B L O C K C A V E MINE INFRASTRUCTURE

RELIABILITY APPLIED TO PRODUCTION

PLANNING

by

ENRIQUE RUBIO

B. Eng., Universidad de Chile, Chile, 1998

M.A.Sc, The University of British Columbia, Canada, 2002

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF

THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

in

THE F A C U L T Y OF G R A D U A T E STUDIES
(Mining Engineering)

THE UNIVERSITY OF BRITISH C O L U M B I A

February 2006

©Enrique Rubio, 2006

 ABSTRACT
The production promise of a mine should reflect the fundamental models that sustain the mining
system. Commonly this promise is formalized by the production schedule of a mine which is a
bankable document that supports the decision of whether or not to pursue (or continue to pursue)
the mining venture. Currently there are several computer based applications that enable mining
engineers to compute a production schedule for a block cave operation. However, several
operational upsets such as hang ups, oversize material, wet muck and rock instability affect the
availability of mining infrastructure jeopardizing the original production estimates. These upsets
can be related to geotechnical properties and caving processes in the rock mass. The current
schedulers do not incorporate or account for geotechnical properties and caving processes. Thus,
they often overestimate the production capacity of the mine.

In this dissertation, a methodology has been devised for using observations of the failure
frequency of mining infrastructure such as draw points, production drifts and ore passes to assess
the reliability of this infrastructure to sustain a given production schedule. The novel aspect of
measuring draw point reliability in this way is that it effectively subsumes complex geotechnical
phenomena that lead to draw point failure such as geological conditions, stress concentration, or
coarse fragmentation. The research found that the rate of occurrence of failure of a draw point
can be characterized by a "bathtub curve" whose shape changes with the geotechnical
characteristics of the rock mass, mining system and stress regime.

The final phase of the research integrated the estimated mining infrastructure reliability into
production scheduling through a reliability model. This integrated model provided the ability to
generate from a number of draw points, a production plan in which a subset of the draw points
will yield the requested tonnage with an associated degree of reliability based on the reliability of
individual components of the mining infrastructure. Validation of the reliability model
demonstrated that it does reproduce the tonnage distribution curve and consequently estimates
the technical uncertainty of a production schedule related to mining infrastructure availability.

ii
 TABLE OF CONTENTS
ABSTRACT. ii
TABLE OF CONTENTS iii
LIST OF TABLES . v
LIST OF FIGURES vi
ACKNOWLEDGMENTS ix
1 INTRODUCTION.. 1
1.1 Statement of the Problem 4
1.2 Research Question 7
1.3 Contribution of the Thesis 8
1.4 Organization of the Work ; 9
2 LITERATURE REVIEW 10
2.1 History of Block Caving 14
2.2 Fundamental Models of Block Caving 16
2.3 Block Cave Production Scheduling : 20
2.4 Uncertainty in Block Cave Production Scheduling 25
3 FAILURE BEHAVIOUR OF MINING INFRASTRUCTURE , 28
3.1 Operational Database ': ." 28
3.1.1 Daily tonnage records 29
3.1.2 Daily status records 31
3.1.3 Convergence records 32
3.1.4 Records of draw point hang up 33
3.1.5 Records of oversize draw points 36
3.1.6 Production schedules 36
3.2 Effect of Geotechnical Events on Production Performance 36
3.2.1 Effect of hang ups on production performance .37
3.2.2 Effect of production performance on induced stresses.'. 40
3.3 Failure of mining infrastructure 47
3.3.1 Data Analysis for computing draw point ROCOF curve 55
3.4 Draw point ROCOF curve as a function of geotechnical domains 56
3.5 Draw Point Reliability as an Indicator of Geotechnical Events 59

iii
 3.5.1 Measurements of draw point reliability 60
3.5.2 Effect of hang up frequency on draw point reliability 63
3.5.3 Effect of induced stress on draw point reliability 63
4 PRODUCTION SCHEDULING INTRODUCING THE CONCEPT OF
RELIABILITY 65
4.1 Definition of System Reliability 65
4.2 Reliability of a System with Redundancy at the Component Level 68
4.3 Block Cave Reliability Model 70
4.3.1 Draw point productivity as a function of draw point yield and draw cycle 76
4.4 Production Scheduling Integrating Draw Point Reliability 78
4.5 Production Schedule Reliability 83
4.6 Sensitivity of Different Inputs to the Reliability Model 87
4.7 Using the Reliability Model to Assess the Value of Overdrawn Draw Points 89
4.8 Using the Reliability Model to Support Tactical Decisions 90
4.9 Definition of Production Capacity of a Block Cave Mine using Reliability 91
4.10 Expected Tonnage versus Adjusted Draw Rate to Assess Production Capacity 93
5 MODEL VALIDATION 95
5.1 Reliability Model Calibration Using Absolute Reliability 96
5.2 Reliability Model Calibration Using Tonnage Distribution 98
5.3 Calibration Based On Expected Tonnage per Period 101
6 DISCUSSION AND CONCLUSIONS 103
6.1 Discussion 103
6.2 Conclusions 106
6.3 Recommendations 108
REFERENCES 110
APPENDICES 120
A Proposed Operational Database in Block Cave Mines 120
B Recursive Algorithm for the Reliability of a k-out-of-n System 123
C Proposed Production Scheduler Workflow 127

iv
 LIST OF TABLES
Table 3.1 Geotechnical information of the mine operations 28
Table 3.2 Features of the mining system 29
Table 3.3 Operational and planning data collected from the mines 29
Table 3.4 Example of daily tonnage record from Mine M l 30
Table 3.5 Example of cumulative monthly tonnage drawn from Mine M l 31
Table 3.6 Example of Height of Draw (HOD) calculation 31
Table 3.7 Example of draw point status extracted from database of mine M l 32
Table 3.8 Convergence measurements taken along production drift 13 from mine M l 33
Table 3.9 Example of monthly hang up records from mine M 2 35
Table 3.10 Example of monthly tonnage from mine M 2 35
Table 3.11 Calculation of hang ups/ton based on hang up and tonnage records 35
Table 3.12 Counting failures as a function of the daily draw point status records 48
Table 3.13 Counting process of failures for draw point D54 49
Table 3.14 Estimation of Draw Point Rate of Occurrence of Failure 50
Table 3.15 Average draw point ROCOF for Mine M l 53
Table 4.1 Draw point reliabilities to compute the entries 73
Table 4.2 Intermediate entry reliability table, R {i,j)
e 74

Table 4.3 Estimation of crosscut production capacity based on reliability estimates 83

Table 4.4 Production profile at 85% reliability 86
Table 5.1 Comparison of actual versus computed reliability 100

Table B . l Draw point reliabilities to compute the entries R (i, j)

e 126

Table B.2 Intermediate entry reliabilities R {i,j)

e 126

v
 LIST OF FIGURES
Figure 1.1 Chuquicamata mine, Codelco - Chile (Flores et al, 2004) 2
Figure 1.2 Production profile of eight operating block cave mines 3
Figure 1.3 Actual versus forecast production of an existing operation 5
Figure 1.4 One month of tonnage reconciliation per draw point 5
Figure 2.1 Mechanized panel caving at Henderson mine (Doepken, 1982) 10
Figure 2.2 Layout of the panel cave mining system (Flores, 2004) 12
Figure 2.3 Production level geometry (Moss et al, 2004) 13
Figure 2.4 Undercutting method used in block caving (Barraza and Crorkan, 2000) 14
Figure 2.5 Fundamental models that affect the planning parameters of a block cave mine (Rubio
et al, 2004) 17
Figure 2.6 Typical production schedule from an operating mine 21
Figure 2.7 Current mine planning process in block cave mining (Rubio et al, 2004) 23
Figure 3.1 Draw point blockage in block caving (Barlett, 2000) 34
Figure 3.2 Oversize at a draw point of mine M2 37
Figure 3.3 Hang up frequency as a function of draw point maturity 39
Figure 3.4 Draw point monthly productivity as a function of hang up frequency 40
Figure 3.5 Illustration of the angle of draw in a draw profile along a production drift 41
Figure 3.6 Effect of angle of draw on the normalized deviatoric stress experienced at the cave
front (Rubio et al, 2004) 42
Figure 3.7 Drift collapse due to a shallow angle of draw 43
Figure 3.8 Drift collapse due to a steep angle of draw 43
Figure 3.9 Drift collapse due to a sudden change on the angle of draw 44
Figure 3.10 Draw point minor apex pillar deformation as a result of uneven draw at mine M4.. 46
Figure 3.11 Total crosscut deformation as a function of the draw cycle 47
Figure 3.12 Operational records of draw point status data 48
Figure 3.13 Cumulative number of failures for a single draw point 50
Figure 3.14 Experimental draw point ROCOF for draw point D54 52
Figure 3.15 Draw point ROCOF curve for mine M l 54
Figure 3.16 Bathtub curve for a mechanical component (Hoyland and Rausand, 1994) 55
Figure 3.17 Draw point ROCOF curve for different production areas of mine M l 57

vi
Figure 3.18 Draw point ROCOF curve comparison between two operating block cave mines... 58
Figure 3.19 Average monthly draw point availability from mine M l 62
Figure 3.20 Comparison of draw point reliability versus draw point availability 62
Figure 3.21 Effect of hang up frequency on draw point reliability 63
Figure 3.22 Effect of convergence on draw point reliability ; 64
Figure 4.1 Components of a traditional mining system (Kazakidis and Scoble, 2002) 66
Figure 4.2 System reliability based on the reliability of the components r ,r ,r
Y 2 3 (Hoyland and

Rausand, 1994) : 67
Figure 4.3 Comparison between series parallel and parallel series for 3 components with 100%
component redundancy 68
Figure 4.4 Selection of 10 out of 15 components to compare with the performance of 10
components connected in series 70
Figure 4.5 Reliability comparison between a series and a k-out-of-n system 70
Figure 4.6 Reliability block diagram of a block cave production system 71
Figure 4.7 Redundancy versus system reliability 75
Figure 4.8 Impact of draw point selection on crosscut reliability 76
Figure 4.9 Draw point yield as a function of height of draw 77
Figure 4.10 Draw cycle time as a function ofk 78
Figure 4.11 Expected evolution of actual tonnage distribution 79
Figure 4.12 Proposed mine planning model including reliability model 80
Figure 4.13 Scheme to compute the reliability of a given production schedule 82
Figure 4.14 Assessment of reliability based on a computed production schedule 84
Figure 4.15 Production schedules including reliability as a target 85
Figure 4.16 Production schedules with integrated reliability 86
Figure 4.17 Effect of redundancy at the draw point and crosscut level 87
Figure 4.18 Sensitivity of the reliability model to the ROCOF curve 88
Figure 4.19 Sensitivity to crosscut productivity 88
Figure 4.20 Assessing the effect of overdrawn draw points in the production schedule 90
Figure 4.21 Reliability model used to prioritize secondary blasting 91
Figure 4.22 Evolution of the production capacity distribution throughout a production schedule
92

vii
Figure 4.23 Stochastic definition of production capacity of a block cave mine 93
Figure 4.24 Expected tonnage computed using the reliability model and the fudge factor
approach 94
Figure 5.1 Data flow to compute the reliability of the original production schedules of mines M l
and M 2 96
Figure 5.2 Calibration of computed reliability using a constant crosscut production capacity.... 97
Figure 5.3 Calibration of computed reliability using variable crosscut production capacity across
the active layout 98
Figure 5.4 Dataflow used to estimate actual reliability of a 60 days period based on the tonnage
distribution curve 99
Figure 5.5 Actual reliability as a function of actual tonnage distribution for a period of 60 days 99
Figure 5.6 Reliability model calibration using mine M l data set 100
Figure 5.7 Reliability model calibration using mine M 2 data set 101
Figure 5.8 Comparison between expected tonnage versus actual tons mined from mine M2.... 102

Figure A . l The reconciliation model 121

Figure B . l Intermediate reliability calculation 125
Figure C . l Process flow for the proposed scheduler including reliability 127
Figure C.2 Draw point list used in the reliability model 128
Figure C.3 Status matrix used in the reliability model 128
Figure C.4 Tonnage matrix used in the reliability model 129
Figure C.5 Draw point yield curve used in the reliability model 129
Figure C.6 Draw point ROCOF curve used in the reliability model 130
Figure C.7 Draw point database used as part of the reliability model 130
Figure C.8 Tonnage targets per period used in the reliability model 131

viii
 ACKNOWLEDGMENTS
I would like to express my gratitude to The Universidad de Chile Departamento de Ingenieria de
Minas, the institution that made possible my M A S c and PhD studies at UBC:. In particular I
would like to thank the individuals that constantly supported my graduate studies in Canada:
Francisco Brieva, Bruno Behn, Aldo Casali, Jaime Chacon and Hans Goefert.

M y research had been aided financially by Codelco - Chile through its support to the mining
technology group at the Universidad de Chile. I would like to express my gratitude to Juan
Enrique Morales, Fernando Geister and Octavio Araneda for enabling my research.

Faculties at The University of British Columbia, Mining Engineering Department who were
always willing to support my research and to contribute to the concepts presented in this
dissertation. M y supervisor Dr. Scott Dunbar for the amazing way of keeping the balance
between science and engineering application on his continuous contributions made to this
research. Also I would like to acknowledge the rest of the committee members Dr. Malcolm
Scoble, Dr. Tony Diering and Dr. Rob Hall for their constant effort to contribute on their best to
the success of the research presented in this dissertation. To my colleagues graduate students at
U B C for their patience to listen many times about block cave production planning and reliability
theory.

To Gemcom software international for giving me the opportunity to collaborate in engineering

several block cave mines operating around the world. In particular I would like to acknowledge
Dr. Tony Diering not only on his quality as a committee member but also as boss, leader and
friend. Many of the concepts expressed in this dissertation have been derived out of discussions
and work that has been done in the last five years as part of the Gemcom team.

There are many mining companies and especially engineers that have contributed with data,
expertise and discussion over the years to make this research applicable to the actual stage of
block cave mining. PT Freeport Indonesia personnel: Eddy Samosir, Rudy Prasetyo, Cahya
Kurniawan, Chuck Brannon, Dave Flient, Widyo Yudanto, Daulat Napitupulu, Eman Widijanto,
Fourmarch Sinaga, Tad Szwedzicki, Husni Sahupala. Palabora Mining Company: Sheperd Dube,

ix
Sergei Diachenko, Sam Ngidi, Lessane Sennanye, Lammie Nienaber, Jaggard Russell and Matt
Gili. Rio Tinto technical services: Allan Moss and Mark Howson. Codelco Chile: Mario
Vickuna, Augusto Aguayo, Gabriel Valenzuela, Patricio Vergara, Jorge Baraqui, Francisco De
La Huerta, Mauricio Barraza, Rigoberto Munoz. I would like to acknowledge specially to my
good friends Mauricio Melendez and Alejandro Moyano for being always willing to collaborate
and discuss mining concepts and ideas. Mauricio has contributed in a great deal specially to the
concept of angle of draw and damage in block cave mines. Alejandro Moyano has shown me
what a real mining engineer is by always bringing applied concepts and solutions to the
operation and planning of a mine.

I would like to specially thank my parents for their constant encouragement to finish my studies
at U B C . Dad your help has been invaluable in the many trips you made to Vancouver to assist
me keeping the balance between my work and family. Finally I would like to thank my children
Natalia and Matias for giving me the energy to continue and ultimately finish my dissertation.

This dissertation is dedicated to my outstanding wife Alex for her awesome way of expressing
her love and support to me over the years. Many of the accomplishments on my career have been
the product of having you as a wife with your understanding and patience over the many trips in
which you had to stay by yourself taking care of everything especially our children. So I dedicate
this work to you.

x
 s

1 INTRODUCTION
Mine planning consists of defining the source, destination and timing of extraction of every
single unit of mineral resource during the life of the mine. The mine plan plays a significant role
in linking the strategic objectives of a mining company to the operation of the mine. The
production schedule is most likely one of the most important components of the production plan
since it defines the tonnages and grades to be mined throughout the life of the mine. This
document is often used by bankers, share holders and other stake holders to assess the potential
benefits of a mining venture. In consequence, the production schedule delineates the business
promise of a mining company. If the production schedule of a company is weak, the economics
of this venture will be rather questionable. Also, if the strategic objectives are not reflected in the
day to day operational performance, it is because the mine plan has been misleading.

Several operations around the world are looking to apply low cost and highly productive mining
methods that could work in a low grade, competent rock mass and deep ore bodies. In particular
several open pit mines around the world such as Chuquicamata mine in Chile shown in Figure
1.1 Chuquicamata mine, Codelco - Chile (Flores et al, 2004) are looking to postpone the closure
of the mine by introducing underground mining methods. In order to maximize the utilization of
existing infrastructure such as power and production plants, the chosen methods need to be
highly productive and cost effective. Thus massive underground mining methods will most likely
be used. One such method, block caving, has gained popularity in recent years due to its
productivity and economic characteristics. Nevertheless, very little research has been conducted
on geotechnical principles which dictate how the mining system works. Thus there is a
considerable amount of uncertainty related to the way how a rock mass behaves in a caving
environment and this leads to uncertainty in planning block cave mines.

1
 Figure 1.1 Chuquicamata mine, Codelco - Chile (Flores et al, 2004)

Many definitions of block caving can be found in the literature, however one of the most
conceptual as well as practical is given by Laubscher (1994) "cave mining refers to all mining
operations in which the ore body caves naturally after undercutting its base. The caved material
is recovered using draw points". One of the main conclusions that one can make from this
definition is that block caving relies mainly on the interaction between the rock mass and the
stresses induced by the cave propagation to surface and on the movement of large quantities of
rock.

Since the knowledge of rock mass behaviour such as caveability, fragmentation, stresses, flow
mechanisms, in block caving is limited, the planning methods used in block caving should
account for the uncertainty in the geotechnical behaviour of the rock mass. However the current
methods used for scheduling block cave mines are rather deterministic and do not include such
uncertainty. This leads to production schedules that do not account for operational upsets
triggered by geotechnical events. In consequence, the current scheduling methods often compute
production targets that are too optimistic and which are difficult or impossible to fulfill during
the operation of the mine. The resulting production schedule computed as part of the production
plan often differs from the actual tonnage mined. Figure 1.2 shows a chart of actual production
from five different mines during an eight to twelve year period. This chart has been constructed

2
as part of the research presented in this dissertation, to show that there is a wide range of possible
production profiles for a given deposit during the life of the mine, depending on the rock mass
properties, size of the footprint, ore body geometry, stress behaviour, availability of capital and
ultimately the strategic objective of the company.

