of the chapter 7 was constructed in a competent massive sulphide rock mass which was characterized by three spaced and low persistent fracture sets. In addition, the ore pass was constructed in a region where progression of two mining zones toward each other, in the abutment of the ore pass system, had resulted in high stress states around the ore pass. In order to investigate the applicability of the synthetic rock mass for stability analysis of ore pass systems in different structural and stress regimes another case study was designed in this appendix.

The stability of a vertical raise constructed in a highly persistent fractured rock mass, within a relatively low stress state region, was investigated. The construction of the vertical raise in the rock mass has resulted in several small to large rock wedges being exposed along the excavation wall sides. The stability of individual wedges was investigated using three different methods. This includes using the limit equilibrium analysis, the 2D particle flow analysis and the 3D synthetic rock mass analysis. The results of different approaches were compared and the advantage and disadvantage of each method were briefly discussed.

£.2 Structurally controlled instability in underground raises

It is recognized that the stability of underground raises is influenced by both geological structure and the stress regime. Although it is convenient to address stress and structure problems separately, this is not necessarily appropriate as these conditions tend to interact with each other. It follows that there are obvious advantages in using analytical tools that can integrate both structure and stress and thus provide more realistic tools that can investigate the stability of the excavations. Implementing such a comprehensive approach has not been easy in the past due to issues associated with data availability, data visualization, model construction and computational problems.

When stress shadows occur by progressive mining of zones adjacent to an underground raise or by relaxation of the confining stresses, instability problems can occur in the form

242

Appendix E

of structural failure. This can lead to significant unraveling and consequently rock falls in the excavation. This mechanism is most likely in blocky rocks.

E.2.1 Traditional wedge analysis

In low to moderate stress environments it may be a reasonable assumption to ignore the influence of stress. This can be justified if there are clearly defined fracture sets that are of infinite length with respect to the size of the excavation. The intersection of fractures with the exposed excavation surfaces can result in the formation of discrete rock wedges. The stability of these exposed wedges is influenced by the geometry of the wedges, their deformability and their mechanical properties. There are several limit equilibrium software packages that facilitate a kinematic analysis investigating the creation of gravity driven wedge failures.

Software packages such as Unwedge, available from Rocscience (2003) are well suited to undertake both sensitivity and parametric investigations. This type of software identifies the maximum size tetrahedral wedge that can form under a given set of conditions assuming fractures of infinite length intersecting the excavation plane. There are two basic shortcomings in this type of analysis that can result in misleading interpretations. The first one is the limited representation of the rock mass structural complexity and the other is the inability to account for the stresses acting on the excavation. Grenon & Hadjigeorgiou (2003a), as well as Liu et al. (2004) illustrated that this approach cannot adequately establish the spatial distribution of wedges created along an exposed excavation surface. Other limiting assumptions include the simplification that fractures are planar and that rock wedges behave as rigid bodies that cannot deform. The influence of stress on the stability of the excavation is ignored or is only considered as a clamping force, Curran et al. (2004).

E.2.2 Fracture system models

Although recent years have seen a subtle change in the vocabulary used to define fracture systems, the basic objective remains the generation of simulated fractures that accurately represent the salient characteristics of a population of fractures sampled in a particular rock mass. The fundamentals of fracture systems have been explained by Dershowitz & Einstein (1988) and more recently by Staub et al. (2002). Although early applications of fracture systems focused on applications in hydrogeology there have been documented applications in hard rock mines such as drift and stope stability, Grenon & Hadjigeorgiou (2003a and 2003b).

Example on the use of fracture systems in the stability of vertical or near vertical excavations has been provided by Stacey et al. (2005) and more recently by Hadjigeorgiou & Grenon (2005). Both approaches rely on the generation of a fracture

243

Appendix E

system to establish the number of potentially unstable generated wedges along the walls of an excavation. The stability of individual wedges can subsequently be established based on limit equilibrium analysis. This approach can provide a series of distributions describing wedge size, number of unstable wedges, etc. It is then possible to overcome limitations of traditional wedge analysis where fractures were assumed to be of infinite length and ubiquitous.

E.3 A case study of a vertical raise in the Canadian shield

A numerical case study is employed in order to illustrate the proposed integrated data representation and analysis techniques that can be used to investigate the stability of vertical excavations in rock. The same data set is used to provide a reference for the impact of the employed tools in the analysis and interpretation of the results.

A 30 m long vertical raise having a cross section of 2.1 m x 2.1 m was excavated in a relatively competent rock mass in the Canadian Shield. The raise was at a moderate depth and the in-situ stress state was estimated as follow (North-South direction: 0">y_5 = 35.5

MPa, East-West direction: &E_W = 38.5 MPa and Vertical stress <JV = 18 MPa). The available structural data identified three fracture sets with the properties summarized in Table E-l. In order to facilitate the analysis in the subsequent sections it was decided to keep the same fracture properties for all fracture sets (cohesion 0.4 MPa and friction angle of 40°).

Table E-l. Fracture sets characteristics.

Characteristics Se t# l Set #2 Set #3

Dip(°) 87 47 81

Dip direction (°) 311 178 70

P32 ( m 1 ) 0.6 0.7 0.8

Area (m2) 646 597 732

Another assumption was that the fracture sets were relatively long with respect to the size of the excavation. In most site investigations, this is the level of information that is available. In order to illustrate the developed concepts, it is necessary to have access to more data in order to undertake more comprehensive investigations. The P32 values were assumed to characterize the fracture intensity in the rock mass.

244

Appendix E

E.4 Structural stability of a raise in a fracture system

In order to illustrate the application of fracture systems in investigating the stability of a vertical raise we employed the case study described in section E-3. A fracture system was generated for a 50 m x 50 m x50 m rock mass using the full set of available structural data reported in Table E-l. In this work the Veneziano model, described in Dershowitz & Einstein (1988), was used to generate 405 discrete fractures, Figure E-l.

Figure E-l. 3D representation of fracture system.

A relatively large model size (50 m x 50 m x 50 m) was selected in order to minimize boundary truncation-errors that would be an issue when the generated rock mass fracture system would be further analyzed. Use of the Veneziano model resulted in fractures defined as polygons.

It should be reiterated that since fracture system models are based on stochastic generation, then, a resulting fracture system is only one of many possible systems. It is recognized that in a comprehensive analysis a series of simulations would be undertaken in order to establish the range of possible fracture systems for any given set of data. This was not undertaken in this work as the main objective was to explore the full integration of 3D fracture systems into numerical stress analysis packages.

245

Appendix E

Once the fracture system was generated, a vertical excavation (2.1 m x 2.1 m x 30 m raise) was constructed in the rock mass model. As the rock mass was fractured, the introduction of the raise exposed a number of rock wedges along the excavation walls, Figure E-2. An in house software package, UBLOCK, developed by Grenon & Hadjigeorgiou (2006), allowed the visualization and tracking of all created rock wedges. There were 15 rock wedges formed along the 4 sides of the excavation. Figure E-2 displays the rock wedges formed along the North and West sides of the raise. The bigger wedge was approximately 3.2 m3 and was created on the side wall facing North while the smallest wedge was 0.005 m3 on the opposite wall.

Figure E-2. Localization of size and distribution of wedges on the West and North walls.

A limit equilibrium analysis was used to determine the stability condition of all individual wedges daylighting into the excavation, assuming that all fractures had an angle of friction equal to 40° and cohesion of 400 kPa. Of the 15 wedges were formed along the raise walls, 5 were defined as unstable (Safety Factor <= 1). The unstable wedges were along the North face of the excavation, sliding along fracture set 2 (477178°). Table E-2 summarizes the number of wedges along each wall, and their maximum and minimum volumes.

246

Appendix E

Table E-2. Number and size of wedges on sidewalls of the raise.