1.3M-,

Month

Figure 1.2 Production profile of eight operating block cave mines

The research presented in this dissertation aims to develop a methodology and a tool to account
the effect of uncertain geotechnical factors in production scheduling by introducing the reliability
of mining infrastructure. This reliability is meant to subsume all or a considerable part of the
geotechnical events that trigger operational upsets on tunnels, ore passes and draw points.
Incorporating the reliability of mining infrastructure in production planning will lead to an
estimation of the system reliability, which represents a measure of the confidence embedded in
the production schedule. Furthermore redundancy will be added to the mining system in order to
plan the amount of resources needed to achieve a certain level of reliability for a given
production plan. Finally, system reliability together with the production schedule redefines the
concept of production capacity of the mine reflecting the underlying uncertainty of geotechnical
events on mining infrastructure.

3
1.1 Statement of the Problem

Block caving has gained increased popularity in recent years due to its ability to produce large
tonnages at low operating cost. However, there are several issues that add considerable
uncertainty to the mining method such as: caveability in competent and highly stressed ore
bodies (De Nicola and Fishwick, 2000); seismicity due the presence of high stresses that could
adversely affect the mining method (Dunlop and Gaete, 1995); stress redistribution due to a
particular draw strategy (Rubio et al, 2004); ultimate rock mass fragmentation that may have
been poorly estimated (Hustrulid, 2000); lack of precision in estimating the grade distribution
within the ore body (Aguayo et al, 2004); and dilution or the manner in which waste is included
in the caved rock mass as it moves toward the draw points (Dolipas, 2000). Inadequate
recognition and understanding of these issues may lead to disruption of production performance.
The ability to integrate the above mentioned issues in production planning plays a significant
role in the success of a block caving operation. The way in which the uncertainty of the block
caving components is treated within the planning stage is crucial to estimate the production
targets that will drive the operation to be a great success or a catastrophic failure (Carew, 1992).
Figure 1.3 shows an operating mine performing with a high compliance between the original
production schedule and the actual production using traditional scheduling methods.

However, the difference between forecast and actual production of a particular draw point
reveals a different performance. For the same mining operation shown in Figure 1.3, Figure 1.4
compares forecast and actual tonnage at individual draw points within an active production area
during one month of production. Note that even though the total production forecast has been
achieved, an even distribution of tonnage per period across the active area has not been achieved.
This tonnage variance induces two well known operational draw point performances: under-
pulling and over-pulling. Under-pulling means that the actual tonnage is less than the production
forecast and over-pulling means that actual production has exceeded the forecast production.

4
 1,200,000- -•—Actual Production
c
o -o — Forecast Production
£
1/5 1,000,000-
c
o
c 800,000 -\
g
o
D
TJ 600,000 4
O

(D
C 400,000 H

o
c 200,000 H
3
DC

OH
i 1
i 1
i 1
i 1
i 1
i i 1
i • i
C M O J O J O J C M C M C O C O C O
o o o o o
d. d.
<0
S 5 5 5

Figure 1.3 Actual versus forecast production of an existing operation

Figure 1.4 One month of tonnage reconciliation per draw point

5
It is generally accepted that under-draw and over-draw behaviour leads to early dilution entry,
excessive induced stresses, and loss of planning abilities (Heslop and Laubscher, 1981). Usually
the simulated mine plans are based upon a production strategy that includes several draw rules,
which tend to draw down the caved ore as evenly as possible. If the production strategy is broken
because under or/and over draw performance, those draw rules would also be broken. Uneven
draw within an active production area creates zones of low or under draw which allow
compaction to occur within the fragmented ore overlying the production level. Frictional forces
are also induced at the boundaries of the under draw zones as a result of the differential draw
between the under and over drawn zones (Rubio et al, 2004). The compaction and the frictional
forces combine to produce high stresses in zones of low or under draw. The gradual compaction
of the under draw zones leads to a density gradient within the fragmented ore, which induces non
uniform movement of particles creating channeling and other oriented flow that may result in
early dilution entry.

There are many reasons for the under and over draw performance. Some authors have attributed
the variance between actual and plan to the "stealing ore" phenomenon which is known as the
operational tendency to extract more ore from productive draw points, near the ore pass, in order
to achieve the production target (Guest et al, 2000). This can be minimized by application of an
appropriate production control system such as underground dispatch (Prasetyo et al, 2004).
However, if the production forecast for a given draw point exceeds its natural capacity, the draw
point will be unable to meet the forecast production and will not be used or will be under-drawn.
To meet the overall production target, a few draw points will have to be over-drawn to
compensate.

Currently production targets are the result of production schedules computed with mine planning
parameters that do not evolve as a function of the operational performance and are not linked to
fundamental models that describe the geotechnical behaviour of the rock mass. These issues lead
to production schedules that do not reflect the actual rock mass behaviour within the mining
system, in particular, they do not incorporate the operational upsets triggered by geotechnical
events that affect the availability of mining infrastructure. Infrastructure availability is directly
related to the production capacity of the mine. Thus a production scheduling methodology that

6
does not integrate the geotechnical behaviour of the rock mass within the block cave mining
system will lead to optimistic production schedules. Optimistic production schedules would
usually force the use of more resources than planned jeopardizing the original value of the mine.
For example, when a production plan is computed, all draw points have the same chance of being
part of the schedule. However the likelihood of a draw point being available when a plan is in
operation varies across the layout due to rock mass properties and production performance.
Consequently there will be draw points that tend to achieve their production target more easily
than others. Therefore the treatment of draw points within the production scheduling algorithm
should reflect the rock mass variability across the layout.

In summary a robust production planning tool should incorporate actual mine behaviour and
make full use of the production data such as production records, infrastructure status records,
stress indicators, fragmentation and geotechnical mappings of the rock mass. If such data can be
shown to reflect geotechnical behaviour, then the uncertainty of a production schedule induced
by geotechnical events would be reduced.

1.2 Research Question

The following research question drives this dissertation

Can empirical observations in block cave mines be used to represent geotechnical effects
on production performance and thus improve the performance of production planning?

Several numerical models for predicting block cave behavior are described in this dissertation.
The effective integration of such models in a dynamic production scheduling environment is
difficult due to uncertainties in the geotechnical models and, more fundamentally, due to the
inability of such models to capture true rock mass behaviour. It is proposed that draw point
reliability could subsume the most important geotechnical events that affect the production
performance of a draw point. Different block cave mines facing operational upsets due to
geotechnical factors have experienced a lack of productivity affecting their ability to meet the
production targets. The reliability model will be used to test the ability to reproduce the
production performance observed at these mines.

7
The integration of draw point reliability together with the mine planning parameters such as draw
cycle and draw point yield in a mine wide reliability model enables the insertion of infrastructure
redundancy of different components of the mining system such as drifts, production crosscuts,
draw points. The introduction of redundancy in the mining system leads to a set of planning
decisions that generate different production schedules with different reliabilities. The redundancy
approach to production planning will be tested using the proposed mine reliability model. It is
expected that a production schedule with a high reliability would tend to minimize the variance
between actual and forecast draw point production, since the production forecast would have
been computed as a function of the true mining infrastructure reliability.

As a result of this research new aspects related to production planning are defined such as system
redundancy, which becomes a planning variable, production schedule reliability, which becomes
the main outcome of the research presented in this dissertation.

1.3 Contribution of the Thesis

The main contributions of this thesis are:

• The demonstration that the rate of occurrence of failure curve at a draw point can
subsume the most important geotechnical effects that lead to production loses. The life
cycle of a draw point is demonstrated to be characterized by a bathtub curve, similar to an
aging component in a mechanical system.
• The system reliability reflects the level of certainty of a given production schedule. It is
demonstrated that neither the draw point productivity nor the crosscut productivity affect
the reliability of a production schedule. It is rather the combination of draw point
productivity, crosscut productivity and infrastructure reliability which dictates the
certainty of a production schedule.
• The reliability of a production schedule is integrated with the components of a production
schedule such as draw point development sequence, draw rate and development rate.
These parameters combined with the system reliability redefine the production capacity
of a block cave mine. This definition includes, the probability of achieving the production
targets stated as part of the production schedule of the mine.

8
1.4 Organization of the Work

This dissertation is organized in six chapters. Chapter 1 has briefly introduced the research topic
and stated the research question and objective. Chapter 2 reviews the background of the block
cave mining system as well as planning methodologies used in block caving and other
underground mining methods. Chapter 3 reviews the geotechnical factors affecting production
performance, also introduces the concept of reliability as a function of the rate of occurrence of
failure of mining infrastructure. Chapter 3 also introduces the bathtub curve as a representation
of a the rate of occurrence of failure of a draw point during its life. Chapter 4 develops the model
in which the block cave mine infrastructure reliability is used to quantify the reliability of a
production schedule of a given mining system to achieve a given production target. Chapter 4
also reviews several applications of the model to planning problems and describes how the
model can be used to improve mine planning decisions. Chapter 5 presents several analyses in
which the reliability model is validated against observed production performance from two
operating mines. Finally, Chapter 6 summarizes the conclusions and recommendations made as a
result of this work.

9
2 LITERATURE REVIEW
Cave mining refers to all mining operations in which the ore body caves naturally after
undercutting and the caved material is recovered through draw points (Laubscher, 1994). In this
method the full block or an approximately equi-dimensional block is fully undercut to initiate
caving. The undercut is drilled and blasted progressively and some broken ore is withdrawn to
create a void into which initial caving of the overlying material can take place. As material is
extracted from draw points located on the production level, the caving propagates upwards
throughout the ore body until the overlying rock also caves and surface subsidence occurs.
Figure 2.1 shows a diagrammatic representation of panel caving at the Henderson molybdenum
mine in Colorado.

Figure 2.1 Mechanized panel caving at Henderson mine (Doepken, 1982)

The size of broken material will dictate which ore-handling system is suitable. For fine ore
fragmentation the full gravity system (Grizzly) is most suitable. For somewhat coarser material,
a slusher system should be implemented. For coarse material, the Load - Haul - Dump (LHD)
system might be the best option. Other parameters should be taken in the selection of an ore-
handling system such as work force sophistication, cost labor, availability and capital cost of
equipment and any other factors that may be unique to the particular mine (Tobie, 1982).

10
Historically, block caving has been used for massive, low strength, and low grade ore bodies
which produce fine fragmentation (Lewis and Clark, 1964). It is a low cost mining method which
is suitable for automation to emulate the concept of a "rock factory" (Tota, 1997). Nevertheless
block caving is highly capital intensive, requiring considerable investment in the development
and preparation of the mine before production begins. Nowadays there is a trend to use block
caving in competent rock masses which results in coarser fragmentation than the traditional
application of the method.

Peele (1941) distinguishes three forms of block caving:

a) Division of the footprint of the deposit into regular squared or rectangular blocks,
drawing evenly to maintain a horizontal plane of contact between broken ore and caved
waste cap.
b) Division of the horizontal area into panels retreating from one end of a panel to the other,
maintaining an inclined plane of contact between broken ore and caved waste cap.
c) No division of the horizontal area of the ore body into blocks or panels, retreat mining
from one wall to the other, maintaining an inclined plane of contact between broken ore
and caved waste cap.

Nowadays a block cave mining is sub-divided into block caving and panel caving. Block caving
is used for small size footprints that can be undercut in a short period of time compared to the
life of the mine. In this case an inclined plane is maintained while undercutting the ore body as
specified according to the second definition of the above list. On the other hand panel caving is
used mainly for large footprints in which undercutting is performed throughout 50-70 % of the
life of the mine. Operations using panel caving also keep an inclined plane to control dilution
entry. deWolf (1981) described a planning technique to control the position of the plane that
define the interface between ore and dilution as a function of the amount of draw. The author
introduced the concept of draw charts that forced the extraction to follow a given draw profile
across the active area.

11
 Figure 2.2 Layout of the panel cave mining system (Flores, 2004)

Figure 2.2 shows a panel cave mining system layout. The mine infrastructure may be composed
of five levels. In a vertical cross section from top to bottom they are: the undercut level, the
production level, the ventilation level, the crusher level and the haulage level. The undercut level
contains a set of parallel drifts running longitudinal or transverse to the ore body. The main
purpose of the undercut level is to provide access to create the initial blast that will induce the
caving of the overlaying rock mass. The production level contains a set of parallel drifts running
in the same orientation and direction as the undercut drifts, these drifts are connected to a rib
tunnel that is used as a main access to the mine. Along a production drift several draw points are
constructed at either side of the drift, at a certain spacing, to access the ore flowing from the
caving zone to the production level through a vertical funnel called draw bell. The draw bell
connects the undercut and production level. A schematic view of the production level and its
interaction with the undercut level is shown in Figure 2.3.

12
 Induced cracking

Figure 2.3 Production level geometry (Moss et al, 2004)

The ventilation level contains a set of parallel drifts running in the same direction and orientation
as the production drifts, along a ventilation drift several crosscuts are created to inject fresh air to
the production level and exhaust contaminated air from the production level using vertical raises
that connect the ventilation with the production level. The haulage level contains a set of tunnels
that connect with the production level through ore passes. The main purpose of the haulage level
is to efficiently haul the ore produced to surface using underground trucks, trains, conveyor belts
or skips. When using conveyor belts or skips underground crushing would be required which is
done either on the crusher level or on the production level.

The mine layout described above is very general and there are variations of the layout which
depend on the size of the ore body and the ultimate productivity of the mining system. For
example there are operations that have neither ventilation nor haulage level. In this case the
ventilation is performed via the production level and the LHDs would dump directly to the
crushers located outside of the access drift surrounding the ore body without ore passes.

Based on the undercutting sequence a block cave mine is classified into conventional
undercutting, advanced undercutting or pre-undercutting as illustrated in Figure 2.4. (Barraza and

13
Crorkan, 2000). The conventional undercutting method consists of blasting the undercut level
once the development and construction of the production level has been finalized. The advanced
undercutting method was introduced to reduce the exposure of the draw points to the abutment
stress zones induced as a result of the undercutting process. For the advanced undercutting
method just the production drifts are developed in advance to the blasting of the undercut. The
pre-undercutting method is such that no development or construction takes place on the
production level before the undercut has been blasted.

Conventional Undercutting

\ / 7 , ' '~ ."^-.^Undercutting

Undercut Level

Production Level

Advanced Undercutting

Undercutting
•t :\1S \lf-
j[ T T - - f h Undercut Level

Production Level

Pre Undercutting

Undercutting _ (

•• I Tfc Undercut Level

No Development Production Level

Figure 2.4 Undercutting method used in block caving (Barraza and Crorkan, 2000)

2.1 History of Block Caving

The first block caving operation recognized as such was the Pewabic mine, Menominee Range,
Michigan (Peele,1941). The method was crude and the only similarity to present day operations

14
was the ore handling. The production level was conditioned as a room and pillar mine in which
the pillars were reduced in size to induce the caving. Desired fragmentation was achieved after
six months of caving. Several other operations in Arizona used a modified Pewabic method such
as the Nowry mine, the Tobin mine and the Detroit Copper Co. The Humboldt mine was
probably the first block cave operation to separate the undercut level from the production level. It
is interesting to see that at this stage there was a clear understanding of the effects of block
confinement and the implication of confinement on the final production performance since the
block was separated from the hangingwall and footwall using drilling and blasting techniques.

The Miami mine in Arizona showed an interesting application of the gravity method that
included an undercut level, a grizzly level, a set of raises for ore handling purposes and a haulage
level. For a drawing strategy the Miami mine used a panel caving concept in which they would
undercut from one end to the other of an entire 120m x 140 m block. The angle of contact
between ore and waste was planned to be between 40 and 60 degrees from horizontal to
minimize dilution and over stress on the crown pillar.

Ore extraction in block caving has two related effects described by Peele (1941). More tonnage
and lower grade than expected is extracted due to dilution of ore from the waste cap. Draw
control tends to minimize dilution entry. Draw control concepts were first introduced by
McClennan (1930) at the Humboldt mine aiming to minimize the amount of dilution as part of
the ore extraction. The author summarized the application of draw control as shown below:
a) the ore should be drawn evenly so that the contact between the broken ore and the broken
waste cap is a horizontal plane and
b) regulate drawing to reduce induced stresses on production openings.

Since 1950 De Beers has used block cave methods for its operations in South Africa (Owen and
Guest, 1994). Not all methods introduced at De Beers operations were successfully initiated and
often plans had to be reviewed and modified according to the performance of the mine.
Mechanized panel caving was introduced at Premier mine in 1990 (Bartlett, 1992).

15
The Henderson mine was the first block cave operation to introduce fully mechanized
equipment. The mine started in 1976 using 4 yd LHDs. The design has evolved to utilize 7 yd
equipment (Rech et al, 2000) and the productivity of the mining system has increased from 136
tonnes/hour to 376 tonnes/hour. The L H D fleet size has been reduced from 30 to 7 while
maintaining the same production rate. Nowadays the Henderson mine also uses underground
haul trucks of 72 tonnes capacity.

The E l Teniente mine first used LHDs in the early 1980s (Chacon et al, 2004) and introduced a
novel way of designing the production level layout that is nowadays called the E l Teniente
layout. Stationary hammers (Moyano and Vienne, 1994) and different alternatives of caving,
such as advance undercut caving, were introduced at the Sub 6 mine part of the E l Teniente
complex (Rojas et al, 2000).

Even though the technology and the methods applied in block caving have evolved dramatically
over the years, concepts for mine planning have not followed the same path of evolution. In
recent years much attention has been given to understand the principles of gravity flow and rock
mechanics without considering mine planning as an important part of the mining system.

2.2 Fundamental Models of Block Caving

Block caving is a mining method that relies on natural processes for its success. Therefore more
detailed geotechnical investigations of the ore body are required than with other methods where
conventional drilling and blasting are employed as part of the production of the mine. The main
geotechnical parameters affecting the planning of the block cave are presented by Brown (2003)
as follows:
• Caveability
• Cave initiation
• Cave propagation
• Fragmentation
• Stress performance surrounding the cave boundary

16
The list of geotechnical parameters presented earlier supports the definition of several aspects
related to the planning of the mine such as undercut sequence, draw rate and development rate.
The ability to represent the variability of the geotechnical parameters throughout the ore body
would result in decreasing the risk of the mining method as well as increasing the ability to
forecast production (Summers, 2000).

A n illustrative representation of the link between the geotechnical and the mine planning
parameters of block caving has been proposed (Rubio et al, 2004) in order to understand the
influence of rock mass, stress regime and mining system in production performance. As shown
in Figure 2.5, the fundamental models of fragmentation, geomechanical, geological and
reconciliation are used to determine mine planning parameters, such as draw rate, undercut
sequence, development rate, tonnage, draw method and production targets.

Figure 2.5 Fundamental models that affect the planning parameters of a block cave mine
(Rubio et al, 2004)

The geomechanical model affects the following aspects of the design and planning of a block
cave mine:

• Draw point sequence will be affected by the structural pattern. Usually, the undercut
sequence will be oriented perpendicular to the major structures in order to produce blocks
that can enhance the caveability of the rock mass. (Rojas, 2000).

17
 • Abutment stress at the cave front will be a function of the pre mining stresses and the
angle of draw. This will affect the stability of the excavations located on the undercut,
production level and haulage level immediately below the front of the caving boundary
(McKinnon and Lorig, 1999).

• Seismicity is the response of the rock mass to the stresses developed at the cave back as
the cave propagates to surface and also the response of the rock mass surrounding the
excavations exposed to the abutment stress such as undercut, production, ventilation and
haulage drifts and rib tunnels. By measuring the properties of the seismic events an
estimation of the energy released can be performed to assess the stresses acting on those
events. This can be useful for forecasting caveability and other caving activity
related.(Glazer and Hepworth, 2004).
• Induced stresses due to uneven draw. B y performing uneven draw high stresses are
transferred to the zones of low draw due to the compaction of the broken rock overlying
the production level (Febrian et al, 2004). It has been shown that the draw performance
influences the distribution of stresses surrounding a draw point (Rubio et al, 2004).

The fragmentation model affects several aspects of the planning of a block cave mine, the most
important aspects are as follows:
• Dilution entry point which is the result of mixing of fragmented material along the draw
column (Heslop and Laubscher, 1981). Large variance of fragmentation along the draw
column leads to an increased amount of chaotic movement of rock within the broken rock
induced by the large density gradient. Gravity will tend to move finely fragmented
material towards the low density areas or coarse fragmented areas until equilibrium is
reached within the broken rock.

• Draw point spacing is the result of the draw column diameter which is believed to be a
function of the ultimate fragmentation of the draw column (Kvapil, 1965). The draw
point spacing is designed in such manner that the material overlying the production level
can be drawn without leaving static areas that may transfer stresses to the production
level infrastructure.
• Draw point secondary breakage activity is the result of the frequency of oversize
boulders, typically larger than 2 m that can not be handled by the L H D .
3

18
 • Hang up frequency is produced by boulders larger than the draw bell opening. Oversize
and hang up frequency will severely affect the productivity of the mining system has
been presented by Moss et al (2004) and Barraza and Crorkan (2000).

• Draw point yield is the maximum productivity of a draw point in the free flow state. As
the draw point matures the fragmentation becomes finer due to secondary fragmentation.
Therefore the void ratio decreases as the draw point matures leading to an increase in
LHD bucket capability, consequently achieving higher draw point productivity
(Esterhuizen et al, 2004).

The geological model links data relating to structure, lithology and mineralogy with the ultimate
metallurgical recovery. This model aims to build a geometallurgical model that can provide a
reasonable estimate of the metallurgical recovery based on the combination of the composite
lithologies. Didyk and Vasquez (1981) showed the effect of different rock types on metallurgical
recoveries at E l Salvador mine.

The reconciliation model captures the production performance of the mine. If this model is
available it is used to feedback key performance indicators to the fundamental models in order to
calibrate their behavior. The reconciliation model is also used to check the validity of different
assumptions made regarding to a production schedule. Thus this model will be used as a guide to
frame the production planning of the mine based on historical performance. The reconciliation
model affects the following aspects of the design and planning of a block cave mine:

• Draw rate is adjusted based on the historical production performance of draw points
located in a given rock mass domain. Usually the adjustment factor will be taken as a
form of draw point availability depending of its history of oversize and hang up
frequency, potential wet muck and potential instability for a given rock mass domain.
• Development rate is adjusted depending on the rock mass, stress regime in which the
construction will take place. Actual mine development performance is important since it
captures the bottlenecks in the mining system that affect the construction cycle of mining
infrastructure.

19
 • Draw strategy is compared against the historical performance of the mine. Based on the
production objective, the draw strategy will be modified in order to compute a schedule
that is realistic and consistent with the production promise.

It is important to note that the reconciliation model has not yet been introduced at the operating
mines, rather it has been conceptualized as the research unfolds over the years. Thus, at this time
there will be some mines applying parts of the reconciliation model as a result of informal
discussion or technical advice. A n interesting example of a reconciliation model was presented
by Kraushacer (1987) who used operational databases as a main source of operational
performance indicators for construction planning. The reconciliation model allowed the
calculation of different operational performance parameters that can be used as feedback for
future construction plans.

A considerable amount of numerical modeling has to take place in order to achieve a

comprehensive understanding of the interaction between the rock mass and the mining system
(Flores et al, 2004). The modeling is normally used to estimate parameters such as: stress
distribution at the front cave to decide upon the mining sequence; stress re-distribution on the
cave back to estimate ultimate fragmentation; fragmentation models to estimate draw point
productivity. Despite the amount of modeling performed at the feasibility stage of a block cave
mine, very few of the results are carried forward and integrated into the ongoing mine planning
activities.

2.3 Block Cave Production Scheduling

The main task of production planning is to define the production rates of a mining system. This
decision will ultimately define the value of the mining project. Scheduling the extraction and
rates of production for an underground mine operation is a common task carried out during the
life of the mine from feasibility through to the final production phase (Russel, 1987). In the case
of a block cave mine, the production schedule mainly defines the amount of tonnage to be mined
from the draw points in every period of the plan to achieve a given planning objective. The mine
plan also defines the number of new draw points that need to be constructed and their sequence
to sustain a given production target. To compute this production schedule many decisions need

20
to be made with regards to accessibility and infrastructure, mine and plant capacity and mining
sequence. Figure 2.6 shows an example of a mine plan in which the tonnage and grade are
predicted during the mine life.

18M-,

Figure 2.6 Typical production schedule from an operating mine

The planning parameters used to compute a production schedule of a block cave mine have been
presented by Rubio (2001) as follows:
• Development rate defines the maximum feasible number of draw points to be opened at
any given time within the scheduled horizon. This constraint is usually based on the
footprint geometry, the geotechnical behaviour of the rock mass and the existing
infrastructure of the mine, which will typically define available mining faces.
• Draw point construction sequence defines the order in which the draw points will be
constructed. This sequence is usually defined as a function of the undercut sequence. This
constraint usually acts on the draw point status activating those that are at the front of the
production face. Commonly the undercutting sequence is defined once the elevation and
horizontal dimension of the footprint has been decided.
• Maximum opened production area at any given time is an operational constraint that
depends on infrastructure and equipment availability as well as on ventilation resources.
A large number of active draw points might lead into serious operational problems such

21
 as excessive haulage distance and problems related to the movement of equipment within
the active draw points.
• Draw rate limits the production yield of a draw point at any given time within the
production schedule. The draw rate is a function of the fragmentation and the caveability
model. The draw rate needs to be fast enough to avoid compaction and slow enough to
avoid air gaps.
• Draw ratio defines a temporary relationship in tonnage between one draw point and its
neighbors. It is believed that this parameter controls the dilution entry point and the
damage of the production level due to induced stresses.
• Period constraints force the mining system to achieve the desired production target
usually keeping it within a range that allows flexibility for potential operational
variations.

The planning parameters shown in Figure 2.7 perform as operational constraints in the planning
stage of a mine. These planning parameters should be based upon the fundamental models
described in Section 2.2 of this dissertation. However, the link between the fundamental models,
particularly the geomechanical model, and the planning parameters has not been formalized or
documented. For example it is often found, in practice, that draw rate is exaggerated to achieve
the desired production targets, which disregards the relationship between draw rate and rock
mass fragmentation.

22
 Development
Rate

Figure 2.7 Current mine planning process in block cave mining (Rubio et al, 2004)

Several methods are currently used to compute production schedules in a block cave mine. They
can be classified in two main categories heuristic methods and operations research methods. The
original heuristic methods were the manual draw charts used at the beginning of block caving.
These methods evolved to the form presented by deWolf (1981) at Henderson where a way to
avoid early dilution entry was described by constraining the draw profile to an angle of draw of
45 degrees. A significant breakthrough was presented by Heslop and Laubscher(1981) who
described a volumetric algorithm to simulate the mixing along the draw column. Carew (1992)
described the use of a commercial package called PC-BC to compute production schedules at
Cassiar mine. Diering (2000) showed the principles behind the commercial tool PC-BC to
compute production schedules, providing several case studies where different draw methods
have been applied depending on the ore body geometry and rock mass behaviour. A n alternative
model was also presented by Kear (2000), in which a floating surface method is used to find the
optimal draw strategy for the Palabora mine using an iterative algorithm.

The application of operations research methods to the planning of a block cave mine was first
described by Riddle (1976). This development intended to compute mining reserves and define
the economic extent of the footprint. The final algorithm did not reflect the operational
constraints of block caving described above since it worked with the block model directly instead
of defining the concept of draw column as an individual entity of the optimization process. Other

23
authors such as Caccetta and Giannini (1988), Wilke et al (1984), and Gershon (1987) have
attempted to develop methodologies to optimize production schedules, but none of them has
satisfactorily produced a robust technique which has an acceptable level of success in block
caving. One of the main reasons for this unsuccessful history has been the difficulty of definning
a multi-period objective function over different mine planning horizons that integrates the
gravity flow mechanism present in block caving. Chanda (1990) developed a model to optimize
production from a slusher block cave method using scrapers as production machines. Chanda
concentrated on a short term planning problems that cover a time horizon of a few weeks to a
few months applying single step optimization rather than multi period optimization.

More sophisticated algorithms have been developed by Guest (2000) to analyze and compute
long term plans. Guest postulates that by following a set of surfaces that conceptually define a
draw control strategy, dilution can be minimized and therefore net present value (NPV) can be
maximized. This algorithm does not integrate the concept of gravity flow as part of the
optimization process. Also Guest assumes that a draw strategy can be summarized by the
definition of a set of surfaces which may be a very difficult task to perform in metal mining due
to grade variability. It is interesting to note that Guest stated the importance of the draw strategy
on dilution control as part of the production scheduling process. Rahal et al (2003) used a dual
objective mixed integer linear programming algorithm to minimize the deviation between the
actual state of extraction (height of draw) and a set of surfaces that tend towards a defined draw
strategy. This algorithm assumes that the optimal draw strategy is known. Nevertheless it is
postulated that by minimizing the deviation to the draw target the disturbances produced by
uneven draw can be mitigated. Diering (2004) presented a non linear optimization method to
minimize the deviation between a current draw profile and the target defined by the mine
planner. Diering emphasizes that this algorithm could also be used to link the short with the long
term plan. The long term plan is represented by a set of surfaces that are used as a target to be
achieved based on the current extraction profile when running the short term plans. Rubio et al
(2004) presented an integer programming algorithm and an iterative algorithm to optimize long
term schedules in block caving integrating the fluctuation of metal prices in time.

The main problems associated with the methods presented above can be summarized as follows:

24
 • The methods do not incorporate the variability and the dynamic behaviour of the
fundamental models throughout the ore body.
• The methods do not have a rational way to link the mine planning parameters with the
fundamental models.
• The methods do not integrate the operational upsets that affect productivity. Therefore
the current systems do not adapt as a function of the historical performance.
• The methods do not incorporate, in a routine basis, operational performance to adjust the
medium and the long term plans.

The issues with the current production scheduling methods listed above reflect that the current
methods do not integrate the technical uncertainty inherent to the mining method. Thus, the
production targets computed with the existing methods lead to unrealistic production targets,
forcing the operation of the mine to break the rules integrated in the original production
schedules. This motivates the development of a methodology that could integrate operational
performance of the mine, as a reflection of the rock mass behaviour to the mining system, into
production planning to compute realistic production targets per draw point during the life of the
mine.

2.4 Uncertainty in Block Cave Production Scheduling

The lack of a formal link between the fundamental models and the planning parameters leads to a
considerable amount of uncertainty in the planning process. Summers (2000) described the main
sources of uncertainty in block cave mining pointing out that there are no clear methodologies to
incorporate the natural variability of the rock mass in the process of design and planning the
mining system.

The treatment of uncertainty in production planning as generally being discussed by several

authors such as Samis and Poulin (1997), who proposed the insertion of contingency plans
associated with parameters that have high level of uncertainty in the long term production
strategy that could aim to avoid production shortages and/ or an increase of production costs.
Singh and Skibniewski (1991), Kajner and Sparks (1992) have also looked at the flexibility
needed in mineral resource industry as a function of the level of uncertainty. Commonly

25
simulation of the mining system has been the main tool used to assess the amount of flexibility
needed in a mine design or a mine plan. The main problem with this approach is that often
simulation models do not integrate the fundamental models such as stress distribution,
caveability and gravity flow. In other mining methods in which the production outcomes are
controlled by drilling and blasting activities discrete event simulation has been used to plan
tactical decisions (Dessureault et al, 2000).

Flexibility or the ability to deal with changes and upsets has often been proposed as a response to
uncertainty in mine planning. Real options have been used to estimate the value of flexibility
(Dunbar et al., 1998; Trigeorgis, 1998). Kazakidis (2003) proposed a flexibility index that uses
simulation to quantify the value of the operational flexibility. However real option concepts do
not provide a methodology to quantify the amount of flexibility needed in a production schedule
nor do they integrate operational performance observations to dynamically adjust the production
schedule.

There are several methods developed to quantify the impact of uncertainty on the financial
valuation of the mine. Often Monte Carlo simulations have been used to quantify the risk related
to metal price uncertainty. Smith (1999) and Dimitrakopoulos (2002) have developed methods to
quantify uncertainty related to ore body modeling and its impact on the production schedule for
open pit mines.

The methods summarized above concentrates mainly in uncertainty derived from metal prices
and grades. These methods do not appear to be the correct approach for quantifying the system
uncertainty involved in block caving. Xiaotian (1989) introduced the concept of caving
parameters to describe the process of caving. The caving parameters have associated probability
distribution to affect different mine planning parameters such as sequence and undercutting rate.
The mine plan is computed using simulation integrating the probability distribution of the caving
parameters. This paper does not discuss how the distributions are constructed and whether or not
those distributions would be stationary.

26
Krantz and Scott (1992) recognized that planning a mine is a dynamic rather than a static
activity. One of the main mine planning activities should be the comparison of actual vs. planned
production and use this comparison to modify the initial assumptions. The mine needs to be built
with enough flexibility to integrate the potential improvements resulting from this comparison.
Thus success of the plan will depend upon the ability of the mine to react to the current
operational situation.

More than 12 operating block caves and projects have been visited during the time that this
research has taken place and none of them reports the production forecasts in a form of a range,
band or confidence interval to represent the technical uncertainty of these forecasts. The
uncertainty of a production schedule is missing in the process of planning a block cave mine. If
the uncertainty could be quantified, different decisions could be made in order to integrate
flexibility in the mining system that could mitigate the risk caused by the underlying uncertainty.

The uncertainty related to rock mass behaviour within the mining system leads to unplanned
operational upsets such as infrastructure failure that tend to jeopardize the original production
estimates. Thus a natural approach would be to incorporate the infrastructure failure
characteristics of a mining component as part of the planning process. Even though this approach
does not integrate a model of rock mass behaviour within the planning process, it does include
the effects of that behaviour in mining infrastructure. This is an empirical approach that will be
reviewed in the following chapter.

27
3 FAILURE BEHAVIOUR OF MINING INFRASTRUCTURE
Mining infrastructure may fail due to geotechnical events that tend to reduce its physical
availability. In particular, there are two geotechnical events, coarse fragmentation and over
stress, that affect the productivity of block cave mining infrastructure. Coarse fragmentation at
the draw point leads to the formation of hang ups and oversize, reducing the effective time that
the draw points have available to achieve a given production target (Barraza et al, 2000). Over
stress on the crown pillar leads to closure that often ends in the collapse of a production area
(Rubio et al, 2004).

In this section the characteristics of infrastructure failure as a result of these two geotechnical
events will be studied. Then a methodology will be devised to quantify the reliability of draw
points as a function of the rate of occurrence of failure. Finally, it will be shown that draw point
reliability correlates well with the underlying geotechnical events that triggered mining
infrastructure failure.

3.1 Operational Database

A n important component of the research was the collection of different pieces of information
recorded at different mines in order to be able to demonstrate the fundamental concepts
presented in this dissertation. A summary of the geotechnical information recorded at four mine
operations, labeled M l to M 4 to maintain confidentiality, is shown in Table 3.1. These four
operations represent 80% of the worldwide block cave underground mine production.

Table 3.1 Geotechnical information of the mine operations

M1 M2 M3 M4
RMR (Laubscher, 1989) 45 80 70 65
Fracture frequency per meter - 0.1-0.8 1-2 3-4
Fragmentation 10%>2m 3
45% >2m 3
40% >2m 3
15%>2m 3

The geotechnical information presented above was taken from technical reports found at the
operations. It may be seen from Table 3.1 that the four mines are operating in different rock mass
environments. For example M 2 and M3 are challenging the traditional application of block

28
caving based on their highly competent rock mass environment while M l and M 4 are operating
in a rock mass environment that is considered to be ideal for block caving.

The main features of the mining systems at these mines are shown in Table 3.2. The number of
draw points and crosscuts constructed at these mines is an indication of the scheduling
complexity required to coordinate all the production units in order to meet the production target.

Table 3.2 Features of the mining system

M1 M2 M3 M4
Advanced Advanced Pre Traditional
Undercutting Method
Undercutting Undercutting Undercutting Undercutting
Panel Block Panel Panel
Mining Method
Caving Caving Caving Caving
Depth (m) 850 1200 500 500
Production (tons/day) 38,000 30,000 32,000 23,000
Draw points 300 320 250 365
Production Crosscuts 14 20 26 20

The operational and planning data used to illustrate the concepts exposed in dissertation are
summarized in Table 3.3.

Table 3.3 Operational and planning data collected from the mines

M1 M2 M3 M4
Amount of data collected 4 years 3 years 3 years 2 years
Tonnage records X X X X
Status records X X X
Convergence records X X
Hang up records X X
Oversize records X X

The different components shown in Table 3.3 are explained in the following sections of the
dissertation.