North East South West

Nb. of Wedges 6 2 2 5

Unstable Wedges 5 0 0 0

Maximum size (m3) 3.20 0.81 0.22 1.20

Minimum size (m3) 0.006 0.33 0.005 0.011

Use of fracture systems in defining a rock mass is arguably more realistic than traditional representations. The limit equilibrium stability analysis assumed that intersecting fractures resulted in tetrahedral wedges. Although this is typical of what is encountered in the field, it is possible that more complex shaped wedges can potentially form. Furthermore, it is possible that generated large rock wedges can engulf smaller wedges. In this example, only the stability of the large kinematically feasible wedges was investigated, ignoring the influence of smaller wedges within a larger wedge. This reasonable simplification can be misleading as failure of smaller wedges can potentially trigger the unraveling of the surrounding rock mass. This is an area where the use of rock support may be very beneficial by preventing the unraveling of the rock mass. Finally, the preceding analysis did not consider the influence of the surrounding stress regime.

E.5 2D Stress modeling

The influence of stress on the stability of an underground raise can be addressed by any number of stress analysis packages. Although it is possible to accommodate the effect of major rock structures in a stress model, there are still important limitations on how to best introduce and interpret fault or fracture behavior.

In the present work the two and three dimensional particle flow codes were used as stress analysis tool to evaluate the stability of vertical raises located in moderate stress states.

E.5.1 Identification of longitudinal section planes

The original choice of a PFC 2D model was dictated by time constraints in generating, running and calibrating a 3D PFC model. This gain in computational time, however, necessitated a series of simplifying assumptions on the representation of a 3D rock mass generated by the fracture system into a 2D PFC model. This was addressed by generating

247

Appendix E

longitudinal sections intersecting the 3D generated rock mass, Figure E-3. In all, 6 longitudinal section planes were used, with sections 2 and 5 intersecting the centre of the raise. The remaining were spaced at 0.5 m across each side of the centre section. This has allowed tracing the defined wedges shown in Figure E-2, along the intersecting planes, Figure E-3.

Figure E-3. Representation of cross section planes (not to scale).

It is recognised that the spacing of the planes is of consequence. Depending on its size, a structurally defined wedge can potentially be intercepted by more than one plane. The 2D shape and dimensions of every wedge created by the fracture system was transposed on each plane. Consequently a PFC2D model was constructed that included all 2D wedges traced along any of the 6 selected planes.

E.5.2 Rock mass simulation in PFC2D

E.5.2.1 Simulation of intact rock properties

The procedure of intact rock simulation in PFC was discussed in chapter 5. The PFC2D (V-3.0) was used to simulate a series of uniaxial compressive tests in order to select the appropriate micro-mechanical properties. After a series of iterations the intact rock

248

Appendix E

properties summarized in Table E-3 were derived based on the micro-properties listed in Table E-4.

Table E-3. Mechanical properties of intact rock.

Property Value

UCS (MPa) 164

Elastic Modulus (GPa) 75

Poisson Ratio 0.22

Table E-4. Micro-properties used to represent intact rock in PFC2D.


Particle density (kg/m ) 2800 Parallel bond modulus (GPa) 63.8

Particle contact modulus (GPa) 63.8 Bond normal/shear stiffness 2.5

Particle normal/shear stiffness 2.5 Bond normal and shear strength (MPa) 137


E.5.2.2 Simulation of fracture properties

The strength of an in situ rock mass is influenced by the mechanical properties of its fracture network. In a PFC model, fractures can be defined as planes along which clusters of particles can slide. In defining fractures that form wedges in PFC it is necessary to assign a series of mechanical properties, which can be different, to all particles that situate on opposite sides of a fracture plane. In this two dimensional case study the ordinary contact model was applied to all particles in the model, as the "smooth-joint contact model" was not available at the time of this 2D simulation. Across the fracture planes zero bond strengths were assumed. This implied that particles adjacent to a fracture planes were assigned zero friction coefficients. Normal and shear stiffness values for particles along a joint plane were set to one-fifth of the values assigned to particle contacts in the intact portion of assembly. The particles along fracture planes were bonded to the next layer of particles, on each side of the fracture plane by bonds that were 10% weaker than the bond strengths in the intact portion of the particle assembly. The undertaken procedure followed the methodology proposed by Wang et al. (2003) where

249

Appendix E

the mechanical properties of fractures were determined using direct shear test simulations in PFC2D models. In the present case study the peak and residual shear strengths were determined for normal stresses of 0.1, 0.3, 0.45, 0.75 MPa. This has resulted in a failure criterion defined by a cohesion of 0.4 MPa and an angle of friction of 40°. These were the same mechanical properties used in the limit equilibrium analysis.

E.5.3 Particle assembly generation

The rock mass was created in the PFC2D model, was transferred along the six longitudinal planes illustrated in Figure E-3. This involved the generation of six rectangular particle assemblies, each 11 m wide and 30 m long. Each assembly was populated by 75087 particles of varying radii. Figure E-4 provides a section of the assembly prior to introducing the excavation. In all 7 groups of particles were used, at all times maintaining the ratio of maximum to minimum radius at 1.66, to generate a uniform particle size distribution. The smaller particles of radius of Rmin=15 mm were generated in the adjacent walls of excavation and the bigger particles (radius of Rmin=57mm) were generated away from the excavation. Large particles inhibit rock wedges from sliding along fracture planes, due to the inherent waviness of any plane (fracture) created by circular particles. Larger particles result in fracture planes of higher roughness. In order to limit this discrepancy finer particles were used along the excavation surfaces.

1 1 1 1 1 1 | _ 1 1 1 1 1 1 1

Width(m)

R,». (mm)

2.45

57

I

34

I

15

2.1

23

1

15

1

34

2.45

57

Figure E-4. Close-up view of a section of the particles assembly model.

250

Appendix E

E.5.4 Installation of in-situ stress

As indicated before the raise was located at a moderate depth in the following stress field North-South (N-S:CTXX=35.5 MPa, 0^=18 MPa)) and East-West (E-W:CTxr= 38.5

MPa and <%= 18 MPa). The stress was applied using a stress installation routine developed by Itasca.

E.5.5 Linking a 2D PFC model to a fracture system

The 3D fracture system was used to extract 2D traces and introduce them to the 2D bonded particle models. For each plane section the traces of fractures that formed wedges on the side walls were introduced to its related particle assembly. Then the mechanical properties of fractures defined during the calibration stage were assigned to the fracture traces in the model. Consequently, gravity was applied to the particle assembly and the raise was excavated by deleting all particles in the area of the raise. Figure E-5 illustrates the resulting wedges along plane 2 (S-N) and plane 5 in the E-W direction.

N cm»

\

1

S^^,»,^ta*<3i

■»»■■ —i

*

" ^

i

N

Figure E-5. Wedges formed along (a) plane No. 2, and (b) plane No. 5.

251

Appendix E

E.5.6 PFC2D stability analysis

At the present time there are no comprehensive quantitative criteria to define the stability of a wedge based on the results of PFC modeling. Consequently the authors used the same approach as by Wang et al. (2003) where the stability was interpreted based on the presence of an "unstable" area defined by displacement vectors. Figure E-6 provides an example of both a stable and an unstable wedge both daylighting into the excavation. It is also possible to establish the direction of movement based on the orientation and magnitude of the displacement vectors. Although this approach seemed satisfactory for the purposes of this work there is clearly much more work that has to be undertaken in developing better quantitative criteria that can be used with PFC. Figure E-7 provides a sequential illustration of the unstable wedge in plane 2. In fact PFC2D provides an interesting visualization option where you can create a "movie" of the displacement history of wedges.

VewSee X 3184».000<=> 3 149..OO0 Y 1 «06e»001 < = > ■« 1O4*«O00

Displacement M»mnum= 4134*402

Figure E-6. Displacement of wedges along plane No. (2).

252

Appendix E

Figure E-7. Evolution of wedge sliding in the plane No. (2).

PFC2D was used to investigate the stability of all defined wedges in the six longitudinal sections, Table E-5. All wedge failures were by sliding along fracture set J2 in the North wall of the excavation. For this analysis a Pentium 4-3.3 GHz computer was used. The running time for each of the six longitudinal sections, for 2,000,000 cycles, was about 70 hours.

253

Appendix E

Table E-5. Stable and unstable wedges along the six longitudinal sections.

Section Wall Wedge size (n?) Stability Failure mode

1

N 3.2 Unstable J2

1 N 0.3 Stable -1

S 0.005 Stable -

2

N 3.2 Unstable J2

2 N 0.35 Unstable J2 2

S 0.22 Stable -

3

N 3.2 Unstable J2

3 N 0.44 Stable -

3 N 0.29 Stable -

3

N 0.006 Stable -

4

E 0.81 Stable -

4 E 0.33 Stable -

4 W 1.2 Stable -

4

W 0.02 Stable -

5

E 0.81 Stable -

5 E 0.33 Stable -

5 W 1.2 Stable -

5

W 0.12 Stable -

6

E 0.81 Stable -

6 W 1.2 Stable -

6 W 0.06 Stable -

6

W 0.011 Stable -

254

Appendix E

This hybrid 3D fracture system and 2D PFC model approach required several assumptions that may not necessarily be always justifiable, including the assumption that only tetrahedral wedges are formed along the excavation. These assumptions can invariably influence the results of any stability analysis. It follows that there are obvious advantages in developing an efficient way to integrate a 3D fracture system to a PFC3D model.

E.6 3D Synthetic rock mass model

The use of a 3D model has several important technical and data visualization advantages. These include more accurate representation of rock wedges but also of the interaction between a 3D stress regime and the defined discontinuous structures in a rock mass.

The present work uses the 3D synthetic rock mass approach to simulate the structural stability of vertical raises. The same 3D fracture system described in Figure E-l was coupled to a 3D bonded particle model. This has overcome all the problems associated with transforming 3D wedges into 2D planes.

E.6.1 Simulation of intact rock and fracture properties in PFC3D

Synthetic rock mass models can be generated by linking the intact assembly of particles with a series of smooth-joints defined by a fracture system. This has the practical implication that the synthetic rock mass is broken into clusters of intact particles. This was achieved by following the same procedure as in PFC2D where the mechanical properties of intact rock and fractures were simulated. Table E-6 summarizes the micro-properties of the intact rock simulation used in the PFC3D model to reproduce the mechanical properties of intact rock listed in Table E-4.

Table E-6. Micro-properties used to represent intact rock in 3D bonded particle model.


Particle density (kg/m3) 2800 Parallel bond modulus (GPa) 84

Particle contact modulus (GPa) 84 Bond normal/shear stiffness 2.5

Particle normal/shear stiffness 2.5 Bond normal and shear strength (MPa) 178


255

Appendix E

The mechanical properties of fractures in the PFC3D model were simulated using the Smooth Joint model. This smooth-joint model simulates the behavior of an interface regardless of local particle contact orientations along the interface, Pierce et al. (2007). Particles that fall along opposite side of a smooth joint plane can overlap to insure particles sliding along the joint plane rather than forcing particles to move around one another as indicated in Figure E-8 from Itasca.

Force ^ T ^ y ^ ^ ^ - ^ Force

Figure E-8. Motion of joints assigned different contact model: (a) ordinary contact, (b) smooth contact, after Itasca (2008a).

In order to allow for a comparison of both the limit equilibrium and the PFC2D results it was decided to assign the same mechanical properties to the fractures (cohesion 0.4 MPa and friction angle 40°) as before. This necessitated the simulation of a series of triaxial tests in PFC3D model. Each test used a synthetic rock mass having a predefined smooth fracture plane at different inclinations (45°, 50°, 55° and 60°). The detail of this test simulation was discussed in chapter 5. In this example, no bonds were assigned to the particles across the fracture surface. Particles directly adjacent to a fracture were assigned a friction coefficient of 0.3. Normal and shear stiffness of particles along the fracture surfaces were 2.8xl013 N/m3 and 2.0xl013 N/m3 respectively. This configuration facilitated the generation of the desired macro-mechanical properties of the fractures.

E.6.2 Particle assembly generation and in-situ stress installation

A 3D bonded particle model of 12 m x 12 m x30 m size was constructed using the micro-mechanical properties listed in Table E-6. In all 131625 spherical particles were generated. In order to decrease the number of particles in the model and reduce the running time for the model, smaller particles were generated near the excavation and larger particles further away from the excavation walls. This resulted in four different zones populated by particles of different size radii, Table E-7. The same stresses, as previously, were applied to the rock mass. In Figure E-9 the Y axis represents the North.

256

Appendix E

Table E-7. Particle size distribution in the different zones of the bonded particle model.

Zone Rmin (mm) Rmax(mm) Nb. of particles Cumulative Nb. of particles

2 m x 2 m 85 141 74334 74334

2.7 m x 2.7 m 116 192 18881 93215

3.5 m x 3.5 m 159 263 12788 106003

6 m x 6 m 212 351 25622 131625

38.5 MPa

18 MPa

35.5 MPa

Figure E-9. Applied stresses on the bonded particle assembly model (12 m x 12 m x 30 m).

257

Appendix E

E.6.3 Linking the 3D fracture system model to 3D bonded particle model

The three fracture sets including of 156 fractures were introduced in the 3D bonded particle model to construct the 3D synthetic rock mass model, Figure E-10. This fracture tessellation of the bonded particle model resulted in 2535 discrete clusters within the generated rock mass.

In a PFC model only contact forces and particle displacements are computed. These quantities are useful when studying the material behavior on a micro-scale, but they cannot be used in a continuum model interpretation. As stress is a continuum quantity, averaging procedures are used to make the step from micro scale to a continuum. Measurement circles (2D) or spheres (3D) are used to quantify the magnitude of stress within the specified measurement area or volume. Figure E-l 1 presents the integration of stress-structure in 3D synthetic rock mass model showing contact force distribution in the cross section of model and stress states in three zones of the synthetic rock mass. Stress magnitudes in three measuring spheres of 4 meters of diameter are: (Measuring sphere #1: cr„= 34.7 MPa, ffyy= 37.5 MPa, a a =18.3, Measuring sphere #2: a „ = 35.9 MPa,

a^= 38.6 MPa, a a =19.3, Measuring sphere #3: (7^= 34 MPa, a^= 37 MPa, cra =18 MPa.)

Having established a representative rock mass, every particle was assigned a gravitational acceleration force. The final step in this procedure was the introduction of the raise in the rock mass, accomplished by deleting the particles that fall within the raise location in the model, Figure E-l2.

258

Appendix E

Figure E-10. The generated 3D synthetic rock mass model illustrating fracture defined particle clusters.

Center. Rotation X 9 537<MX>7 X 90000 Y, -4 292e-0O6 Y: 0.000 Z 9537e-007 Z 0000

Dis l 9 784e+001 Mag.: 3.05 Ang 22500

Cluster

CForce Chains

■Compression Tension

Maximum = 1 3206+007 Measurement Spheres

Figure E-l 1. Integration of stress-structure in 3D synthetic rock mass model.

259

Appendix E

Figure E-l 2. The ore pass dimensions introduced in 3D synthetic rock mass.

E.6.4 PFC3D stability analysis

Once the excavation was introduced in the 3D synthetic rock mass, the individual particles were allowed to move. This has resulted in the creation of rock wedges by clusters of bonded particles. As before the model was allowed to run over 2,000,000 cycles observing the generated wedges. It took 375 hours, to run 2,000,000 cycles in the PFC3D model.

In the particle flow codes, damage is defined as particle bonds that break under load. Any series of particles, with broken bonds, can result in the creation of macroscopic fractures in the intact rock. Particle size influences both the damage mechanism and model resolution.

In the present work there was no intact rock damage associated with bond breakage. The intact rock was quite competent, and the raise was located in a low to medium stress environment. In this numerical experiment the observed failure mechanisms were sliding along fracture planes and fall of rock blocks (wedges) towards or into the raise excavation.

This process provided a valuable insight into the mechanics of the stability of the raise by visualizing any wedges that move. Recognizing and defining all wedges is a difficult task. Although it can be facilitated by using smaller particles near the excavation walls to provide more resolution this would have an adverse impact on the required computational time. The failure mechanisms in the model are illustrated in Figure E-l3 where an

260

Appendix E

unstable wedge formed in the North wall of the raise and in Figure E-l4 where a stable wedge is situated in the East wall of the raise.

Figure E-13. Plan view and side view of an unstable wedge at the North side wall, (a) initiation of displacement of the wedge immediately after raise excavation, (b) sliding of wedge after

running of the model for 1500,000 cycles.