3.1.1 Daily tonnage records

Tonnage records correspond to the daily records of tonnage drawn from every active draw point
across the mine layout. Usually this data is collected by production systems such as Underground

29
Dispatch (Barraza et al, 2004) which operate interactively with the L H D operator to record the
number of buckets drawn from a given draw point at a given time. The number of buckets is
converted to tonnage based on the weight reported by the mill. A n example of daily tonnage
1

records is shown below in Table 3.4.

Table 3.4 Example of daily tonnage record from Mine M l

Date Draw Point Name Tonnage (tons)
5/1/2003 D1 108.8
5/1/2003 D2 97.4
5/1/2003 D3 97.4
5/1/2003 D4 129.9
5/1/2003 D5 97.4
5/1/2003 D6 120.2
5/1/2003 D7 175.5
5/1/2003 D8 129.9
5/1/2003 D9 129.9
5/1/2003 D10 129.9
5/1/2003 D11 108.8
5/1/2003 D12 107.1
5/1/2003 D13 107.1
5/1/2003 D14 129.9
5/1/2003 D15 129.9
5/1/2003 D16 129.9
5/1/2003 D17 129.9
5/1/2003 D18 99.1
5/1/2003 D19 148.1

The daily tonnage records are used to compute the cumulative tonnage drawn per draw point per
month as shown in Table 3.5. This tonnage is commonly used to represent draw point maturity.
The cumulative tonnage drawn from a draw point can be converted to Height of Draw column
(HOD) at a given period by dividing the cumulative tonnage drawn by the in situ density and the
effective draw point area. For example using the cumulative tonnage shown in Table 3.5 the
HOD is computed in Table 3.6 using a draw point area of 280 m and an in situ density of 2.7
tonnes/m (values associated with mine M l ) . HOD will be used extensively to represent draw
point aging.

' The actual trend is to install weightometers on the LHD buckets to measure the tonnage drawn from a draw point
in real time. However, this system has not yet been delivered to operations since the weightometers are still too
sensitive to the oscillations of the hydraulic system of the LHD (van Hout et al, 2004).

30
 Table 3.5 Example of cumulative monthly tonnage drawn from Mine M l

Draw Point Name Apr,01 May,01 Jun,01 Jul,01 Aug,01 Sep,01

D1 10,182 10,843 13,132 13,969 14,918 16,346
D2 11,481 12,181 13,660 14,472 15,625 17,736
D3 16,482 17,067 18,576 19,094 21,766 23,236
D4 15,379 16,255. 17,868 18,276 19,451 21,408
D5 12,512 13,581 15,155 17,204 20,291 22,529
D6 13,039 14,011 16,331 18,552 20,184 22,830
D7 16,355 18,149 21,251 23,188 27,404 32,340
D8 10,946 12,851 16,874 19,080 25,497 29,889
D9 13,143 16,245 22,551 25,207 30,900 35,248
D10 20,945 25,497 30,308 32,351 39,728 44,951
D11 20,612 33,594 36,418 40,236 47,161 51,266
D12 27,194 39,263 41,411 47,563 53,461 56,752
D13 1,969 2,167 4,330 11,042 21,422 30,813

Table 3.6 Example of Height of Draw (HOD) calculation

Draw Point Name Apr,01 May,01 Jun,01 Jul,01 Aug,01 Sep,01

D1 13 14 17 18 20 22
D2 15 16 18 19 21 23
D3 22 23 25 25 29 31
D4 20 22 24 24 26 28
D5 17 18 20 23 27 30
D6 17 19 22 25 27 30
D7 22 24 28 31 36 43
D8 14 17 22 25 34 40
D9 17 21 30 33 41 47
D10 28 34 40 43 53 59
D11 27 44 48 53 62 68
D12 36 52 55 63 71 75
D13 3 3 6 15 28 41

3.1.2 Daily status records

Status records correspond to information collected daily on regarding the physical status of the
draw points. Table 3.7 shows an example of the draw point status data collected from mine M l .
Currently there is a trend in the industry to implement the recording of draw point status as part
of the Underground Dispatch system so that every time a L H D is dispatched to a draw point the
L H D operator would have to communicate the initial and the final status of the draw point. Also
the secondary breakage equipment and crews would be dispatched by the Underground Dispatch
system in order to keep records of the draw points that have been blasted. Usually the
information recorded by the Underground Dispatch system is checked at the beginning of the day

31
as part of the production control tasks. If there is a difference between the information recorded
by the system and what is seen in the field, the status is corrected and imported manually into the
production database. The draw point status recorded at the draw points can be classified as
active, oversize, wet muck, hang up, and closure (high deformation). Currently there is one draw
point status recorded per day, although a drawpoint may experience different status conditions
within a day. For instance a draw point could change from active to oversize and return to active
status within a day. In this case the draw point would be recorded as active, thus ignoring the
draw point repair that took place during the day. More sophisticated systems will be needed to
record the changes of draw point status during the day in order to more accurately capture the
actual availability of a draw point.

Table 3.7 Example of draw point status extracted from database of mine M l
Draw Point\ Day 1 2 3 4 5 6 7 8 9 10 11 12 13 14
D1 A A A A A A A A A A/H A/H A/H A/H A/H
D2 B/H B/H B B B B B B B B B B B B/H
D3 A A A A A A A A A A A A A A
D4 A A A A A A A A A A A A A A
D5 A A A A A A A A A A A A A A
D6 A A A A A A A A A A A A A A
D7
D8 0 0
D9 0 0
D10 0 0
D11 A/H A/H A A A/H A/H A/H A/H A A/H A/H A/H A/H A/H
D12 C C C C C C C C C C C C C C
D13 A/H A/H A/H A A/H A/H A/H A A/H A A A/H A/H A
D14 A A A A A A A A A A A A A A

The draw point status conditions shown above are: active draw point (A), oversize draw point
(B, B/H), temporarily closed due to high convergence draw point (C) or hanged up draw
point(A/H).

3.1.3 Convergence records

Convergence records result from the monitoring of displacements at different locations along the
production drift using a tape extensomenter (Febrian et al, 2004). Table 3.8 shows an example of
convergence data collected from mine M l . The purpose of convergence monitoring is to provide

32
an early warning of excessive displacement of the back of drifts located on the undercut and
production levels. Another benefit of this monitoring is to provide information to assess the state
of the induced stresses and rock behavior during undercutting and production stages, providing
valuable information for mine planning and operations to maintain ground stability (Febrian et
al, 2004). Several convergence stations are installed along production and undercut drifts and are
monitored at least once per week. If a station showed high convergence, daily monitoring would
be performed in this area.

Table 3.8 Convergence measurements taken along production drift 13 from mine M l

Date Station Hzjncr(mm)

3/16/2000 P13C01 -0.30
3/16/2000 P13C02 0.05
3/16/2000 P13C03 0.20
3/16/2000 P13C04 -0.30
3/16/2000 P13C05 -0.60
3/16/2000 P13C06 0.95
3/22/2000 P14C01 0.50
3/22/2000 P14C02 0.25
3/22/2000 P14C03 0.10
3/22/2000 P14C04 -0.55
3/22/2000 P14C05 -0.10
3/22/2000 P14C07 -0.10
3/22/2000 P15C02 -0.85
3/22/2000 P15C03 0.10

3.1.4 Records of draw point hang up

Hang up records are extracted from the draw point status database. Figure 3.1 shows an
illustration of hang ups and oversizes experienced at draw points. Hang ups are stable arches
formed inside the draw bell as a result of the movement of a large boulder. Oversize is
considered to be any boulder at a draw point greater than 2m , which seems to be the maximum
3

size that a L H D bucket can load (Moss et al, 2004).

33
 Hang Ups

Oversize

Figure 3.1 Draw point blockage in block caving (Barlett, 2000)

The hang up status is of such importance that it is usually stored in a separate database in order
to keep track of the amount of explosives used to remove the hang up as well as the machines
used to drill and place the explosives. Given these records, a list of hang ups per draw point per
month could be constructed as shown in Table 3.9. The tonnage drawn per draw point in the
same period of time could be computed using the daily tonnage records. A n example of monthly
tonnages from mine M 2 is shown in Table 3.10. By combining the number of hang ups with the
tonnage drawn per period, it is possible to compute the hang up frequency, hang ups/ton, for a
given period of time as shown in Table 3.11. For the periods in which there is no tonnage drawn
this calculation is disregarded. This indicator is extremely important for the study of the effect of
hang ups on production performance.

34
 Table 3.9 Example of monthly hang up records from mine M2
Draw Point Name Jan,02 Feb,02 Mar,02 Apr,02 May,02 Jun,02 Jul,02
P1 7 4 4 1 6 7 13
P2 1 - - - 1 1 2
P3 1 2 1 - 10 10 8
P4 3 2 5 - 6 9 9
P5 4 2 2 - 4 7 5
P6 2 3 4 - 2 1 8
P7 1 5 5 - 2 11 3
P8 5 4 2 - 2 2 9
P9 2 2 - - - 4 7
P10 - - - - 1 3 4
P11 - - - - 5 6 8

Table 3.10 Example of monthly tonnage from mine M2

Draw Point Name Jan,02 Feb,02 Mar,02 Apr,02 May,02 Jun,02 Jul,02
P1 116 1,573 1,384 243 2,363 1,196 2,192
P2 379 45 455 117 1,505 1,141 887
P3 163 2,067 1,303 - 1,798 2,254 2,572
P4 2,423 1,842 1,263 88 2,626 2,529 2,660
P5 511 2,525 1,808 - 2,151 3,274 1,637
P6 1,091 2,229 1,798 - 1,283 2,391 789
P7 472 2,525 2,414 - 1,929 2,502 468
P8 2,539 1,195 1,475 - 1,060 1,766 97
P9 968 144 152 10 959 2,502 536
P10 1,780 162 1,313 126 2,212 3,495 419
P11 1,788 - - - 3,787 2,548 2,173

Table 3.11 Calculation of hang ups/ton based on hang up and tonnage records

Draw Point Name Jan,02 Feb,02 Mar,02 Apr,02 May,02 Jun,02 Jul,02
P1 0.0603 0.0025 0.0029 0.0041 0.0025 0.0059 0.0059
P2 0.0026 - - - 0.0007 0.0009 0.0023
P3 0.0062 0.0010 0.0008 0.0056 0.0044 0.0031
P4 0.0012 0.0011 0.0040 - 0.0023 0.0036 0.0034
P5 0.0078 0.0008 0.0011 0.0019 0.0021 0.0031
P6 0.0018 0.0013 0.0022 0.0016 0.0004 0.0101
P7 0.0021 0.0020 0.0021 0.0010 0.0044 0.0064
P8 0.0020 0.0033 0.0014 0.0019 0.0011 0.0924
P9 0.0021 0.0139 - - - 0.0016 0.0131
P10 - - - - 0.0005 0.0009 0.0095
P11 - 0.0013 0.0024 0.0037

35
3.1.5 Records of oversize draw points
Oversizes are treated in the same manner as hang ups, given their relevance to production
performance. By combining the number of oversizes with the tonnage drawn from a draw point,
it is possible to compute the oversize frequency, number of oversizes/ton, for a given period of
time.

3.1.6 Production schedules

The original production schedules are the original production estimates per draw point through
the life of the mine completed as part of the feasibility study of the mine. This information was
not found at the mines M l to M4. It was usually found that the mines had kept the original global
production targets per year or per quarter. However the detail per draw point was not available.
This information was reconstructed based on the production targets reported in the feasibility
study, draw point undercut sequence derived from the draw point status, draw rate reported in the
feasibility study, development rate from draw point status and mine design (draw point spacing)
from the feasibility study. The result was the tonnage planned per draw point per month during
the life of the mine.

It was observed that there was no standardized way of recording operational upsets, besides the
draw point status, that could be used for detailed back analysis or for general study of production
performance. Considering that block caving is a complex mining method, more attention should
be paid to capturing different operational and geotechnical situations that could assist mining
engineers and geologists to understand the phenomenon of caving and its production
implications. A proposed design for an operational database to be used in block cave mining is
presented in Appendix A .

3.2 Effect of Geotechnical Events on Production Performance

Production activities in a block cave mine are continuously interrupted due to infrastructure
damage as a result of geotechnical events. One of the most significant interruptions is the
secondary breakage activity to clear either a hang up or an oversize event at a draw point. Since
secondary breakage usually involves the use of a drill and explosives to remove a hang up or an
oversize, to clear one or several draw points along a crosscut, an entire production drift may be

36
closed to avoid hazards. Another important operational upset triggered by geotechnical events is
the damage of the production infrastructure due to cracking or rock sloughing, or the collapse of
the production drift due to induced stresses. The damage produced by induced stresses requires
repair that could cease production from that drift for a period ranging from a few weeks to
several months. It has also been found that usually after a few weeks without drawing from a
production drift, damage occurs, sometimes resulting in the collapse of the production drift. This
phenomenon is the result of rock mass compaction and subsequent stress transfer from the
material overlying the production level to the production crown pillar. The fragmentation and the
pillar convergence geotechnical events will be studied in the following sections, with a view to
understand their implications on production performance.

3.2.1 Effect of hang ups on production performance

The secondary breakage activities taking place in an underground mine constantly interact with
the production activities of the mine. In a block caving environment every production drift is
usually engaged in production, secondary breakage or rework at any given time. Secondary
breakage activity is often viewed as a bottle neck to reach production goals. Oversize events, as
shown in Figure 3.2, are usually cleared within a shift and with production resuming
immediately after blasting or breaking the boulder with mobile hammers.

Figure 3.2 Oversize at a draw point of mine M2

37
The situation is very different with hang ups which usually require a day or at least two shifts to
clear. At the moment there is no a single rule regarding how many draw points should hang up
until the status of a production crosscut is changed to secondary breakage. There are at least two
strategies to attack the problem: the first is to wait for a given number of hanged up draw points
in a crosscut, the second is to have a fixed secondary breakage sequence in which, regardless of
the number of hanged up draw points, the status of the crosscut is changed to secondary breakage
at a certain time. At the moment there is no clear evidence to say which strategy is more
productive. Most likely the strategy to follow would depend upon the changes on rock mass
fragmentation during the life of the mine.

Mine M 2 has experienced extremely coarse fragmentation as a result of the geotechnical

environment (high R M R ) in which block caving has been applied. Consequently the scheduling
of secondary breakage activities has been crucial to the productivity of the mine. The mine is
divided into three geotechnical domains: D l , D2, D3 which lead to different fragmentation
characteristics across the layout. Figure 3.3 shows the hang up frequency (hang ups/1000 tons) of
these three areas as a function of draw point maturity. The cumulative tonnage drawn from a
draw point converted to a height of draw column (HOD) is used as a proxy for draw point
maturity in this case.

Figure 3.4 shows that the hang up frequency decreases as the draw point matures for the three
geotechnical domains. This shows that material drawn from higher up in the draw column will
have more secondary fragmentation than material located at the bottom. The second interesting
observation from Figure 3.4 is that the hang up frequency tends to be fairly uniform across the
active layout once a particular value of HOD is achieved, about 100 m in this case. This will lead
to a totally different production scenario that could eventually facilitate expansion of the current
production rates. Given that the hang up frequency will become uniform, the strategy to treat
hang ups should evolve accordingly.

38
 •
12-,
o D1
O • D2
o
2 ioH
A D3
o O
o O
O
o
o
c A A
CO A ° • A O O
X 8
>, A
o
c
CD
D A
O
O" •
2 A
Q.
2H •
o>
c •
CO A
X
20
I
40
I
60
-r- —1—
100
— I —
120 140
— i
160
80
Height of Draw ( m)

Figure 3.3 Hang up frequency as a function of draw point maturity

The second analysis performed at mine M 2 was to correlate hang up frequency with draw point
productivity per month with the objective of illustrating the impact of this geotechnical event on
draw point productivity. Figure 3.4 shows the results of this analysis. There is little doubt that
productivity is inversely correlated to hang up frequency. The more hang ups a draw point
experiences in a given time period, the less time each draw point will be available for production.
Figure 3.4 shows that hang up frequency, as an indicator of geotechnical events, affects draw
point productivity. Consequently it could be expected that hang up frequency would affect the
productivity of the whole mining system.

It should be noted that sometimes the correlation between hang up frequency and draw point
productivity is positive. This can be explained by the fact that a draw point that is excessively
drawn (i.e., yields high production) leads to differential draw with its neighbors and facilitates
the development of a hang up in the draw bell.

39
 6000-,

c
o 50004 •
E
03
C
H 4000- •
•>
o 3000 4
T3
O

zr
c
2000-
o
Q. 1000-
CO

04
i i i i i i i I

1 10

Hang up frequency (Hang Ups /1000 tons)

Figure 3.4 Draw point monthly productivity as a function of hang up frequency

3.2.2 Effect of production performance on induced stresses

There are several operational aspects that could lead to stress redistribution across the active
layout. However a distinction should be made between stresses observed at the cave front and
stresses observed within the active production area. A discussion of the nature of these two stress
redistributions is given below.

Stress at the cave front

The main cause of over stresses at the cave front is related to the angle of draw which affects the
stress pattern at the cave front. The angle of draw is commonly measured in a vertical cross
section perpendicular to the mining sequence displaying the height of draw (HOD) of the draw
points as shown in Figure 3.5. A line is fitted to intersect the HOD of all draw points shown on
this section. Then the angle of draw is measured from the horizontal to the fitted line.

40
 Vertical Cross Section

Level

Draw Points

Figure 3.5 Illustration of the angle of draw in a draw profile along a production drift

Mine M3 had experienced a significant number of collapses at the cave front. As part of their
production database this mine has recorded the time and location of the collapses for the first five
years of operation. These data together with the production records were used to back-analyze
the angle of draw in association with the historical collapses experienced at this mine. Collapse
of the cave front was found to occur under three conditions:

• Shallow angle of draw: A shallow angle of draw would produce an open span that
exceeds the rock mass strength and thus the roof of the undercut level would collapse as a
result of the stress acting on the back of the undercut area.
• Steep angle of draw: As shown below a steep angle of draw induces a rotation of the
stress tensor as well as increases the deviatoric stress (difference between principal
stresses) experienced at the cave front (Rubio, et al, 2004).
• Sudden changes of the angle of draw between periods: This is explained below.

Figure 3.6 shows the effect of different angles of draw on the deviatoric stress which has been
normalized by the deviatoric stress experienced at the pre mining stage. For a steep angle of
draw the cave front will experience a significant increase in deviatoric stress with respect to the

41
pre mining stage. The change in deviatoric stress will induce shear stress which, depending on
the rock mass strength, could produce the collapse of the cave front. A collapse of the cave front
would affect the productivity of the mining system since draw points could not be
commissioned.