The PFC3D model has allowed a better insight into the mechanisms of the interaction of smaller and larger wedges. It was possible to visualize the resulting wedge displacements triggered by the movement of smaller wedges. It was also possible to visualize both sliding and gravity fall wedge failure, Figure E-l5. Using a longitudinal section in the North-South direction, as used for section 2 in the PFC2D model (Figure E-5) it was possible to identify three unstable wedges, two failing by gravity and the other by sliding.

261

Appendix E

N

^— B̂-̂ » ^ ~ »» • e».e, 1

>

Figure E-14. Plan view and side view of a stable wedge at the East side wall, (a) initiation of displacement of the wedge immediately after raise excavation, (b) wedge displacement stopped after

running of the model for 1500,000 cycles.

Figure E-l 5. Longitudinal section represents unstable wedges on the North and South walls.

262

Appendix E

The results of the stability analysis using a 3D fracture system integrated into a 3D bonded particle model are summarized in Table E-8. In the 3D model the minimum ball radius was selected based on the minimum wedge size, indicated in Table E-2 (0.005 m3). To detect the rock wedges in the PFC models as cluster of particles, the minimum ball size must be chosen smaller than the minimum wedge size. In the present work the minimum ball size in PFC3D model is 85 mm that has a volume of 0.0025 m3. Therefore the minimum ball volume is about two times smaller than the minimum wedge volume (0.0025 m versus 0.005 m3). However due to porosity between particles and particle size distribution wedges smaller than 0.02 m3 could not be detected. It is interesting to note that, based on the available structural data, the model predicts that there will be more problems along the North wall of the excavation.

The next step in this investigation looked into comparing the results obtained by the three approaches. The first approach used a 3D fracture system and a limit equilibrium structural analysis. The second was a hybrid approach where traces from the 3D fracture system were introduced in the PFC2D model. Finally, the last approach used the integrated 3D fracture system with the PFC3D model. The results of this comparison are summarized in Table E-9.

It is important to note that in this case study the integrated 3D model suggested a much greater potential for instability than the PFC2D analysis. This is explained by the fact that the 2D PFC model required that 3D structural data were transposed upon a 2D plane. As indicated this does not take into account the true shape of wedges as they are represented by apparent as opposed to real dip potentially resulting in shallower dip angles. This results in more stable wedge geometries. Furthermore, the number and spacing of cutting planes also influenced the results of a 2D PFC analysis. Furthermore, the fracture surface in PFC2D was not smooth. The 3D analysis also took into consideration the progressive instability as a result of smaller wedges engulfed in larger wedges. This is another plausible explanation for the greater number of unstable wedges predicted by the PFC3D analysis.

263

Appendix E

Table E-8. Stability analysis using the PFC3D model.

Wall Wedge size (m3) Stability Failure mode

North

3.2 Unstable Falling + Sliding along J2

North

0.44 Unstable Falling + Sliding along J2

North 0.35 Unstable Sliding along J2

North 0.3 Unstable Falling

North

0.29 Unstable Falling

North

0.006 Not-detected -

South 0.22 Unstable Falling + Sliding along Jl &

J3 South 0.005 Not-detected -

East 0.81 Unstable Falling + Sliding along J2

East 0.33 Stable -

West

1.2 Stable -

West

0.12 Unstable Falling

West 0.06 Unstable Falling West

0.02 Not-detected -

West

0.011 Not-detected -

Table E-9. Comparison of the stability condition of the wedges.

North East South West

Total Nb. of Wedges 6 2 2 5

Unstable Limit Equilibrium (SF <=1) 5 0 0 0

Unstable PFC2D 2 0 0 0

Unstable PFC3D 5+1 Unknown 1 1+ 1 Unknown 2+2 Unknown

264

Appendix E

E.7 Conclusions

Fracture systems provide a more holistic approach in investigating the stability of underground excavations that can be attained by using a particular wedge. Applying limit equilibrium analyses on all wedges formed on the sides of an excavation is arguably more appropriate than focusing on one wedge. This has been demonstrated by Hadjigeorgiou & Grenon (2005).

The next challenge is to link a comprehensive fracture system to a stress analysis package. This has been problematic in the past as it has been difficult to maintain the structural complexity of a model and successfully link it with a stress analysis package.

The use of a PFC code in tandem with a fracture system allows for a better insight on the interaction of structure and stress. Furthermore, it is possible to consider the structurally defined wedges are deformable. Finally it was possible to visualize both gravity fall and sliding of unstable wedges. This approach, however, has its own limitations. As fracture system models are based on stochastically generation of fractures then a number of simulations must be undertaken to account for the inherent variability of the rock mass associated with the probability of fracture intersections. Another concern was the necessary time of execution. This was the reason that the author linked the fracture system to a 2D PFC model.

The use of 2D model, as opposed to the 3D model, necessitated several simplifications on the way 3D wedges can be transposed upon a 2D plane. This is in itself problematic as it cannot take into account the shape of wedges. In practice any transposed wedges are represented by apparent, as opposed to real dip, potentially resulting in shallower dip angles. This can result to more stable wedge geometries. Moreover, the number and spacing of the longitudinal planes will be of influence on the results of the PFC2D analysis.

Another concern is the inherent waviness of fractures generated by particles. Although small diameter particles were used to define fracture planes in the PFC2D model there is an inherent waviness along the fracture planes which may influence the results of the wedge stability analysis. Although these limitations were overcome using the PFC3D model, computation time is still a concern for routine stability investigations.

This document demonstrated that it is possible to provide a complete 3D approach in investigating the stability of vertical excavations in hard rock. This has drawn from experience in 3D fracture systems and the use of the Particle Flow Code both in 2D and 3D.

265

References

References

Andrieux, P., O'Connor, C , & Cotesta, L. 2006. Destressing Options for the 18-21 Ore passes Complex and Associated Issues at Brunswick Mine. Technical Memorandum, p. 23.

Andrieux, P., & Simser, B. 2001. Ground stability based mine design guidelines at the Brunswick mine. In W.A. Hustrulid & R.L. Bullock, (Eds.), SME Underground Mining Methods Handbook, (pp. 207-214). Colorado: The Society for Mining, Metallurgy and Exploration.

Autio, J., Wanne, T., & Potyondy, D. 2002. Particle mechanical simulation of the effect of schistosity on strength and deformation of hard rock. In Proc. NARMS-TAC 2002, Toronto, 275-282.

Azzoni, A., & de Freitas, M.H. 1995. Experimentally gained parameters, decisive for rock fall analysis. Rock Mech. Rock Eng. 28 (2): 111-124.

Baecher, G.B., & Lanney, N.A. 1978. Trace length biases in joint surveys. In Proc. of 19th US, Symp. On Rock Mechanics, Vol. 1, pp. 56-65.

Baecher, G.B., Lanney, N.A., & Einstein, H.H. 1977. Statistical Description of Rock Properties and Sampling. In Proc. of 18th U.S. Symp. On Rock Mechanics, Colorado, pp. 5C1.1-5C1.8.

Barton, N., & Bandis, S. 1980. Some Effects of Scale on the Shear Strength of Joints. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. Vol. 17, pp. 69-73.

Barton, N., Lien, R., & Lunde, J. 1974. Engineering classification of rock masses for the design of tunnel support. Rock Mechanics, 6:189-236.

Beus, M.J., Iverson, S., & Stewart, B. 1998a. Application of Physical Modeling and Particle Flow Analysis to Evaluate Mine Ore Pass Design. Inst. Min. Metal. Trans., Sect. A: Min. Ind. 106: A110-A117.

Beus, M.J., Iverson, S.R., & Stewart, B. 1998b. Design Analysis of Underground Mine Ore Passes: Current Research Approaches. Proc. of the 100th Annual Meeting of CIM., Montreal, May 3-7, 1998.

Beus, M.J., Iverson, S.R., Dreschler, A., & Scott, V.A. 1999. Static and dynamic loads in ore and waste rock passes in underground mines. Rock Mechanics for Industry -

266

References

Proceedings of the 37th U.S. Rock Mechanics Symposium. Vail, Colorado, June 6-9, 1999. Netherlands: A. A. Balkema Publishers.