Angle of Draw
- o - 80° 4 2.0
-•-70°
-v-60° 4 1.8 in
<n
<D
-A-50°
-o-40° A 1.6
o
'*

A 1.4 o
>
<X>nv o
Q
1.2 T3
<D
N
^ J ^ P / / /° V
76
,O-0-O-n-DVV A P 1.0
E
D-D^V-V-VA-AA'' p o
O^^O-^A-A^O-OO O 0.8
o o-o
— I —
—
1— 0.6
-100 -80 -60 -40 -20

Distance from the cave front (m)

Figure 3.6 Effect of angle of draw on the normalized deviatoric stress experienced at the
cave front (Rubio et al, 2004)

To show that such collapses occur, the data available from Mine M3 was used to compute the
monthly angle of draw at each production crosscut for an approximate four year period. Then the
angle of draw was correlated with the collapses experienced at this mine. It was shown that
shallow or steep angles of draw induce collapses at the cave front leading to periods of "no
draw". Figure 3.7 shows the draw profile along a production drift in which the collapse of the
cave front was induced by a shallow angle of draw below 30 degrees.

42
 80-

70 H

°~ 60 - |

CO CO

2o 50
Shallow
Angle of Draw
CtXOOOCCO
5 40 O Collapse
ccP\ d\ /
30 H
cooco
20

3
OL
CO

<
CD
2 CO

Month

Figure 3.7 Drift collapse due to a shallow angle of draw

Figure 3.8 shows a case in which the collapse of the cave front was induced by a steep angle of
draw over 60 degrees.

801

70- Collapse

£ 60-|
CO

o
Steep
"5 5 (
H
Angle of Draw
? 40 • •
< -fxt\
30

20

CO
o o o
> d.
o
CO

2 z < CD

Month

Figure 3.8 Drift collapse due to a steep angle of draw

43
Figure 3.9 shows the third mechanism of failure related to the sudden change of angle of draw
from period to period. A theoretical explanation for this behaviour has not been found either in
the literature. However it is believed that the stress tensor suffers several short-term changes as a
result of modifying the angle of draw which would induce corresponding short-term changes in
shear loads leading to rock mass failure.

™]
65-| Collapse
60- ooooooooooooooooo
/
55-
5
CO
50-
A
o 45-

c 40- Changes of the

< oooooooooo
35- Angle of Draw
30-
25 H
20-
C\I CM co CO
O o o_ o
o o"
CD o CO
Q O 2
Month

Figure 3.9 Drift collapse due to a sudden change on the angle of draw

Based on the draw charts shown above it is possible to see that at this operation the range for the
angle of draw should be in the interval 30 to 60 degrees. Obviously the ultimate range for the
angle of draw is site dependent since this parameter is a function of in situ stress tensor, rock
mass strength, undercutting method among others. Nevertheless the aim of this exercise was to
show that production performance represented by the angle of draw may trigger geotechnical
events that affect the stability of the cave front.

The cave front stability affects the undercutting process which influences the development rate.
Araneda and Gaete (2004) derived an interesting relationship between development rate and
block cave productivity

44
 T{t)=D Hp
v 1-exp (3.1)

The parameters of the model shown above are listed as follows:

T(t) tonnage production per unit time

D v development rate in area per unit time
H height of the economic draw column
p the in situ specific gravity
vd the draw rate in tons/m /day
t the time variable

From the above relationship it is possible to see that any constraint on the development rate will
affect the production tonnage and therefore the productivity of the mining system.

Induced stress on the production level

Stress is redistributed on the production level as a function of the draw pattern performed across
the production area. It is well established that even draw leads to a more uniform stress
distribution on the production level than isolated draw. Isolated draw consists of performing
excess ore extraction from a draw point or a group a draw points in isolation from the rest of the
production area. Isolated draw leads to high concentration of stresses often resulting in large roof
displacements. This phenomenon has been described by Verdugo and Ubilla (2004) who used a
F L A C 2D model to represent the induced stress as a result of the arch effect produced in a
hanged up draw point. Encina et al (2004) developed a model derived from first principles to
analyze the effect of induced stress as a result of isolated draw. Also Rubio et al (2004) showed a
discrete model constructed using PFC 2D to demonstrate the induced stress as a result of
different draw strategies.

Although all this evidence presented in different papers it was desired to show the effect of
production performance on the deformation of the major production apex pillar. Thus
convergence observations, taken from mine M l , were used to study this phenomenon. Figure
3.10 shows the deformation of a draw point major apex pillar as a result of uneven draw. In this
case the deformation led to the collapse of the production drift which later triggered closure of
the entire production crosscut.

45
 Figure 3.10 Draw point minor apex pillar deformation as a result of uneven draw at mine
M4

Despite the empirical evidence and theoretical explanations for stress redistribution on the
production areas of a block cave mine, none of the mines studied showed a clear indicator to
assess the uniformity of the draw pattern. A n indicator called "draw cycle" was developed as part
of the research to represent the uniformity of a given draw pattern. The draw cycle of a
production crosscut is defined as the time it would take to draw at least a certain number of
buckets from every draw point of the crosscut. The longer the crosscut draw cycle, the more
chance there is of areas of no draw within the crosscut leading to instability. A n alternative
definition of the draw cycle would be the fraction of the total number of draw points in a crosscut
that are drawn in a given period of time.

The Mine M l database was used to study the relationship between the convergence
displacements observed at different places along the production drift and the draw cycle
computed at this mine. The analysis was conducted in periods of four days given three years of
production and convergence records. In this case the crosscut draw cycle was estimated as the
fraction of the total number of draw points drawn (at least 500 tons) in a time period of four
days. Figure 3.11 shows a relationship between the fraction of the production crosscut (XC)
drawn in a four day period and the total convergence observed along the crosscut for the same
period of time. Four days was used as the time period in order to obtain the whole range of

46
possible draw cycles. Figure 3.11 suggests a direct inverse correlation between crosscut draw
cycle and convergence observed on the active production area of a block cave mine.

12-,

o 11

10-
o 9-
>^
o
8-
OB
T3
7-
5f-
E 6-
E, 5-
CD
o
c 4-
<x>
cn ^ i . ®n n •
>
c 2-
o
O
1 -

I— —I— -1— —I— —I— —I— —I— -1— -1—

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Fraction of the XC drawn

Figure 3.11 Total crosscut deformation as a function of the draw cycle

3.3 Failure of mining infrastructure

It has been shown that geotechnical events such as convergence and fragmentation affect the
productivity of the mining system. In this section the failure characteristics of a draw point will
be studied to achieve a better understanding of the relationship between failure and production
performance. To begin the study, consider the draw point status recorded at mine M l as shown
on Figure 3.12.

47
 Draw PointA Day 1 2 3 4 5 6 7 8 9 10 11 12 13 14
D1 A A A A A A A A A A/H A/H A/H A/H A/H
D2 B/H B/H B B B B B B B B B B B B/H
D3 A A A A A A A A A A A A A A
D4 A A A A A A A A A A A A A A
D5 A A A A A A A A A A A A A A
D6 A A A A A A A A A A A A A A
D7
D8 0 0
D9 - 0 0
D10 0 0
D11 A/H A/H A A A/H A/H A/H A/H A A/H A/H A/H A/H A/H
D12 C C C C C C C C C C C C C C
D13 A/H A/H A/H A A/H A/H A/H A A/H A A A/H A/H A
D14 A A A A A A A A A A A A A A

Figure 3.12 Operational records of draw point status data

Note that in this case failure is recorded when the draw point status changes from active (A) to
any of the other status such as oversize (B, B/H), temporarily closed due to high convergence (C)
or hanged up (A/H). Table 3.12 shows the method that has been used to count the number of
failures of a draw point. In this example draw point D54 is under analysis. Column "Code"
contains a binary variable with value of 1 if the draw point is Active or 0 otherwise. Column
"Failures" contains a binary variable with value of 1 if a draw point has changed its status Active
to any other status within two consecutive days.

Table 3.12 Counting failures as a function of the daily draw point status records

Time (Days) Draw Point Name Status Code Failures

1 D54 A/H 0
2 D54 A/H 0 0
3 D54 A/H 0 0
4 D54 A/H 0 0
5 D54 A/H 0 0
6 D54 A 1 0
7 D54 A/H 0 1
8 D54 A 1 0
9 D54 A 1 0
10 D54 A 1 0
11 D54 A/H 0 1
12 D54 A 1 0
13 D54 A 1 0
14 D54 A 1 0
15 D54 A 1 0
16 D54 A/H 0 1

48
Table 3.13 that shows the times during the life of a draw point at which failure occurs and the
cumulative number of failures as a function of time, N(t).

Table 3.13 Counting process of failures for draw point D54

Cumulative Number of
Time of Failure (Days)
Failures N(t)
7 1
11 2
16 3
31 4
37 5
41 6
46 7
61 8
67 9
71 10
76 11
91 12
97 13
101 14
106 15
121 16

Figure 3.13 shows a plot of N(t) versus time. This plot is very relevant to the understanding of

the evolution of failure in a draw point and the general mining infrastructure. For instance one
can see that after 150 days of operation the tendency to failure changes reflected by the shape of
N(t) which becomes more concave.

At the moment draw point failure has been related to time. However, mining infrastructure ages
as a function of the tonnage that has been processed through it. For example two draw points
could have been commissioned at the same time but one of them may have produced 150,000
tons and the second one only 40,000 tons. The first draw point would have been exposed to more
wear and tear than the second draw point which is less mature (in terms of tonnage drawn).
Therefore it is expected that the first draw point would have a greater tendency to fail than the
second one. This suggests that time is not the correct indicator to represent the tendency of draw
point failure.

49
 _2
30

E 20

CD

E
O 04

— I — -1— — I —
100 200 300 400 500 600 700

time of failure (days)

Figure 3.13 Cumulative number of failures for a single draw point

The production history of draw point D54 was taken from the tonnage production database and
cross related with the failures recorded per month as shown on Table 3.14.

Table 3.14 Estimation of Draw Point Rate of Occurrence of Failure

Time of Failure N(t) Cumulative Tonnage Incremental Tonnage W(T)

(Days) (tons) (tons) (Failures/tons)
31 4 6,623 6,623 0.0006
61 8 11,073 4,451 0.0009
91 12 18,492 7,419 0.0005
121 16 20,606 2,114 0.0019
151 20 28,949 8,343 0.0005
211 24 34,326 5,377 0.0007
312 25 42,884 8,558 0.0001
334 26 45,713 2,829 0.0004
427 34 55,065 9,352 0.0009
519 40 64,562 9,497 0.0006
574 41 73,069 8,507 0.0001
609 46 77,075 4,006 0.0012
631 47 79,254 2,179 0.0005

The first column of Table 3.14 shows the chronologic time at which the draw point failed. The
second column shows the cumulative number of failures. The third shows the cumulative

50
tonnage drawn from the draw point at the time of failure. The fourth column shows the
calculation of the incremental tonnages between times of failure. The last column shows the
calculation of the draw point rate of occurrence of failure per tons.

The draw point rate of occurrence of failure (draw point ROCOF) represents the frequency at
which a draw point fails in a given tonnage interval A r . ROCOF is used rather than failure rate
since a draw point is a repairable component that behaves as new after repair has been
conducted. Failure rate is usually used to represent the failure behaviour of a non-repairable
mechanical component (Hoyland and Rausand, 1994). The draw point ROCOF is given by:

w[T) = -^-E[N{T)] (3.2)

at

where W(T) is the draw point ROCOF after drawing T tons from the draw point, is[iV(r)] is the

expected value of the number of failures, N(T), in the tonnage interval (0,r]. For a discrete

tonnage interval A r Equation 3.2 may be rewritten as follows:

^ ) = l i m £ M l ± A T ) - M r ) l (3.3,
v
' AT^O AT

If A r is small, E[N{T + AT) - N{T)] = M (T, T + A r ) , the mean number of failures in the tonnage

interval between draw point repairs. Then a natural estimator for w(T) would be
M T H M ( T , T + . A T )
V
' AT

Equation 3.4 will be used to estimate the draw point ROCOF for a given tonnage interval A r .
Table 3.14 shows the application of Equation (3.4) to estimate the draw point ROCOF during the
production history of draw point D54.

Figure 3.14 shows a plot of draw point ROCOF against the cumulative tonnage drawn. From this
plot is possible to see at least three areas in which the tendency of draw point D54 to fail
changes. Between 0 and 20,000 tons drawn there is a decrease in the tendency of failure.
Between 20,000 and 60,000 tons there is a relatively constant tendency to failure. Over 60,000
tons there is an increase in the tendency to failure. This behaviour could be explained by the
nature of the mining method in which at the beginning of the life of the draw point there is a high
probability of experiencing hang ups and oversize due to primary fragmentation with very little

51
or no influence of secondary fragmentation. Between 20,000 and 60,000 tons drawn the draw
point enters a steady state regime in which there is an even probability of experiencing failure,
most likely low probability of hang ups and medium probability of over size due to the influence
of secondary fragmentation. Over 60,000 tons drawn in the case of D54 the draw point start to
wear out showing damage on the concrete and the draw point brow tends to retreat towards the
production drift.

2.5x10"

S 2.0x10

J2
2=- 1.5x10 H 3

LL

o
O
o
OC 1.0x10"H
c
o
0-
| 5.0x10"*-
Q

0.0-

§ 0
0
0
0
0
0
0
0
0
0
O" Q Q O" O" O"
CM CO * U5 co

Cumulative Tonnage Drawn, T, (tons)

Figure 3.14 Experimental draw point ROCOF for draw point D54

The average draw point ROCOF curve of mines M l and M 2 was computed using the following
procedure;
1) Compute the cumulative tonnage drawn per draw point per month for the time horizon
under study.
2) Compute the number of failures of the draw points for each month during the time
horizon under study
3) For each 10,000 ton interval of cumulative tonnage drawn, compute the average number
of failures
4) Plot the computed average ROCOF versus the cumulative tonnage.

52
Table 3.15 shows a detailed calculation of average draw point R O C O F for mine M l . The second
and the third columns represent the total number of draw points and the total number of failures
of draw points with tonnage less or equal to the lower bound tonnage. The average number of
failures per draw point is the ratio between the differentials of column 3 and 2. Finally the draw
point R O C O F for different tonnage intervals is the result of dividing the 4 column by 10,000
th

tons.
Table 3.15 Average draw point ROCOF for Mine M l

Average Number of
Lower Bound Number of Number of Failures per Draw w(T)
Tonnage Draw Points failures Point
(Failures/tons)
10000 327 592 1.81 1.8E-04
20000 666 1056 1.37 1.4E-04
30000 1002 1486 1.28 1.3E-04
40000 1338 1944 1.36 1.4E-04
50000 1650 2335 1.25 1.3E-04
60000 1965 2710 1.19 1.2E-04
70000 2273 3090 1.23 1.2E-04
80000 2554 3430 . 1.21 1.2E-04
90000 2807 3722 1.15 1.2E-04
100000 3010 3954 1.14 1.1E-04
110000 3201 4161 1.08 1.1E-04
120000 3342 4325 1.16 1.2E-04
130000 3445 4460 1.31 1.3E-04
140000 3513 4556 1.41 1.4E-04
150000 3548 4623 1.91 1.9E-04
160000 3565 4655 1.88 1.9E-04
170000 3575 4681 2.60 2.6E-04
180000 3578 4689 2.67 2.7E-04

Figure 3.15 shows the resulting draw point R O C O F curve for mine M l . This curve represents
the evolution of the tendency of a representative draw point in Mine M l to fail.

53
 2.8x10 -

2.6x10^-

2.4x10""-

2.2x10""-
w(T)
2.0x10""-

1.8x10""-

1.6x10""-

1.4x10""-

1.2x10""-

1.0x10""-
I T I
o O o o o o o o o o
o o o o o o o o
o o o o o o o o
o o" o o" o" cf o"
CM CO o CD 00 o
CM

Cumulative Tonnage Drawn, T (tons)

Figure 3.15 Draw point ROCOF curve for mine M l

Across an active layout there could be draw points that do not follow the same trend showed in
Figure 3.15. More detailed analysis can treat these outlier draw points using a different draw
point ROCOF curve representing the failure tendency of that subset of draw points.

The tonnage-dependent behaviour of failure of a draw point shown by the R O C O F curve is

interesting. At the beginning of a draw point's production life, it is exposed to several factors that
may induce it to fail such as poor construction quality, pillars left on the undercut level due to
poor undercut blasting, or blast damage of the crown pillar due to poor design of the drilling and
blasting at a draw bell. However the most serious geotechnical event would be the cave
propagation after blasting the draw bell. After drawing the ore blasted as part of the undercutting
a draw point moves to a regime in which the failure rate is relatively constant and dependent on
secondary fragmentation and factors such as oversize and minimum hang up events. At the end
of the productive life of a draw point the draw point brow will have been eroded. There will be a
few steel arches lost, as a result of the constant friction between the concrete and the flow of
rock, and the production drift will have experienced severe damage. These three behaviours are
observed in Figure 3.15. At the beginning there is a high rate of failure which decreases until
about 40,000 tons are drawn. Then the draw point moves to a stable rate of failure depending on

54
secondary fragmentation. Finally when the draw point reaches 120,000 tons drawn it starts to fail
due to wear out of concrete and loss of steel arches on the draw point brow.

It is interesting to draw an analogy between the draw point ROCOF curve and a failure rate of a
mechanical component as shown in Figure 3.16. The failure rate of a mechanical component,
often called "bathtub" curve (Hoyland and Rausand, 1994) due to its characteristic shape,
represents the evolution of the tendency of a mechanical component to fail during its life.

Period Useful Life Period Period

Figure 3.16 Bathtub curve for a mechanical component (Hoyland and Rausand, 1994)

There are three main areas on the bathtub curve. During the burn-in period there are
undiscovered defects which usually manifest themselves when the units are activated. When the
units have survived the burn-in period, the failure rate often stabilizes at a level where it remains
for a certain period of time called the useful life period until it starts to increase as the unit begins
to wear out. The similarity to the ROCOF curve of a mine is evident. This analogy is used to
introduce the concept of draw point reliability later in this chapter.