Bieniawski, Z.T. 1989. Engineering rock mass classifications. New York: Wieley.

Brady, B.H.G. 1987. Boundary element and linked methods for underground excavation design. In Analytical and computational methods in engineering rock mechanics. Allen & Unwin, London, pp. 164-204.

Brenchley, P.R., & Spies, J.D. 2006. Optimizing the Life of Ore Passes in a Deep-Level Gold Mine. The Journal of South African Institute of Mining and Metallurgy. 106, 11-16.

Brown, E.T. 1981. Rock characterization, testing and monitoring: ISRM suggested methods. Pergamon Press.

Brummer, R.K. 1998. Design of Ore passes: Methods for Determining the Useful Life of Ore passes Based on Previous Experience and Case studies. Report to CAMIRO Mining Division, p. 55.

Castro, L.A.M., Grabinsky, M.W., «fe McCreath, D.R. 1997. Damage initiation through extension fracturing in a moderately jointed brittle rock mass. Int. J. Rock Mech. Min. Sci., 34: 3-4, paper 110.

Chandler, N., Read, R., Potyondy, D., Young, P., «fe Hazzard, J. 2002. Computing Brittle Rock Fracture and Excavation Stability using the Particle Flow Code. 2nd Canadian Speciality Conference on Computing in Geotechnique, Winnipeg, April 28-30, 2002.

Chau, K.T., Wong, R.H.C., & Wu, J.J. 2002. Coefficient of restitution and rotational motions of rockfall impacts. Int. J. Rock Mech. Min. Sci., 39: 69-77.

Cho, N , Martin, CD., «fe Sego, D.C. 2007. A Clumped Particle Model for Rock. Int. J. Rock Mech. Min. Sci. AA: 997-1010.

Cundall, P.A. 2000. A Discontinuous Future for Numerical Modeling in Geomechanics. Geotech. Eng; 149: 41-47.

Cundall, P.A. 2007. Synthesizing rock mass behavior by combining a discrete fracture network with intact rock, Keynote Lecture, Special session: Numerical modeling of brittle fracture, 1st Canada-US Rock Mechanics Symposium, May 2007, Vancouver, BC, Canada.

Cundall, P. A., Pierce, M., <& Mas Ivars, D. 2008. Quantifying the size effect of rock mass strength. In Proc. SH1RMS, 2008, Perth, Australia, Australian Center for Geomechanics.

267

References

Curran, J.H., Corkum, B., Hammah, R.E. 2004. Three dimensional Analysis of underground wedges under the influence of stresses. Proceedings, Gulf Rock, the 6th North American Rock Mechanics Symposium, Houston, Texas, 5-9 June 2004.

Dershowitz, W.S., «fe Einstein, H.H. 1988. Characterizing Rock Fracture Geometry with Fracture System Models. Rock Mech. Rock Eng, 21: 21-51.

Dips, 2006. Dips Manual-V5.1, Rocscience.

Dukes, G.B., Van Oort, J., «fe Van Heerden, F.D. 2006. Rockpass Stability and Rehabilitation at AngloGold Ashanti's Tau Lekoa Mine. The Journal of South African Institute of Mining and Metallurgy. 106, 5-10.

Elmo, D., «fe Stead, D. 2009. An integrated numerical modeling-discrete fracture network approach applied to the characterization of rock mass strength of naturally fractured pillars. Rock Mech. Rock Eng. DOI: 10.1007/s00603-009-0027-3.

Esmaieli, K. <& Hadjigeorgiou, J. 2009. Investigations of finger configuration on degradation of ore pass walls. 3 r Canada-U.S. Rock Mechanics Symposium (Rock Engineering in Difficult Conditions), May 8-13, 2009, Toronto, ON, Canada.

Esmaieli, K., Hadjigeorgiou, J. «fe Grenon, M. 2009a. Investigating the influence of scale on rock mass strength using a synthetic rock mass model. International Conference on Rock Joints and Jointed Rock Masses, 4-11 January 2009, Tucson, AR, USA.

Esmaieli, K., Hadjigeorgiou, J. «fe Grenon, M. 2009b. Estimating representative elementary volume (REV) size, for structural and mechanical rock mass properties, using synthetic rock mass models: A case study from the Brunswick Mine. Submitted to Int. J. Rock Mech. Min. Sci.

Esmaieli, K., Hadjigeorgiou, J., Grenon, M., «fe Harrisson, R. 2008. Ore pass stability analysis at the Brunswick Mine using PFC3D. Is ' International FLAC/DEM Symposium on Numerical Modeling, 25-27 August 2008, Minneapolis, MN, USA.

Everitt, R.A., <& Lajtai, E.Z. 2004. The influence of rock fabric on excavation damage in the Lac du Bonnett granite. Int. J. Rock Mech. Min. Sci., Al: 1277-1303.

Fakhimi, A. 2004. Application of slightly overlapped circular particles assembly in numerical simulation of rocks with high friction angles. Eng. Geo!., 74: 129-138.

Fakhimi, A., Carvalho, F., Ishida, T., <& Labuz, J.F. 2002. Simulation of failure around circular opening in rock. Int. J. Rock Mech. Min. Sci. (39): 507-515.

Ferguson, G. 1991. Ore Pass Design Guidelines. Report for the Mining Research Directorate, Bharti Engineering and Ass., 99 pp.

268

References

Gardner, L.J., «fe Fernandes, N.D. 2006. Ore Pass Rehabilitation- Case studies from Impala Platinum Limited. The Journal of South African Institute of Mining and Metallurgy. 106, 17-23.

Gibbon, J.A. 1976. The Repair of the Main Rock-pass Systems at Kloof Gold Mining Company Limited. Ass. Mine Managers S.Afr., Papers and Discussions, 1976-1977, pp. 283-300.

Gibbs, G.R. 1979. The Experimental Repair of a Main Rock-pass on Doornfontein Gold Mining Company, Limited. Ass. Mine Managers S.Afr., Papers and Discussions, 1978-1979, pp. 413-425.

Godin, R. 1987. Structural Model of Brunswick #12 Mines as an Aid to Mining and Exploration. Presentation at Annual Meeting of the CIM's New Brunswick Branch, Bathurst, NB. Canada.

Golder, 2005. FracMan Technology Group, Golder Associates, www.fracman.golder.com

Goodfellow, W. 1975. Rock Geochemical Exploration and Ore Genesis at Brunswick No. 12 Deposit, New Brunswick. PhD. Thesis, The university of New Brunswick.

Goodman, R., Heuze, F.E., <& Bureau, G.J. 1972. On modeling techniques for the study of tunnels in jointed rock. The 14th U.S. Symposium on Rock Mechanics, University Park, PA. June 11-14 1972.

Goodwill, D.J., Craig, D.A., <& Cabrejos, F. 1999. Ore pass design for reliable flow. J. Bulk Solid Handling. Volume 19,No.l: 13-21.

Grenon, M., «fe Hadjigeorgiou, J. 2003a. Drift reinforcement design based on discontinuity network modeling. Int. J. Rock Mech. Min. Sci., 40: 833-845.

Grenon, M., «fe Hadjigeorgiou, J. 2003b. Open stope stability using 3D joint networks. Rock Mech. Rock Eng, 36: 183-208.

Grenon, M. «fe Hadjigeorgiou, J. 2006. Fracture-Underground. A fracture system based underground excavation wedge stability analysis software package, Version 1.23.

Grenon, M., <& Hadjigeorgiou, J. 2008a. Fracture-SG. A fracture system generator software package. Version 2.17.

Grenon, M., <& Hadjigeorgiou, J. 2008b. A Design Methodology for Rock Slopes Susceptible to Wedge Failure Using Fracture System Modeling. Eng. Geol, 96: 78-93.

Grenon, M., Hadjigeorgiou, J., «fe Harrison, J.P. 2001. Drift Stability in Moderately Jointed Rock. CIMBulletin, 94: 245-302.

269

http://www.fracman.golder.com

References

Hadjigeorgiou, J., Esmaieli, K., «fe Grenon, M. 2007. Investigating raise stability using fracture system and particle flow tools. In Proceedings of the 1st CA-US Rock Mechanics Symposium, Vancouver, May 27-31, 2007, eds. E. Eberhardt, D. Stead, and T. Morrison, 341-349. Taylor «fe Francis.

Hadjigeorgiou, J., Esmaieli, K., & Harrisson, R. 2008. Observations of Ore Pass System Performance at Brunswick Mine. CIM Bulletin, Vol. 3, No.5.

Hadjigeorgiou, J., Esmaieli, K., <& Grenon, M. 2009. Stability Analysis of Vertical Excavations in Hard Rock by integrating a fracture system into a PFC model. Tunne. Undergr. Sp. Tech. 24 (3), 296-308.

Hadjigeorgiou, J., «fe Grenon, M. 2005. Design of Ventilation Raises Using Fracture Systems. The 40th US., Symposium on Rock Mechanics. 25-29 June, Alaska, USA.

Hadjigeorgiou, J., <& Lessard, J.F. 2003. The Case for Liners in Ore Pass Systems. 3rd International seminar on Surface Support Liners. 25-26 August, Quebec City, Canada.

Hadjigeorgiou, J., <& Lessard, J.F. 2007. Numerical investigations of ore pass hang-up phenomena. Int. J. Rock Mech. Min. Sci. AA: 820-834.

Hadjigeorgiou, J., Lessard, J.F., «fe Flament, F. 1995a. Characterizing in situ Block Distribution using a Stereological Model, Canadian Tunneling, pp. 201-211.

Hadjigeorgiou, J., Lessard, J. F., <& Mercier-Langevin F. 2004. Issues in selection and design of ore pass support. In Fifth International Symposium on Ground Support in Mining and Underground Construction. Perth, Australian.

Hadjigeorgiou, J., Lessard, J.F., <& Mercier-Langevin, F. 2005. Ore pass practice in Canadian mines. The Journal of South African Institute of Mining and Metallurgy. 105, 809-816.

Hadjigeorgiou, J., Lessard, J.F., Villaescusa, E., «fe Germain, P. 1995b. An appraisal of Structural Mapping Techniques, Mechanics of Fractured and Faulted Rock. In Mechanics of Jointed and Faulted Rock, Rossmanith (edt.), Balkema, pp. 193-199.

Hadjigeorgiou, J., «fe Mercier-Langevin, F. 2008. Estimating Ore Pass Longevity in Hard Rock Mines. The 42nd US Rock Mechanics Symposium and 2nd US-Canada Rock Mechanics Symposium, 29 June-2 July 2008, San Francisco, USA.

Hagan, T.O., «fe Acheampong, E. 1999. Current Design, Support and Maintenance of Rockpasses and Assessment of Practices Applicable at Depth. In Proc. of the 2nd South African Rock Eng. Symp., Johannesburg, South Africa, 13-15 September, 62-79.

270

References

Hambley, D.F. 1987. Design of Ore Pass Systems for Underground Mines. CIM Bulletin, Vol. 897, pp. 25-30.

Hambley, D.F., Pariseau, W.G., & Singh, M.M. 1983. Guidelines for open-pit ore pass design. Contract Report US Bureau of Mines, Volume 1, 168pp.

Hamrin, H., «fe Kessedy, B., 1982. Choosing an underground mining method. Atlas Copco MCT AB Sweden, (Publication 15073a).

Harrison, J.P., <& Hudson, J.A. 2000. Engineering rock mechanics. Part 2: Illustrative worked examples. Pergamon, Oxford.

Harrisson, R. 2008. Personal Communication.

Hart, R. 2006. Case Study of the Rockpass System at Kloof No. 3 Shaft. The Journal of South African Institute of Mining and Metallurgy. 106, 1-4.

Heuze, F. 1980. Scale effects in the determination of rock mass strength and deformability. Rock Mech., 12: 167-192.

Hilner, N.E. 1966. The Re-Development of a Main Shaft Ore Pass System. Ass. Mine Managers S.Afr., Papers and Discussions, 1966-1967, pp. 25-37.

Hoek, E. 2001. Rock mass properties for underground mines. In Underground Mining Methods: Engineering Fundamentals and International Case Studies. Edited by W. A. Hustralid, R. Bullock, Society for Mining, Metallurgical and Exploration (SME), 467-474.

Hoek, E.T, Grabinsky, M.W., <& Diederichs, M.S. 1990. Numerical modeling for underground excavation design. Trans. Instn Min. Metal I. Sect. A: Min. industry. Vol. 100. pp. A22-A30.

Hudson, J.A., <& Harrison, J.P. 1997. Engineering Rock Mechanics, Pergamon, 444 p.

Hutchings, I.M. 1992. Tribology: Friction and wear of engineering materials. CRC Press, 133-197.

Imre, B., Rabsamen, S., «fe Springman, S.M. 2008. A coefficient of restitution of rock materials. Computers & Geosciences, 34: 339-350.

Itasca, 1999. Particle Flow Code 2D/3D-V3.0, Itasca Consulting Group, Inc.

Itasca, 1999. Universal Distinct Element Code (UDEC), Three Dimensional Distinct Element Code (3DEC), Itasca Consulting Group, Inc.

271

References

Itasca, 2008a. Particle Flow Code, PFC2D/3D-V.4.0, Itasca Consulting Group, Inc.

Itasca, 2008b. Fish-Lab, V-l, Itasca Consulting Group, Inc.

Itasca, 2008c. FLAC, Itasca Consulting Group, Inc.

Iverson, S.R, Jung, S.J., «fe Biswas, K. 2003. Comparison of ore pass computer simulations for design against dynamic load. SME Annual Meeting, 24-26 February, Cincinnati, USA.

Jing, L. 2003. A review of techniques, advances and outstanding issues in numerical modelling for rock mechanics and rock engineering. Int. J. Rock Mech. Min. Sci., (40): 283-353.

Jing, L., & Stephansson, O. 2007. Fundamentals of Discrete Element Methods for Rock Engineering Theory and Application. Elsevier Publication, p. 545.

Joughin, W.C, «fe Stacey, T.R. 2005. Risk Associated with Rockpasses in Deep-level Tabular Mines Based on Historical Pass Performance. The Journal of South African Institute of Mining and Metallurgy. 105, 795-802.

Jung, S.J., <& Iverson, S.R. 2004. Investigation of coefficient of restitution for rocks impacting other rocks and steel structures in ore passes. In Gulf Rock, Houston, Texas, June 5 - 9 , 2004.

Kachanov, L.M. 1986. Introduction to continuum damage mechanics. M. Nijhoff, Dordrecht, The Netherlands.

Kazakidis, V.N., «fe Morrison, D.M. 1994. Analysis of Key Factors Affecting Ore Pass Stability in High Stressed Rock Masses: A Case Study. The ISRM Int. Symp. 10-14 May 1994, Santiago, Chile.

Kirsten, H.A.D., «fe Klokow, J.W. 1979. Control of Fracturing in Mine Rock Passes. Proceeding of 4th Int. Cong. On Rock Mechanics, Vol.1, Montreux, Swiss.

Krauland, N., Soder, P., & Agmalm, G. 1989. Determination of rock mass strength by rock mass classification-some experience and questions from Boliden mines. Int. J. Rock Mech. Min. Sci., 26(1), 115-123.

Kulatilake, P.H.S.W., Malama, B., & Wang, J. 2001. Physical and particle flow modeling of jointed rock block behavior under uniaxial loading. Int. J. Rock Mech. Min. Sci., (38): 641-657.

272

References

Kulatilake, P.H.S.W., Park, J., «fe UM, J. 2004. Estimation of rock mass strength and deformability in 3D for a 30 m cube at a depth of 458 m at Àspô Hard Rock Laboratory. J. Geotech. Geol. Eng, (22): 313-330.

Kulatilake, P.H.S.W., Wathugala, D.N., «fe Stephansson, O. 1993. Joint network modeling with a validation exercise in Strip mine, Sweden. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 30, 503-526.

Kulatilake, P.H.S.W., «fe Wu, TH. 1984. Estimation of mean trace length of discontinuities. Rock Mech. and Rock Eng. 17:215-232.