3.3.1 Data Analysis for computing draw point ROCOF curve

The estimation of ROCOF based on operational records of draw point status may be misleading
as a result of noise contained on the raw data. For example it has been found at several
operations that a draw point shuts down due to operational logistics rather than geotechnical
problems. Also when a draw point has been cleared and is operational, the status of the draw
point will change in the database only when the draw point has a L H D allocated to it. This could

55
be several days. Until that time, the draw point would still be in repair status. It is expected that
in the future the allocation of secondary breakage equipment will be driven by the dispatch
system which could enhance the way draw point status is recorded.

The methods used in this research to reduce the effect of noise in the recorded draw point status
are summarized as follows:
• Disregard raw data outside the 95% confidence interval, that is data that is outside of the
average plus or minus two times the standard deviation is ignored
• Crosschecks such as between draw point status and tonnage records for the same day, or
between draw point status and production crosscut status.

3.4 Draw point ROCOF curve as a function of geotechnical domains

The aim of this section is to study a possible relationship between the shape of the draw point
ROCOF curve and the underlying geotechnical domain. The concept of the R O C O F curve would
suggest that as the geotechnical characteristics of the rock mass become more adverse from a
mining system point of view, the more failures a piece of mining infrastructure would
experience. In the case of draw points and the main components of block caving infrastructure,
as the rock mass becomes weaker, typically with a mining rock mass rating (MRMR)
(Laubscher, 1989) less than 30, the draw point ROCOF curve should be shifted upwards as a
result of high frequency of collapses experienced on the production level as a result of the low
rock mass rating.. The same effect is observed when the rock mass is competent ( M R M R greater
than 70) the draw point ROCOF curve would be shifted upwards due to draw point
fragmentation characteristics, resulting in high frequency of hang ups and oversize as a result of
the high rock mass rating.

The relationship between draw point ROCOF curve and geotechnical domains has been tested by
constructing the draw point ROCOF curve for every production crosscut. It was expected that
eventually a crosscut would be developed in a different geotechnical environment (different
M R M R ) area respect to the rest of the crosscuts. This should be reflected by the shape of the
draw point ROCOF curve. Figure 3.17 shows an analysis conducted using the data from mine

56
M l , in which the draw point ROCOF curve has been constructed for three different production
crosscuts.

0.00028 -
- • - X C 13
0.00026 -
- o - X C 15
0.00024 - - x - X C 16
0.00022 -

w(T) 0.00020 -

0.00018-

0.00016-

0.00014-

0.00012-

0.00010-
T
o o o o o o o
o o o o o o o
o o o o o o o
o" o o
9 •a- <o

Cumulative tonnage drawn, T (tons)

Figure 3.17 Draw point ROCOF curve for different production areas of mine M l

Based on Figure 3.17 the crosscut XC13 exhibits a higher likelihood of failure based on its draw
point ROCOF curve. This agrees with the fact that the mine started undercutting in this area due
to its weak rock mass characteristics. Hence, the same geotechnical characteristic that was used
to decide the place where to start undercutting was harmful to the long term stability of the draw
points. Therefore different draw point ROCOF curves have to be used to represent appropriately
the draw point rate of failure of different geotechnical domains found across the mine.

A second test consisted in comparing the average draw point ROCOF curve of two different
mines ( M l and M2) which happen to be operating in two different geotechnical environments.
M l operates in a rock environment that is considered to be the optimal for the use of block cave,
since the rock is strong enough to sustain the production infrastructure but weak enough to
produce the desired fragmentation for the mining method. On the contrary, M 2 is developed in a
competent rock mass environment that has delivered a very coarse fragmentation inducing a high
rate of hang ups and oversize during the life of the mine. Figure 3.18 shows two different draw
point ROCOF curves from mines M l and M2.

57
 1.4x10-]

1.2x10
° Draw Point Rocof mine M2
1.0x10° H * Draw Point Rocof mine M1

8.0x10"

6.0x10"

4.0x10"

2.0x10"

0.0 T
"T" T T -1
o
o
o
O
O
O
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o"
CM
o" o" o" o" o" o" o o
<o CO o C\J •0- (D CO

Cumulative Tonnage Drawn, T (tons)

Figure 3.18 Draw point ROCOF curve comparison between two operating block cave
mines

It is seen that mine M 2 has a higher average draw point R O C O F curve than mine M l . In other
words a M2 draw point fails more frequently than a M l draw point. This is effectively reflecting
the fact that M 2 is operating in one of the most competent rock mass environments found in
block cave mining today (MRMR=80). On the other hand, M l operates in what is considered an
ideal rock mass environment for block caving with a M R M R = 45 across the layout, facilitating
cave propagation and fragmentation. In summary it is possible to see that both curves have
similar shapes but the gap between them is related to the rock mass condition that triggers
different geotechnical events and subsequent production losses.

Finally the draw point ROCOF curve has been identified as a very important caving performance
indicator since it could provide insights as to how the production system will evolve in time
without explicitly modeling the geotechnical behavior of the system. It is not suggested that
geotechnical modeling should be excluded from the engineering process of designing a block
cave mine. On the contrary, the author believes that such modeling may contribute a great deal to
the understanding of the rock mass and its role in the mining system. Nevertheless the draw point
R O C O F curve is a production control/planning tool that enables engineers to quantify the effect

58
of geotechnical events on the likelihood of mining infrastructure to achieve a desired production
target.

The next step consists of introducing a tool that characterizes the life cycle of a draw point from
a reliability point of view. This concept is important since it facilitates the implementation of
draw point failure performance as part of a production scheduling process.

3.5 Draw Point Reliability as an Indicator of Geotechnical Events

In the previous section, two geotechnical events were studied: hang ups and displacements along
the production drift. It was shown that these two geotechnical events are correlated with draw
point production performance. In this section, draw point reliability will be introduced as a
planning parameter that could subsume the effects of at least these two geotechnical events.
Draw point reliability is defined as the probability that the draw point will be available to
produce certain amount of tonnage in a given period of time. Since the production capacity of a
draw point will be determined in part by the underlying geotechnical events that induce failure
on mining infrastructure, the failure characteristics of a draw point will determine the reliability
of a draw point. The aim of this definition is to use the draw point reliability as a planning
parameter that would facilitate the estimation of the mining system reliability by modeling the
mine as if it were an aging mechanical component. Thus, concepts used in reliability engineering
could be applied to better forecast the failure behaviour of a draw point, consequently enhancing
the ability to forecast draw point production capacity.

Lakner and Anderson (1985) have defined reliability as the probability that an item will perform
a required function under stated conditions for a stated period of time. Note that "functioning"
implies that an item could be either in active operation or able to operate if required. In that
context, reliability of a given component, r, is related to the probability of failure, p, through the
equation
r = \-p (3.5)
The reliability of a draw point is defined as the probability that the draw point does not fail to
produce a certain tonnage AT. This probability will be computed as a function of the draw point
R O C O F curve which is the natural estimator of draw point tendency to failure.

59
3.5.1 Measurements of draw point reliability
A n approach similar to that used by Ascher and Feingold (1984) can be used to define a
relationship between the estimated ROCOF and an estimate of draw point reliability. The
reliability of a draw point to produce r tons can be expressed as:
R{T) = 1 - F(T) = P{T0 > T) (3.6)

Then knowing that the rate of occurrence of failure is the probability that the draw point would

fail within the tonnage interval (T,T + AT], W(T)can be written as:

, P(T<T <T + AT\T >T)

= lim °- LJ 1 (3.6)
N () N

w(T)

Using Bayes theorem the conditional probability can be re-written as:

V P{T<T <T+AT\T >T)

0 0 F(T+AT)-F{T) 1
w[T) = lim = lim — - — — r (3.7)
Ar->o AT A T ^ O Ar R(T)

From Equation 3.6 R(T) = —F(T) . Substituting in Equation 3.8

w (r) = -4^ (3.8)

Integration of the above differential equation gives

T+AT

\n[R(T + AT)]-ln[R{T)] = - ^w{u)du (3.9)

T •

The draw point is functioning after drawing Ttons, then fl(r) = l . Thus, equation (3.9) can be

re-written as

R(T + AT) = exp ^w(u)du (3.10)

In order to compute Equation 3.10 one would need to know the form of the function w(u).
Several models can be fit to this function such as the Poisson or renewal processes (Rigdon and
Basu, 2000). In the following w[u) is read directly from the draw point ROCOF curve and it is
assumed to be constant for the tonnage interval at which the production schedule is run. The
tonnage interval for a typical long term production schedule run in a monthly basis would vary in
the range of 3,000 to 8,000 tons which is less than 10,000 tons, the tonnage step for the
construction of the draw point ROCOF curve. This assumes that the draw point tendency to

60
failure remains constant in a tonnage interval of 10,000 tons which is about 15m of draw column.
This assumption is reasonable since in block cave mining system the rock mass is often very
uniform. Thus, the draw point ROCOF curve will be replaced in Equation 3.11 resulting in the
following expression for estimating the draw point reliability:

R{T + AT) = exp[-w{T)AT] (3.11)

The reliability per draw point per month was estimated for the mines M l and M 2 using the
above formulation. Mines M3 and M4 were not analyzed since there were no records of draw
point status per day or hang up frequency to construct the draw point bathtub curves. The
estimation of historical draw point reliability was useful to compare against geotechnical factors
such as hang up frequency and convergence. It is expected that this will show that draw point
reliability could subsume the effects of geotechnical events that trigger operational upsets.

Note that the measurements of reliability have been made independent of the tonnage drawn per
draw point or crosscut. This provides two independent variables: tonnage and reliability that aim
to characterize the mining system at any given time.

To corroborate the estimation of draw point reliability as a function of the monthly R O C O F the
draw point availability was computed. It was expected that there should be a direct correlation
between draw point reliability computed using draw point monthly ROCOF and monthly draw
point availability. The monthly draw point availability was computed as the ratio of the time that
a draw point has been active to operate over a time period of a month. Figure 3.19 shows the
average monthly draw point availability during the life of a draw point of mine M l . It is seen
that the shape of the monthly draw point availability is the inverse of the draw point ROCOF
curve. This will lead to a direct correlation between the draw point reliability derived from
R O C O F and draw point availability as shown in Figure 3.20.

61
 1.0
0.9
0.8
0.7
• •
0.6
'cc
> 0.5 • •
<
c • •
o 0.4 •
D_ •
0.3
co
Q 0.2
0.1
0.0
XL XL XL XL XL XL
O O O o o O O O o
C\J •<* CO 00 o CM CO 00

Cumulative Tonnage Drawn (tons)

Figure 3.19 Average monthly draw point availability from mine M l

Figure 3.20 shows a close correlation (0.99) between draw point reliability and draw point
availability. Therefore it would be possible to use Equation 3.11 within the production schedule
to compute draw point reliability as a function of the draw point ROCOF curve.

1.0-1

Li. 0.9-
O
OC

0.8-
cc
E 0.7-
o
0.6-
&
3 0.5-
m
Rel

0.4-
c
0.3-
o
D_

5 0.2-
CO

Q 0.1 -

0.0-
0.0

Draw Point Availability

Figure 3.20 Comparison of draw point reliability versus draw point availability

62
3.5.2 Effect of hang up frequency on draw point reliability
Figure 3.21 shows a relationship between hang up frequency and draw.point reliability at
different stages of draw point maturity for mine M 2 . There is a negative correlation between
hang up frequency and draw point reliability as a measure of its ability to provide the planned
tonnage. Note that the available data is mainly in the interval of 2 to 10 hang ups/1000 t which
results in a reduction of draw point reliability from 0.7 to 0.4.

1.0-

0.9-

&
0 8
'
2
CO

'a 0.7-
or PB
CD
8. o.6 H

I
Q
0.5 4 •

0.4 H -D

10 15 20 25 30 35

Hang Ups/ 1000t

Figure 3.21 Effect of hang up frequency on draw point reliability

3.5.3 Effect of induced stress on draw point reliability

Induced stress across the active layout has been recognized as the second most important cause
of draw point breakdown after draw point oversized and hanged up. Figure 3.22 shows also a
negative correlation between measured convergence, as a result of stress redistribution, and draw
point reliability for mine M l .

Based on the empirical charts constructed with data from mines M l and M2, it was possible to
conclude that draw point reliability does represent the most relevant geotechnical factors that
trigger draw point failure with consequent loss in production. This conclusion is very important

63
since it implies that geotechnical factors can be represented by a simple operational measure
such as the draw point reliability.

1.0n

0.9 H •

• •
cc
0.8 A
0)
CC
•
-*—
c
1
•
~~tr-
o 0.7 H
Q.

cc
D
0.6-

0.5- 1 ' 1 ' 1

4 6 8 10 12

Convergence (mm/day)

Figure 3.22 Effect of convergence on draw point reliability

The fact that hang up frequency and convergence observations correlate well with draw point
reliability as shown on Figure 3.21 and Figure 3.22 respectively, indicates that draw point
reliability could subsume the geotechnical events that lead to infrastructure failure. This result is
of interest since draw point reliability, which is an operational indicator derived from empirical
measurement of failures, would allow modeling the rock mass response to mining as a
realization of the fundamental geomechanical models. The introduction of draw point reliability
into a mine wide reliability model would facilitate the calculation of how realistic a production
target is as a function of the system reliability. This will be further explained in the next chapter
of the research.

64
4 PRODUCTION SCHEDULING INTRODUCING THE CONCEPT OF
RELIABILITY
One of the main objectives of the research presented in this dissertation is to compute the
probability of success of a given production plan. This can be computed as the reliability of a
production schedule, which represents the inherent uncertainty of the mining system to achieve a
desired production target as a result of the geotechnical events that affect the availability of
mining infrastructure. Chapter 3 showed how the draw point ROCOF can be used as an
operational indicator of the likelihood of failure of a draw point. It was also shown that the draw
point reliability, derived from the draw point ROCOF curve could subsume geotechnical events
that trigger failure of the draw point. In this chapter the individual draw point reliabilities will be
introduced in a mine wide reliability model to compute the system reliability. Then this model
will be used to compute the probability of success of different production schedules as a function
of planning variables such as production targets, draw point yield, and development rate.

4.1 Definition of System Reliability

The concept of reliability applied to mining was first introduced by Dotson (1966), describing
mainly the failure modes of equipment and its influence on productivity. Later Kumar and
Granholm (1988) constructed reliability models to support the design, planning and operation
mainly oriented to equipment performance. These authors concentrate on using reliability to
identify the most unreliable subsystems and the economic implications of these subsystems.

Ramani, et al (1989) also discussed the application of the concepts of reliability, maintainability
and availability to study the subsystems of longwall mining system. Kazakidis and Scoble (2002)
used continuous distributions to model the reliability and hazard function of the mining system
components presented in Figure 4.1. The authors used different probability distributions to model
the failure rates of different mining infrastructure. Nevertheless, the authors stated that there was
not enough operational data to support the use of a given set of distributions for different
components of the mining system. Also this research does not consider whether the components
of a mining system are non repairable or repairable components. As shown in Chapter 3 the
components of the block cave mining system, particularly draw points, are repairable

65
components. This leads to a different formulation than the one presented by Kazakidis and
Scoble (2002)

Figure 4.1 Components of a traditional mining system (Kazakidis and Scoble, 2002)

Reliability modeling of a system is based on the failure characteristics of the individual

components of the system represented by the reliability of the system's component. To compute
the reliability of a system based on the reliability of its components a "Reliability Block
Diagram" (Hoyland and Rausand, 1994) is used to represent the hierarchical relationships among
the different components of a system. Individual components of a system can be connected in
series, parallel and series-parallel combinations, examples of such combinations are shown in
Figure 4.2.

66
 1
(a) (b)
2
1 2 3
3

R = rrr-
{ 2 3 /? = l-{(l-r )(l-r )(l-r )}
1 2 3

(C) (d)

1 2 3 1 h H 3 h

1 2 3 H 1 H 3

/? = l-{(l-r r r ) }
1 2 3
2
/? = [l-(l-r ) Il-(l-r ) Il-(l-r ) ]
1
2
2
2
3
2

Figure 4.2 System reliability based on the reliability of the components r , r , r x 2 3

(Hoyland and Rausand, 1994)

The systems shown in Figure 4.2 show different levels of reliability and redundancy. The series
system in Figure 4.2(a) is the least reliable since all components must function for the system to
function. The parallel system in Figure 4.2(b) is more reliable since at least one component must
function for the system to function. Figure 4.2(c) is a parallel-series system which has
redundancy at the subsystem level while Figure 4.2(d) is a series-parallel system that has
redundancy at the subunit level.

Figure 4.3 shows a comparison between the parallel-series system and a series-parallel system of
three components with 100% component redundancy at the subunit level. From this figure is
possible to see that the series parallel system is a more reliable system given the same amount of
redundancy. This analysis leads to the statement that redundancy of components at the subunit
level rather than at the subsystem level serves to increase system reliability.

67
 0.8 A —a— Parallel series
—o— Series parallel

0.2 H
,o

0.0
0.0 0.2 0.4 0.6 0.8 1.0

Components Reliability

Figure 4.3 Comparison between series parallel and parallel series for 3 components with
100% component redundancy

If the block cave mining system is viewed as a mechanical system composed of different pieces
of infrastructure with different reliabilities, the reliability of the mining system could be
computed by linking its components in a manner similar to the examples shown in Figure 4.2.

4.2 Reliability of a System with Redundancy at the Component Level

Another way of modeling systems that contain components redundancy is by using the k-out-of-n
models. A k-out-of-n model consists of a system that has n components to perform a function
that only needs k to function. This kind of models will be further study on the next section of the
dissertation since they will be used extensively to model the reliability of the block cave
production system.

A system that contains n independent components of which k < n are needed is redundant and is
referred to as a k-out-of-n system. The system functions if and only if k of the n components
function (Boland and Proschan, 1983). The quantity (n-k) is known as the system redundancy.
The particular case of a parallel system would be a 1-out-of-n system and a series system would

68
be a n-out-of-n system. The reliability of a k-out-of-n system is computed by the following
equation

/?(M) = £ c , „ r V - ' (4.1)

i=k

where C in is the number of combinations of i functioning components of n available given by

c, n ]
~("
i\(n — i)\ yi y

r is the reliability of each component and q is the unreliability 1 - r . The product r'q"~' is the
reliability of i components connected in series having n components available. Figure 4.4 shows
a diagrammatic representation of system composed of 10 components connected in series to
perform an activity and a second system with 50% of components redundancy whereby, there are
15 components available to perform an activity that just needs 10 components. The second
system is called a 10 out of 15 system.