Larson, M.K., Iverson, S.R., Stewart, B.M., & Walker, K. 1998. Preliminary Assessment of Particle Flow Code as a Tool to Assess Ore Pass Safety. Int. J. Rock Mech. & Min. Sci. 35: 4/5, paper 092.

Laslett, G.M. 1982. Censoring and edge effects in areal and line transect sampling of rock joint traces. Mathematical Geology, 14(2): 125-140.

Lessard, J.F. 2004. Influence La Configuration Des Cheminées À Minerai Sur L'Écoulement Du Matériel. PhD. Thesis, Laval University, Quebec, Canada.

Lessard, J.F., <& Hadjigeorgiou, J. 2003. Design tools to minimize the occurrence of ore pass interlocking hang-ups in metal mines. ISRM 2003-Technology roadmap for rock mechanics, South African Institute of Mining and Metallurgy, 8-12 September 2003, Gauteng, South Africa.

Lessard, J.F., «fe Hadjigeorgiou, J. 2006. Ore pass Database: Quebec Underground Metal Mines. CIM Bulletin, Vol. 99, NO. 1093.

Liu, J., Zhongkui, L., <& Zhuoyuan, Z. 2004. Stability analysis of block in the surrounding rock mass of a large underground excavation. Tunn. Undergr. Sp. Tech. 19: 35-44.

Loughran J.G., Anderson S.I., «fe Owen, D.R.J. 2003. Explicit Predictive Technology for Studying the Evolution of Damage in Ore Passes. 1st Australian Conference of Ground Control in Mining-Technology and Practice, p. 95-102, 10-13 November, Sydney, Australia.

Luff, W. 1977. Geology of Brunswick No. 12 Mine. CIM Bulletin, Volume 70, No. 782, pp. 109-119.

Mardia, K.V. 1972. Statistics of Directional Data. Academic Press, London-New York.

Martin, CD., Kaiser, P.K., «fe McCreath, D.R. 1999. Hoek-Brown Parameters for Predicting the Depth of Brittle Failure Around Tunnels. Can. Geotech. J. 36: 136-15.

273

References

Martin, CD., Read, R.S., «fe Martino, J.B. 1997. Observations of brittle failure around a circular test tunnel. Int. J. Rock Mech. Min. Sci., 34(7): 1065-1073.

Mas Ivars, D., Potyondy, D., Pierce, M., <& Cundall, P. 2008. The smooth-joint contact model. In 8th World Congress on Computational Mechanics, 5th European Congress on Computational Methods in Applied Sciences and Engineering, Jun 30- Jul 5, 2008, Venice, Italy.

Mauldon, M. 1998. Estimating Mean Fracture Trace Length and Density from Observations in Convex Windows. Rock Mech. Rock Eng, 31(4), 201-216.

Mauldon, M., Dunne, W.M., & Rohrbaugh, M.B. 2001. Circular scanlines and circular windows: new tools for characterizing the geometry of fracture traces. J. Struct. Geol, 23:247-258.

McCracken, A., <& Stacey, T.R. 1989. Geotechnical Risk Assessment for Large Diameter Raise Bored Shafts. Trans. Instit. Min. Metal. (Section A: Mining Industry) 98, A145-A150.

McCutcheon, S.R., Gower, S.J., Lentz, D.R., Langton, J.P., Walker, J.A., Wilson, R.A., Fyffe, L.R., Hamilton, A., <& Luff, W. 1997. Geology and Massive Sulphides of the Bathurst Camp. New Brunswick, Field Trip Report, Geological Association of Canada! Mineralogical Association of Canada.

Mercier-Langevin, F., <& Hadjigeorgiou, J. 2004. Design Considerations for Ore pass Systems in Massive Sulphide Nickel Operations. Report submitted to INCO Limited.

Miller, F., «fe Jacob, D. 1996. Évaluation tridimensionnelle des monteries avec un laser. Proc. 2nd N.A.R.M.S. Rock Mechanics Tools and Techniques, Montreal, pp. 323-329.

Mine Modeling Pty. Ltd., 2008. Map3D-V-50. Australia.

Min, K.B., <& Jing, L. 2003. Numerical determination of the equivalent elastic compliance tensor for fractured rock masses using the distinct element method. Int. J. Rock Mech. Min. Sci. 40: 795-816.

Morrison, D.M., <& Kazakidis, V.N. 1995. Orepasses in Burst-Prone Ground: Current Practice and Alternatives. 12th Mine Operators Conference. Timmins, Canada.

Musto, G. 2004. A numerical study of the performance and safety issues for ore pass systems. Final WMRC Research Report.

Nazeri, H. 2001. Development of a distinct element methodology for simulation of gravity flow of ore in ore passes. PhD. Thesis, Colorado School of Mines, Co. USA.

274

References

Nazeri, H., «fe Rozgonyi, T.G. 2003. The modeling of gravity flow of ore in ore passes. SME Annual Meeting, 24-26 February, Cincinnati, USA.

Noranda Technology Center (NTC), 1998. Propriétés Mécaniques Du Matériau Intact et des Discontinuités de Type A à la Mine Brunswick. Reported by CANMET and URSTM.

Olofsson, L, «fe Fredriksson, A. 2005. Strategy for a numerical rock mechanics site descriptive model: future development of the theoretical/numerical approach. SKB Report, R-05-43, Stockholm, Sweden.

Osae, W., «fe De Lange, J.A.J. 2003. Ore pass Rehabilitation- A Case Study at Elandsrand Gold Mine. 3rd International seminar on Surface Support Liners, 25-26 August, Quebec City, Canada.

O'Shaughnessy, J.P. 2001. Use of Spray-on Liners in High Impact Areas of an Orepass, Case Study of Stobie Mines # 7 Ore pass. Reported to Inco.

Park, E.S., Martin, CD., <& Christiansson, R. 2004. Simulation of the mechanical behavior of discontinuous rock masses using a bonded particle model. In Proceeding of Gulf Rock, the 6th North American Rock Mechanics Symposium, 5-9 June 2004, Houston, Texas, USA.

Park, H.J., West, T.R., «fe Woo, I. 2005. Probabilistic analysis of rock slope stability and random properties of discontinuity parameters, Interstate Highway 40, Western North Carolina, USA. Eng. Geo!., 79: 230-250.

Peck, W. A., «fe Lee, M.F. 2008. Stability of Raise Bored Shafts in Australian Mines. In 13th Australian Tunneling Conference, 4-7 May 2008, Melbourne, Australia.

Pierce, M., Cundall, P.A., «fe Potyondy, D. 2007. A synthetic rock mass for jointed rock. In Proceedings of the 1st CA-US Rock Mechanics Symposium, Vancouver, May 27-31, 2007, eds. E. Eberhardt, D. Stead, and T. Morrison, 341-349. Taylor <& Francis.

Pine, R.J., Coggan, J.S., Flynn, Z.N., «fe Elmo, D. 2006. The Development of a New Numerical Modelling Approach for Naturally Fractured Rock Masses. Rock Mech. Rock Eng. 39(5), 395-419.

Potyondy, D., «fe Autio, J. 2001. Bonded-particle simulations of the in-situ failure test at Olkiluoto. Proc. Of the 38th US Rock Mechanics Symposium, Washington, DC, pp. 1553-1560.

Potyondy, D., «fe Cundall, P.A. 1998. Modeling notch formation mechanisms in the URL Mine-by Test Tunnel using bonded assemblies of circular particles. Int. J. Rock Mech. Min. Sci., 35(4-5) paper 067.

275

References

Potyondy, D., «fe Cundall, P.A. 2001. The PFC Model for Rock: Predicting Rock-Mass Damage at the Underground Research Laboratory. Itasca Consulting Group, Inc.; Ontario Hydro, Nuclear Waste Management Division, Report 06819-REP-01200-10061-ROO.

Potyondy, D., «fe Cundall, P.A. 2004. A bonded particle model for rock. Int. J. Rock Mech. Min. Sci., 41: 1329-1364.

Priest, S.D. 1993. Discontinuity Analysis for Rock Engineering. Chapman «fe Hall Publication, London.

Priest, S.D., <& Hudson, J.A. 1976. Discontinuity spacings in rock. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. 16, 339-362.

Priest, S.D., <& Hudson, J.A. 1981. Estimation of Discontinuity Spacing and Trace Length Using Scanline Survey, Int. J. Rock Mech. Min. Sci. Geom. Abs., 18: 182-197.

Rech, J.R., Otto, W.M., «fe Hagan, T.O. 1992. Attempts to Combat Scaling and Blockage Problems in Deep Mine Rockpasses. Proc. Symp. On Orepasses and Combustible Materials, p. 221-240.

Rockfield, 2008. ELFEN, Swansea, UK.

Rocscience, 2003. Unwedge-V3. Toronto, Canada.

Rosso, R.S. 1976. A Comparison of Joint Stiffness Measurements in Direct Shear, Triaxial Compression, and In Situ. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. (13): 167-172.

Schoombee, A. du P., <& Van Wyk, J.J. 1982. The Support and Rehabilitation of an Ore pass at Vaal Reefs Exploration and Mining Company, Limited. Ass. Mine Managers S.Afr., Papers and Discussions, 1982-1983, pp. 343-364.

Schultz, R. 1995. Limits on strength and deformation properties of jointed basaltic rock masses. Rock Mech. Rock Eng, 28 (1): 1-15.

Schultz, R. 1996. Relative scale and the strength and deformability of rock masses. J. Structural Geol. Vol. 18, No. 19, pp.1139-1149.

Sellers, E.J., «fe Klerck, P. 2000. Modeling of the effect of discontinuities on the extent of the fracture zone surrounding deep runnels. Tunn. Undergr. Sp. Tech., 15(4): 463-469.

Sen, Z., <& Kazi, A. 1984. Discontinuity spacing and RQD estimates from finite length scanline. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 21 (4), 203-212.

276

References

Simser, B., <& Andrieux, P. 1999. Seismic Sources Mechanisms at the Brunswick Mine. In 14th CIM Underground Operators Conference, Bathurst, NB., Canada.

Simser, B., Andrieux, P., Falmagne, V., oulson, A., & Grenon, M. 2002. The Rock Mechanics Design for a 352800 tonne mass blast at the Brunswick Mine, Bathurst, New Brunswick. In 3rd North American Rock Mechanics Symposium, Toronto, Canada.

Sjoberg, J., Lundman, P., Nordlund, E., <& Quinteiro, C. 2003. Stability Analysis of Ore Passes in the Kiirunavaara Mine. ISRM 2003, Technology Roadmap for Rock Mechanics, South African Institute of Mining and Metallurgy.

Smith, E.L., O'Reilly, M.K., «fe Langston, R. 2006. Increasing Service Life of Muck Passes at the East Boulder Mine. In Golden Rocks, The 41st U.S. Symposium on Rock Mechanics, 17-21 June 2006, Golden, Colorado, USA.

Stacey, T.R., <& Erasmus, B.J. 2005. Setting the Scene: Rockpass Accident Statistics and General Guidelines for the Design of Rockpasses. The Journal of the South African Institute of Mining and Metallurgy. 105, 745-752.

Stacey, T.R., «fe Swart, A.H. 1997. Investigation into Drawpoints, Tips Orepasses and Chutes. Vol. 1. Report to the Safety in Mines Research Advisory Committee, 112 pp.

Stacey, T.R., Wesseloo, J., & Bell, G. 2005. Predicting the Stability of Rockpasses From the Geological Structure. The Journal of The South African Institute of Mining and Metallurgy. 105, 803-808.

Starfield, J.L., «fe Cundall P.A. 1988. Towards a methodology for rock mechanics modelling. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr. (25): 99-106.

Staub, I., Fredriksson, A., «fe Outters, N. 2002. Strategy for a rock mechanics site descriptive model. Report R-02-02., Swedish Nuclear Fuel and Waste Management Company.

Stringer, R.P., <& Lategan, J.C 1982. Solution to Problem Encountered with the Main Rockpasses on No. 8 Shaft, St-Helena. Ass. Mine Managers S.Afr., Papers and Discussions, 1982-1983, pp. 365-380.

Tannant, D.D., «fe Wang, C. 2004. Thin tunnel liners modeled with particle flow code. Eng. Computations, 21, 2-4, 318-342.

Turcotte, P., Hadjigeorgiou, J., & Lessard, J.F. 2003. Modélisation numérique de l'écoulement du matériel dans les cheminées è minerai et è stérile. 18e Colloque en contrôle de terrain de l'Association Minière du Québec, Val d' Or.

277

References

Université Laval, Rock Mechanics Laboratory, 2007. Results of Laboratory Tests-Xstrata Zinc-Brunswick Mine.

Van Heerden, D., Fryda, H., «fe Saucier, F. 2005. Elements for Effective Design of Abrasion Resistant Concretes. The Journal of South African Institute of Mining and Metallurgy. 105,771-782.

Vieira, F.M.C.C, «fe Durrheim, R.J. 2005. Design and Support of Rockpasses at Ultra-Deep Levels. The Journal of South African Institute of Mining and Metallurgy. 105, 783-794.

Villaescusa, E., <& Brown, E.T. 1992. Maximum Likelihood Estimation of Fracture Size from Trace Length Measurements. Rock Mech. Rock Eng, 25: 67-87.

Warburton, P.M. 1980. A Stereographical Interpretation of Joint Trace Data. Int. J. Rock Mech. Min. Sci. & Geomech. Abstr., 17:181-190.

Wang, C, Tannant, D.D., <& Lilly, P.A. 2003. Numerical analysis of the stability of heavily jointed rock slopes using PFC2D. Int. J. Rock Mech. Min. Sci., (40): 415-424.

Zhang, L., & Einstein, H.H. 2000. Estimating the Intensity of Rock Discontinuities. Int. J. Rock. Mech. Min. Sci., 37: 819-837.

Zhang, L., <& Einstein, H.H. 1998. Estimating the mean trace length of rock discontinuities. Rock Mech. Rock Eng, 31 (4), 217-235.


http://maps.google.com/

278


http://maps.google.com/

List of Contributions

List of Contributions

Esmaieli, K., Hadjigeorgiou, J. and M. Grenon, 2009. Estimating geometrical and mechanical REV, based on sysnthetic rock mass models at Brunswick Mine. Int. J. Rock Mech. Min. Sci (2010), dot 10.1016/j.ijrmms.2010.05.010.

Esmaieli, K. and J. Hadjigeorgiou, 2009. Investigations of finger configuration on degradation of ore pass walls. 3 r Canada-US. Rock Mechanics Symposium (Rock Engineering in Difficult Conditions), May 8-13, 2009, Toronto, ON, Canada.

Esmaieli, K, Hadjigeorgiou, J. and M. Grenon, 2009. Investigating the influence of scale on rock mass strength using a synthetic rock mass model. International Conference on Rock Joints and Jointed Rock Masses, 4-11 January 2009, Tucson, AR, USA.

Hadjigeorgiou, J., Esmaieli, K., and M. Grenon, 2009. Stability Analysis of Vertical Excavations in Hard Rock by integrating a fracture system into a PFC model. Tunnel. Underg. Space Technoi. 24 (3), 296-308.

Hadjigeorgiou, J., Esmaieli, K. and R. Harrisson, 2008. Observations of Ore Pass System Performance at Brunswick Mine. CIM Bulletin, Vol. 3, No.5.

Esmaieli, K., Hadjigeorgiou, J. Grenon, M. and R. Harrison, 2008. Ore pass stability analysis at the Brunswick Mine using PFC3D. Is ' International FLAC/DEM Symposium on Numerical Modeling, 25-27 August 2008, Minneapolis, MN, USA.

Hadjigeorgiou, J., Esmaieli, K. and R. Harrisson, 2008. Observations of Ore Pass System Performance at Brunswick Mine. Maintenance Engineering/ Mine Operators (CIM-MEMO) Conference, 23-25 February 2008, Val d'Or, QC, Canada.

Hadjigeorgiou, J., Esmaieli, K. and M. Grenon, 2007. Investigating raise stability using fracture system and particle flow tools. Is ' Canada-US Rock Mechanics Symposium, 27-31 May 2007, Vancouver, BC, Canada.

279

Documents

stability analysis of ore pass systems at brunswick mine