Figure 4.5 shows a comparison between a series system of 10 elements and a I0-out-of-l5
system in which every component has the same reliability. It is clear that the system reliability
increases by adding redundancy to the system. Certainly, to achieve this increased reliability, it
must be physically possible to combine the k components. For a block caving system this is
significant; it means that n draw points must be developed and available to provide all the
possible combinations.

69
 Er-EHIHIHIH^
10 components in series

Figure 4.4 Selection of 10 out of 15 components to compare with the performance of 10

components connected in series

1.0-1

0.9- - • - 10-out-of-15
— o — 10 series •
0.8-
/
0.7-
-
•
« 0.6- o
"53 •
CC 0.5-
E
CD
+-'
CO
0.4-
• o
>,
CO 0.3- /
•
0.2- / o
0.1 - •
J3
0.0- mama cp-B'rj ® Q Q Q-Q-°', l—>—i—«—i
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Component Reliability

Figure 4.5 Reliability comparison between a series and a k-out-of-n system

4.3 Block Cave Reliability Model

The reliability model associated to the block cave production system consists of production
infrastructure components such as draw points, ore passes, crushers, tunnels. These components

70
are linked through different relationships depending on the goal of the mine plan. To improve the
understanding of the model, the mining system has been divided into: production areas,
production crosscuts and draw points as shown in Figure 4.6.

Production Production
k^out-of-n.
Crosscut 1 Drift 1

DP 1

Production Production DP 2 . ki-out-of-n;

Production
Unit Crosscut / ' Draw Points of
Drift /'
crosscut /
DPn,

Production Production
Crosscut N Drift N k -out-of-n
N N

K-out-of-N
Crosscuts to achieve
production target

Figure 4.6 Reliability block diagram of a block cave production system

A production unit is an area of the mine that has a particular geology, geotechnical
characteristics and the location facilitates the logistics of the mine. Every production unit is can
be modeled as a k-out-of-n system composed of production crosscuts. Every production crosscut
consists of one production drift connected in series with a k-out-of-n system of draw points.

Usually a production target is met by the tonnages produced from the different production
crosscuts. Based on the nominal crosscut productivity there are usually more production
crosscuts than needed to achieve a desired production target. This means that there is
redundancy, more components than needed, at the crosscut level. The same can be observed at
the draw point level. To achieve the crosscut nominal production target there are more draw
points than needed based on the draw point nominal productivity. Thus the block cave mining
system contains redundancy at the crosscut and draw point levels.

71
Usually fc-out-of-n systems in this dissertation consist of systems with identical components, i.e.
all components have the same reliability. In block caving, the infrastructure components of the
mining system such as draw points and drifts may have different reliabilities since they may be
located in different geotechnical domains or may have different ages. For example in the case of
a production crosscut there could be draw points at different stages of maturity that would lead to
different draw point reliabilities along the production crosscut. The same can be observed at the
crosscut level in which there are several crosscuts with different reliabilities yielding the
production target. These systems are called k-out-of-n system with independent and non-
identical components. Computing the reliability of such systems is more complicated than
systems with identical components. The notation used to compute the reliability of a fc-out-of-n
system with independent and non identical components is presented below:
n number of components in the system
^ minimum number of components that must function for the k-out-of-n system
to function
rt reliability of component i, i = 1,2, • • •, n
r reliability of each component when all components are identical.
q
{ unreliability of component /, q = 1 — p , i = 1,2, • • •, n
i t

q
unreliability of each component when all components are identical q= \ - p
g x
intermediate reliability entry which represents the probability that exactly i out
e
of n components are functioning
R(k ri) y °f k-out-of-n system or probability that at least k out of the n
r e n a D u u a

components are functioning, where 0 < k < n and both k and n are integers
unreliability of a k-out-of-n or probability that less than k out of the n
Q(k,ri) components are functioning, where 0 < k < n and both k and n are integers,
Q(k,n) = l-R(k,n)

Suppose that in a given crosscut there are n draw points available and depending on the average
draw point yield and the crosscut production target, k out of the n draw points are needed to meet

the target. Define a subset of / functioning in series out of n available as s' with T = l,2,---C
T IN

and k <i<n (i< k will not be a feasible system). Then the reliability of a given subset s' is T

R{S' )=
T /'(components ts s' available)x /'(componentste s""' unavailable)
T

72
Denote the set of all s' subsets is by S" .Then the reliability of the k-out-of-n system with non-
T

identical and independent components is given by

^.»)=zz[ri -.( *i)n*(^*r)] | fe

(4.3)
i-k S"

Incorporating the tunnel or production drift reliability R , the production crosscut reliability
T

Rcx is defined as

R =R R(k,n)
cx T (4.4)

If explicit enumeration were to be used to compute Equation (4.4) 30,827 draw point
combinations would be required to compute the reliability of a 5-out-of-\5 system. However, a
recursive algorithm developed by Barlow and Heidtmann (1984) is available to compute the
intermediate entry reliabilities. (See Appendix B for a derivation of this algorithm.) The
intermediate entry reliability R {i,j)
e is defined as the reliability of a system composed of i
functioning components out of j available with i< j . Given the individual draw point
reliabilities presented in Table 4.1, the intermediate reliabilities R {i,j) e computed by the
recursive algorithm are shown in Table 4.2.
Table 4.1 Draw point reliabilities to compute the entries

Draw Point n
1 0.49
2 0.78
3 0.63
4 0.51
5 0.52
6 0.34
7 0.64
8 0.58
9 0.53
10 0.40

73
 Table 4.2 Intermediate entry reliability table, R (i, j) e

Hel'j) 0 1 2 3 4 5 6 7 8 9 10
0 1.00 0.18 0.04 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00
1 0.00 0.82 0.32 0.10 0.03 0.01 0.00 0.00 0.00 0.00 0.00
2 0.00 0.00 0.65 0.39 0.17 0.06 0.02 0.01 0.00 0.00 0.00
3 0.00 0.00 0.00 0.50 0.42 0.23 0.11 0.05 0.02 0.01 0.00
4 0.00 0.00 0.00 0.00 0.38 0.41 0.28 0.15 0.08 0.04 0.02
5 0.00 0.00 0.00 0.00 0.00 0.29 0.38 0.30 0.20 0.11 0.06
6 0.00 0.00 0.00 0.00 0.00 0.00 0.21 0.33 0.31 0.23 0.15
7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.15 0.28 0.30 0 25
8 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.11 0.23 0 28
9 0.00 0.00 0.00 0.00 0.00 . 0.00 0.00 0.00 0.00 0.08 0 18
10 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 o.oo 0 05

Based on the intermediate reliabilities presented in Table 4.2 the reliability of a 5-out-of-10 draw
point system is computed by adding the rows 5 to 10 of column 10 of Table 4.2. Thus the system
reliability would be 97.9%.

The intermediate entry reliability table is a useful result. This table could allow mine planners to
analyze the effect of different amounts of redundancy added at the crosscut and draw point level
on the overall reliability. Figure 4.7 shows a relationship between draw point redundancy and
crosscut reliability for different draw point reliabilities. Clearly adding draw point redundancy
increases crosscut reliability. However, the increase depends significantly on the individual draw
point reliabilities.

74
 14% 29% 43% 57% 71% 114% 200%

% draw point redundancy within the XC

Figure 4.7 Redundancy versus system reliability

A n aspect that has not been discussed so far in this section is how to select the n planned draw
points within the open draw points available in the crosscut. Usually, this selection will be
facilitated by the status of the draw points, i.e. there will be hang up, over size or repair draw
points at any given time in the crosscut. Nevertheless, it has been found that usually all the
available draw points should be used in a production schedule for all different planning horizons.
This results in a robust estimation of the system reliability. Leaving active draw points out of the
schedule over long periods (more than a week) is not realistic since even draw is required to
minimize dilution entry as well as stresses. However, in daily or shift by shift planning horizons
it is possible to select the draw points that should be in operation. This selection of draw points
could be facilitated by the use of a priority system (Diering, 2004).

75
 3
T3 n dpts with highest reliability
2
CL
o.i4 • —o— n dpts with lowest reliability
0.0 O'
8 10 12 14 16 18

# of planned draw points within the crosscut (n)

Figure 4.8 Impact of draw point selection on crosscut reliability

Figure 4.8 shows the effect on the system reliability of choosing n within the available draw
points for a given number of required draw points (k=7). It is possible to see that if the available
draw points are sorted in descending order of individual reliability, the "highest reliability"
performance is observed. If the draw points are sorted in ascending order of individual reliability,
the "lowest reliability" behaviour is observed. Also, Figure 4.8 shows the impact of different n
on the overall crosscut reliability. This criterion for draw point selection should be used together
with other draw control rules such as angle of draw and maximum differential draw to avoid
isolated draw in the production area of the mine.

4.3.1 Draw point productivity as a function of draw point yield and draw cycle
A n important component of the reliability model is the estimation of the number of draw points
k needed to achieve the crosscut production target. This estimation is performed using the
average nominal draw point productivity within the crosscut. The following will explain how this
parameter is computed.

The draw point productivity is a function of the draw point yield which is defined by the particle
size distribution at the draw point, L H D bucket size and production cycle time. The particle size

76
distribution of a draw point will affect the bulk density of the loose material at the draw point. As
the draw point matures, the secondary fragmentation together with the mixing along the draw
column will lead to a wide distribution of particle sizes which will result in an increase in the
bulk density of the rock. Thus, the draw point yield would be a function of the draw point
maturity as shown in Figure 4.9 which has been constructed from raw data taken from mine M2.
Note that the curve has been constructed without including the hang ups or oversize that may
have occurred at the draw points.

12000

ioooo4 •

9_ 8000

Iz
§, 6000 - |
CD

£ 4000
c
o
Q_
g 2000
CO
O
o-l

10 20 30 40 50 60

Height of Draw (m)

Figure 4.9 Draw point yield as a function of height of draw

Production cycle time also affects draw point productivity. The production cycle time is a
function of the speed of the L H D and the average distance that the L H D has to travel in a time
period. The average distance that the L H D has to travel to achieve a given production call would
be a function of the number of draw points that the L H D has to muck in the call within the
crosscut. If the whole crosscut had to be drawn, the productivity of the L H D will be less than if
only the closest draw points to the ore pass had to be drawn to fulfill the crosscut production call.
Therefore there is an inverse relationship between the L H D productivity and the number of draw
points to be drawn within a crosscut. Consequently a highly productive schedule would tend to
use less draw points to meet the crosscut target, concentrating the mucking activity on those
draw points close to the ore passes or crushers. Nevertheless if just a few draw points are drawn
within a crosscut the draw cycle would increase and trigger convergence as shown in Chapter 3

77
Figure 3.11. For a given amount of acceptable convergence a draw cycle can be computed. This
computed draw cycle can be used to estimate the minimum number of draw points that need to
be drawn in a given period of time using as shown in Figure 4.10.

u
i — ' — i — ' — i — ' — i — ' — i — • — i — « — i — < — i — > — i — • — i — >
0 2 4 6 8 10 12 14 16 18

Minimum Number of Dpts (k)

Figure 4.10 Draw cycle time as a function of k

The draw point yield curve and the minimum number of draw points to be drawn per period will
be used to compute the number of draw points needed to achieve the crosscut production call.
This number is computed per crosscut and per every period of the production schedule during a
given planning horizon.

4.4 Production Scheduling Integrating Draw Point Reliability

It is important to define what is meant by production schedule reliability. The reliability of a

production schedule is defined as the probability of achieving a given production target. This
probability can be assumed to be directly proportional to the probability that there will be enough
mining infrastructure available to produce a given production target. This probability will be
time dependent since the production targets as well as the failure characteristics of mining
infrastructure change from period to period. Thus the reliability of a production schedule will not
be represented by a single number but rather a vector of reliabilities, one for each period of the
schedule. Equation 4.5 illustrates the definition of reliability

78
 R[TT(t)]= P[TA(t)>TT(t)] (4.5)

where /?[7T(r)] is the reliability of achieving the tonnage target TT(t) at time t, TA(t) is a
random variable representing the actual tonnage produced from the mine at a given time. Figure
4.11 shows the expected variation of tonnage distribution for different times during the life of the
mine. Note that the lower the tonnage target the greater the probability of achieving it. From this
figure it is possible to see that the tonnage distributions change over time since the slope of the
curves shown in Figure 4.11 varies from period to period.

Production target 7T(t) (tons/day)

Figure 4.11 Expected evolution of actual tonnage distribution

Consequently the aim of this chapter is to present a method to compute the reliability of a
production schedule together with the methods to compute the tonnage distribution curve
throughout the life of a mine.

The mining system reliability model described in the previous section of this research will be
used to compute the reliability of a given production plan as a function of the reliability of the
mining infrastructure. Figure 4.12 shows the proposed mine planning model including the
reliability model.

79
 10.000 15.000 20,000 25.000 30,000 35.000 40,000
Production target

Figure 4.12 Proposed mine planning model including reliability model

There are two main operational indicators used in the reliability model: the draw point R O C O F
curves and the draw point yield curves. Both of these parameters should be computed using the
operational records of the mine. The derivation of the draw point yield will be explained later.

A prototype production scheduler was constructed to integrate characteristics of the draw points,
production targets, crosscut relationships, ROCOF curves and draw point yields. The parameters
used in the model are presented below

r(r) production target for period t

7](f) tonnage target for crosscut / at time t
K(t) minimum number of crosscuts heeded to achieve the production target at period t
k^i) minimum number of draw points needed in crosscut i to achieve 7](f)
k c
minimum number of draw points needed to achieve a given draw cycle to control
convergence of crosscut /. Computed from relationship shown in Figure 4.10
N(t) available crosscuts at period t
n (t)
t available draw points in crosscut i at period t
M{t) nominal production capacity of a crosscut at period t
d (t)
t average productivity of draw points in crosscut i at period t
r/(t) draw point reliability of draw point j of crosscut i at period t

80
 CT/{t) cumulative tonnage of draw point j of crosscut / at period t
w[cr/ (t)] ROCOF curve for draw point j of crosscut i
, /?.(f) reliability of crosscut i at time t for k (t)draw points needed and n^t) draw points
t

available in crosscut /
f {t)
t Expected production from crosscut i at period t
f(t) Expected total production capacity at period t

The minimum number of draw points to achieve the production target of crosscut i is

fc,.(0 = m a x j - ^ , f c ; j (4.5)

There are n .(f)draw points available and only fc.(f)draw points are needed to achieve the
(

crosscut i production target, 7](f) at time t. The minimum number of draw points to achieve a

desired draw cycle per crosscut,^, is assumed constant during the production schedule to

simplify the model. Nevertheless it is possible that k\ is time dependent reflecting the dynamic

stress conditions during the life of the mine.

The production crosscuts form one k-out-of-n system with non identical components and result in
the crosscut reliabilities, R {t), i = l,---N(t), as a function of the draw point ROCOF curves,
t

'/(')•

Once the crosscut reliabilities,/?.^), have been computed another k-out-of-n system is formed

among the crosscuts. The minimum number of crosscuts needed to achieve the production target,

T(t), is K(t) which is computed using the following procedure:

• Sort crosscuts in ascending order of crosscut production capacity, T^t)

• Add the crosscut production targets, T (t) until the production target T(t) is reached. This
t

determines the number of crosscuts K(t) needed.

Note that this approach is very conservative since it will always compute the worst possible
production scenario. If priorities are used to quantify the minimum number of crosscuts needed

81
to achieve the production target, the crosscuts should be sorted in descending order of priority
before computing K(t).

The final part of the algorithm computes the total system reliability by computing the reliability

of a k-out-of-n system with non identical production crosscuts with reliabilities, R (t), in which
t

there are N(t) crosscuts available and K(t) crosscuts needed to achieve the production target,

T(t). Figure 4.13 shows a scheme of how the different components of the production schedule as

well as the availability of mining infrastructure are linked to compute the production schedule

reliability.

N(t) Crosscuts
Crosscut i

Drawpoint reliabilities r'

n (t) draw points available

i

N(t) available
k (t) draw points required
i

K(t) required
T (t) Tonnage target
t

\
R[TT{t)}

Figure 4.13 Scheme to compute the reliability of a given production schedule

The second output of the reliability model is the expected tonnage f(t). The estimation of the

expected crosscut production, T^t), is performed by embedding a sampling procedure into the k-

out-of-n model. A number of draw points z, k < z < n, in a crosscut (sub indices have been
dropped to simplify the notation) are chosen from the n available draw points using a random
sampling process. The productivity of these z draw points is computed by adding their draw
point productivities. The sampling of the z draw points is repeated a fixed number of times. At

the end the productivity of the z draw points, f (t),

z
is estimated by taking the average of all

82
samplings. This procedure is repeated for all k<z<n. Then the expected production of crosscut

fit)
i, is estimated using the intermediate reliability entries R (z,n) as shown below:
e

f (t) =
i ±V(tX(z,n)
(4.6)
Table 4.3 shows the method to compute crosscut productivity based on reliability measures. Note
that in this example the nominal crosscut productivity was 80,000 tons/month. Then the expected
crosscut productivity should be 80,000x0.4 tons/month assuming a deterministic behavior of the
draw point productivity. Nevertheless by incorporating the stochastic behaviour of tonnage
together with the reliability of different draw point configurations the expected tonnage is
computed to be 29,141 tons/ month as shown below. The former estimation is more realistic
since it represents the stochastic behaviour of draw points active rather than the deterministic
traditional approach.

Table 4.3 Estimation of crosscut production capacity based on reliability estimates

9
n
l = 12
z R (z.n)
e V{t)
9 0.215 67,500
10 0.131 75,000
11 0.049 82,500
12 0.008 90,000
29,141

Finally to compute the expected production capacity of the mine the same procedure as
illustrated in Table 4.3 is followed. In this case the components are the production crosscuts with

estimated productivities T^t). Appendix C shows the details of the data flow related to the

proposed scheduler algorithm shown in this section.

4.5 Production Schedule Reliability

The first application of the reliability model consists of estimating the reliability of a given
production schedule. In order to compute the reliability of a production schedule the draw point

83
ROCOF and draw point yield curves must be computed. Another input for the reliability model
consists of the production schedule which specifies the amount of tonnage to be produced per
period and the draw point opening sequence as shown in Appendix C. The final input for the
reliability model is the crosscut production target per period. These targets can be computed
either from the production schedule or from the maximum haulage capacity of a crosscut. After
setting up the inputs the reliability model is run to compute the reliability per period as shown in
Figure 4.14.

— • — Tonnage target
VZZft Reliability 1.0
1.1M4 ••••••••••••••••a
0.9
no1
900.0k H 0.8
c
o 0.7
E 750.0k H
"35 0.6
c
33
2 600.0k -A

CD 0.5 i:
03
450.0k - CT

CD 0.4 | i
D>
CO 300.0k -
C
c 0.3
o
150.0k
0.2

0.0- 0.1

0.0

Period

Figure 4.14 Assessment of reliability based on a computed production schedule

In the example shown in Figure 4.14 the reliability per period is fairly low during the production
schedule. The variance of the reliability per period is due to the draw point and crosscut
availability per period, which depends on the draw point sequence used in the production
schedule.

Once the reliability of a production schedule is known the model can be used to recommend
changes to the draw point opening sequence and to the production target per period that could
improve the overall reliability during the production schedule. For example Figure 4.15 shows
the reliability for different production targets of a mine, maintaining all the production schedule
related parameters constant. Evidently lower production targets will produce more reliable

84
production schedules as a result of the amount of redundancy contained in the mining system.
The redundancy in the mining system is reflected by the number of draw points and crosscuts
commissioned in a given period of the production schedule.

1.1M-,
•••••••••
1.0M-
/ v w
900.0k - •••••••••r^g^WWVWAAA
c •
/
o 800.0k - • wvfoX*A4cooooooooooo /

E
/ y A' ../
get i(tons/

700.0k -
P w AA' /O O
•• /Y Y J I
600.0k - I yAAA O O O
• ••wC^booooo /

CO
+—>
500.0k - y /<pd
wv^Ado
nage

400.0k - Reliability
•
A O -o-90%
c
o 300.0k - otiod -A-70%
1-
200.0k - -v-50%
-•-20%
100.0k-

0.0-
-r -
—T—
10 15 20 25 30

Period

Figure 4.15 Production schedules including reliability as a target

A n analytical tool resulting from applying the reliability model to production scheduling is the
production forecast as a function of reliability. Figure 4.16 shows an example in which the
reliability model has been used to forecast production capacity. In this case the reliability has
been computed to assess the maximum production capacity at different periods during the life of
the mine.

85
 1.0
28 Ktr^d —•"B=6= 5=g=g=g
n
=

0.9
-o"
30 Ktp®'
v—v-
0.8

0.7
A'
0.6 H
| 0.5 H 32 Ktr/d
A .o-O'
CC 0.4 H
0.3-
34 KJpci
,0"
0.2- 38Ktoer
0.1 - XT'*
0.0 —I—'—r- n—1—i—<—i— —r- n——i—<—i—'—i——i
1 1 1 1

0 1 2 3 4 5 6 7 8 9 10 11

Period

Figure 4.16 Production schedules with integrated reliability

If 85% was the desired reliability for the mine shown in Figure 4.16. The recommended
production profile is shown in Table 4.4.
Table 4.4 Production profile at 85% reliability

Period Target
(vear) (tons/day)
1 28,000
2 29,000
3 30,000
4 31,000
5 32,000
6 32,000
7 32,000
8 33,000
9 33,000
10 34,000

Another analytical tool developed as a result of applying the reliability model to production
scheduling is presented in Figure 4.17, which shows the effect of redundancy on production
schedule reliability, adding redundancy at the crosscut and draw point level. This chart is
important since it allows mine planners to asses the benefits of allocating redundancy at the
crosscut and the draw point level. The ultimate amount of redundancy to introduce in the mining
system can be planned in order to balance the reliability of the production schedule with the
amount of capital cost needed to support the production schedule.

86
4.6 Sensitivity of Different Inputs to the Reliability Model
A sensitivity analysis was performed in order to understand how the reliability model would
behave in different applications. Figure 4.18 shows the reliability profile of a production
schedule for different draw point ROCOF curves. To analyze the impact of different ROCOF
curves in the reliability model, three different curves with constant values of rate of occurrence
of failure 0.1, 0.2 and 0.3 were used to represent the effect of the draw point tendency to failure
in the production schedule reliability. The resulting reliabilities were compared to the reliability
profile using the actual draw point ROCOF curve obtained from mine M l . The production target
was set to be 360,000 tons/month constant during the life of the mine. Figure 4.18 shows that the
R O C O F curve affects not only the absolute reliability value but also the shape of the reliability
profile.

The second parameter that has a tremendous influence on the reliability model is the nominal
productivity of a crosscut. For the same production target as used above and the actual draw
point ROCOF curve of mine M l the crosscut nominal productivity has been changed in the

87
range of 60-150 K tons/day. Figure 4.19 shows the effect of this parameter on the performance
reliability during the life of a mine.

1.0-1

~ o // a d \ ° /
0.9-
0.8-
jO p
0.7 I /
= 0.6'
I 0.5 H
cc
0.4-
/ o Y

0.3-
/ v°"\ -
-o -
ROCOF
ROCOF
©0.1
@0.2
0.2- -A - ROCOF @0.3
0.1 - -v- ROCOF curve

0.0 10 15
Period

Figure 4.18 Sensitivity of the reliability model to the ROCOF curve

A . / 7

- • - T(t)= 150 Ktpm

-o-r/f>=100Ktpm
- A - r/y=80 Ktpm
- v - 7/r>=60 Ktpm

Figure 4.19 Sensitivity to crosscut productivity

88
4.7 Using the Reliability Model to Assess the Value of Overdrawn Draw Points

Overdrawn draw points are defined as those draw points that have depleted their mining reserves
according to the long term model but are still producing grade above the predictions made with
the long term model. This behaviour has been observed at several block cave mines. When this
situation arises at an operation the fundamental question is whether to consider this resource as
part of the short term plans, 2 to 4 months, or simply exclude this material from the planning
model. It is often found that the overdrawn draw points will be included in the plan since this
action will bring some flexibility to the operation which may be under pressure to achieve the
production target. If the overdrawn draw points are considered in the plan, the question is what
strategy should be followed in order to maintain the mine throughput stability. Usually there is a
trade off between a short term increase in production rate versus a medium term steady state
production rate.

Mine M l faced the issue of deciding what strategy to use to treat the overdrawn draw points. The
challenge was to find a production strategy to mine 23 overdrawn draw points in a time horizon
of 9 months. The strategies studied were: mine the 23 draw points at their fastest capacity so
there could be a production increase in the following 2 months, or mine a few of those 23 draw
points at their maximum capacity and save the rest of the draw points to offset an eventual loss
of future scheduled production due to an unstable production crosscut, called X C G U .

The reliability model was used to quantify the number of overdrawn draw points needed to offset
an expected loss in production due to a potential unstable production crosscut. In this example
the draw points located in the potential unstable crosscut have been assigned a reliability of 0.2.
The overdrawn draw points have been included in the plan with a reliability of 0.35. Figure 4.20
shows how the reliability model has been used to evaluate the different options and the
throughput that is supported by every one of these options. The base case scenario is shown by
the line labeled "with X C G U " which shows a reliability of 0.8 to produce 36,000 tons/day. The
reliability of 36,000 tpd drops to 0.65 if X C G U collapses, as shown by the line labeled "without
X C G U " . However, in the event that crosscut X C G U collapses, by adding 23 overdrawn draw
points to the medium term plan the reliability of 36,000 tons/day target increases to 0.9, which is
represented by the line tagged "wo X C G U , 23 acc". In the event that crosscut X C G U collapses

89
and just 10 overdrawn draw points are added to the medium term plan the reliability of a 36,000
tons/day target increases to 0.75, which is represented by the line tagged "wo X C G U , 10 acc".
The analysis presented here was used to support the decision regarding the number of overdrawn
draw points that can be used to increase short term production and maintain the reliability of the
medium term production schedule when faced with the potential collapse of a production area.

wo XC GU, 23 acc

Production target (tons/ day)

Figure 4.20 Assessing the effect of overdrawn draw points in the production schedule

4.8 Using the Reliability Model to Support Tactical Decisions

One of the tactical decisions that affects the productivity of a block cave mine is the allocation of
secondary breakage resources, crews and equipment, to different production crosscuts. This
decision is usually left to operators to accommodate the tasks of production and secondary
blasting. Yet it has been noticed at operating block caves that placing a draw point or a crosscut
on hold for secondary blasting has tremendous production implications (Dessureault et al, 2000).
Then mine planners are starting to include the secondary blasting chart as part of the daily
production call. Nevertheless there has been no study of the best strategies for allocating
secondary breakage resources in an operating environment. A method to allocate secondary
breakage resources based on the production crosscut reliability is proposed.

90
As draw points within a crosscut begin to hang up, the draw point redundancy decreases with the
number of available draw points approaching the minimum required draw points k at which point
the crosscut is sent to secondary breakage and then returned to operation. Figure 4.21 shows how
the reliability model can be used to monitor the reliability of the crosscut as a function of the
draw points reliabilities and the crosscut redundancy. When a crosscut reaches a secondary
breakage threshold it will be put on secondary breakage status.

1.0-1 • — •

0.9-

0.8-

0.7- Optimum Performance

s •
CO 0.6-
"55
CC 0.5-
E
CD
•4—•
03 0.4-
(f)
0.3-

0.2-

0.1 -

0.0-
10 12 14 16
# of Planned Draw Points

Figure 4.21 Reliability model used to prioritize secondary blasting

4.9 Definition of Production Capacity of a Block Cave Mine using Reliability

The traditional approach to define production capacity in block caving has been by constructing
a curve that shows the tonnage profile during the life of the mine. The use of the reliability
model in production planning allows estimation of the uncertainty associated with a production
profile. The reliability profile produced by the reliability model can be interpreted as an
estimator of the uncertainty of a production profile for a given available mining infrastructure. If
several production profiles are simulated and passed through the reliability model, a family of
curves relating tonnage to reliability could be plotted in which a single curve would represent the
expected tonnage distribution of that period. Figure 4.22 shows an example of defining

91
production capacity by introducing the probability of different production profiles during the life
of the mine.

1.0-,
0.9- —•—5 years
—o— 10 years
0.8- —A— 15 years
20 years
0.7-
0.6-

to 0.5-
a> •

CC 0.4-

0.3-
0.2-
0.1 -
0.0- •. • i O O CT
S A
Ur
10.( 15.0k 20.0k 25.0k 30.0k 35.0k 40.0k
Production target

Figure 4.22 Evolution of the production capacity distribution throughout a production

schedule

Figure 4.22 shows the distribution of the production capacity throughout the life of the mine for
a given draw point opening sequence and development rate. However if several planning
strategies are under evaluation the mine planner could be able to plot every strategy as a surface
in which the X axis is the production target, Y axis the time and Z axis the reliability. Figure 4.23
shows the distribution of the production capacity for a given strategy in a surface format.

92
 Figure 4.23 Stochastic definition of production capacity of a block cave mine

Based on Figure 4.23 a production strategy could be defined as a surface rather than a curve.

4.10 Expected Tonnage versus Adjusted Draw Rate to Assess Production Capacity

One possible result of this research could be to use the estimated draw point reliability to adjust
the draw point yield to achieve a more realistic prediction of draw point productivity. In fact at
the moment several of the mines visited as part of the research were using different kinds of
"fudge factors" to adjust draw point productivity, based on the availability of the area, the status
of the draw point and even sometimes "engineering judgment". At some other operations the
final production forecast was adjusted without a clear criterion to accommodate historical
performance. It is important to understand how use of reliability per draw point alone can lead to
an erroneous estimate of mine productivity. Figure 4.24 shows a comparison among the actual
tonnage drawn from mine M2, the expected production capacity from the reliability model and
the expected production capacity using an adjusted draw point productivity.

93
 1.2M n

1.1M
1.0M CC03
0 O -
CO 900.0k
c
o 800.0k O•O n_ ^ ^ A
A

c d t ^ n
A A'A
o 700.0k /
-t—'
o
Z3
600.0k
•o
o 500.0k
400.0k c o 4 n v • A A
300.0k / AQA^ A' A X - • - Expec. tons
200.0k - o — Production fudge
- A — A c t u a l production
100.0k k 7

0.0 i | i—| i | i | i | i | i | i | i | i | i | i |
0 3 6 9 12 15 18 21 24 27 30 33 36
Period

Figure 4.24 Expected tonnage computed using the reliability model and the fudge factor
approach

From Figure 4.24 is seen that by adjusting the draw rate by a factor such as the draw point
reliability the production capacity of the mine is excessively overestimated. There is a much
better agreement between the actual production and the expected production capacity derived
from the reliability model than using the draw point reliability as a fudge factor.

In summary the reliability model produces two main outputs: the reliability of the system and the
expected tonnage per period. The system reliability per period has to do with the minimum
number of days in which the tonnage target would be achieved within the period. The expected
tonnage would be the most likely tonnage that would be produced in a given period of the
production schedule. In the next chapter of the dissertation both of these parameters will be
calibrated using data from mines M l and M 2 in order to validate the approach proposed in this
thesis.

94
5 MODEL VALIDATION
Validation of the model was considered to be significant because it would demonstrate that the
introduction of reliability measures related to mining infrastructure in a production scheduling
algorithm could facilitate the estimation of the system reliability as previously defined in Chapter
4. In order to perform this analysis several methods of validation were tested using the
production databases from mines M l and M2.

The validation process started with the construction of the draw point ROCOF curves of both
mines M l and M2. This construction was performed using the draw point status available from
the operational databases. The original production targets used at mines M l and M 2 were also
estimated using the historical draw point sequence. Then all the inputs: draw point ROCOF
curves, draw point yield curves and original production targets were used to compute the
reliability of these targets. Figure 5.1 shows the data flow to compute the system reliability based
on operational data.

The methods used to estimate the actual reliability obtained at mines M l and M 2 respect to their
original production targets are listed below:
• As the ratio between the actual total production and the original production target per
period.
• As the probability of achieving the production target based on the actual tonnage
distribution curve per period. In this case every period was set up to be 60 days of
production.

The following sections will review in detail the validation process in detail based on estimates of
actual reliability mentioned above.

95
 Dpt. Bathtub
Status Curves
Operational
Database Reliability Model
Dpt. Dpt. • Computed Reliability
Tons Yield • Expected Tonnage

Production Production
Scheduler Targets

Estimated Actual
System Reliability

Figure 5.1 Dataflowto compute the reliability of the original production schedules of
mines M l and M2

5.1 Reliability Model Calibration Using Absolute Reliability

The first method of calibration consisted in estimating the actual reliability as the ratio between
the actual tonnage drawn and the tonnage target at time t

(5.1)
T{t)

The actual reliability, R'(t) was plotted against the reliability computed by the model. The aim

of this first analysis was to estimate how well the computed reliability can represent the

dispersion of the actual tonnage with respect to the plan. This first analysis was performed using

a nominal crosscut production capacity, T cx (t), of 85,000 tons per month. Figure 5.2 shows the

result of this first analysis.

96
 0.0030 -\

0.0025 H

-g 0.0020 - |

DC
•o 0.0015-
a>
CL
E 0.0010-
o
O
0.0005 •

0.0000 •

-1— I—
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Actual Reliability

Figure 5.2 Calibration of computed reliability using a constant crosscut production

capacity

The results shown on Figure 5.2 were very disappointing since there was no correlation between
the actual and computed reliability. However, it was discovered that the nominal crosscut
capacity should not be constant since the number of draw points constructed per crosscut evolves
together with the production schedule. Crosscuts that are not developed fully at the beginning of
the periods used in the calibration will not have the draw points to reach their nominal capacity.
Thus it was decided to introduce a production target per crosscut, T {t) which was computed as a
t

function of the actual number of draw points commissioned at time t. The following equation
shows how the production target per crosscut is computed as

^(0 = (5-2)
Ji

Where h^t) is the number of draw points that have been commissioned at time t in crosscut i, f. t

is the total number of designed draw points in crosscut i, and T (t) is the nominal crosscut
cx

production capacity at time t. Figure 5.3 shows how the calibration is improved by making the
nominal crosscut production capacity variable during the production schedule.

97
 Actual Reliability

Figure 5.3 Calibration of computed reliability using variable crosscut production capacity
across the active layout

5.2 Reliability Model Calibration Using Tonnage Distribution

The second method to estimate actual reliability was based on a production tonnage histogram
which was constructed based on the daily tonnage records. For each mine, M l and M 2 , the
actual daily production records were divided into two month periods in order to have enough
information (60 data points) to construct the tonnage distribution within a period. One year of
Mine M l production data and almost three years of Mine M2 production data were used. The
actual reliability for a given period was computed from the histogram as the probability of
achieving at least the target defined for that period. Figure 5.4 shows a schematic dataflow that
illustrates how the daily production records were used to compute the actual reliability of a
period of 60 days.

98
 Operational
Database
Daily Tonnage 60 days Tonnage Estimated Actual
— ^

Records Records Histogram Reliability

Y. Production
-Dpt. Production
Target for a 60
Sequence Scheduler
days

Figure 5.4 Dataflow used to estimate actual reliability of a 60 days period based on the
tonnage distribution curve

Figure 5.5 shows a relationship between tonnage target and actual reliability for a period of 60
days.

1.0-1
0.9-
0.8-
0.7-
0.6-
-

or 0.5-
0.4-
0.3-
0.2-
0.1 -
0.0-
0.0 5.0k 10.0k 15.0k 20.0k 25.0k 30.0k
Tonnage Target (tons/day)

Figure 5.5 Actual reliability as a function of actual tonnage distribution for a period of 60
days

Table 5.1 shows a comparison between the actual versus computed reliability for a given
production profile using data from mine M l .

99
 Table 5.1 Comparison of actual versus computed reliability

Target Computed. Actual.

Period (tons/day) Reliability Reliability
1.00 35000 0.97 0.84
2.00 37000 0.68 0.74
3.00 39000 0.53 0.62
4.00 41000 0.47 0.41
5.00 43000 0.13 0.22
6.00 45000 0.04 0.09
7.00 47000 0.01 0.01
8.00 49000 0.00 0.01

The validation of the reliability model using actual reliability from the tonnage distribution curve
for mine M l is presented in Figure 5.6. The correlation coefficient for this validation was 0.97
for 8 points.

1.0-1

0.9-

0.8-

>* 0.7-
!5 -
to 0.6-
"55
CC
0.5-
-o
w -

"5 0.4-
E
Q_ -
Co

0.3-

0.2-

0.1 -

0.0-
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Actual Reliability

Figure 5.6 Reliability model calibration using mine M l data set

Similar results for mine M2 are presented in Figure 5.7. The correlation coefficient found in this
comparison was 0.98 for the 40 compared points.

100
 Actual Reliability

Figure 5.7 Reliability model calibration using mine M2 data set

The results shown in Figure 5.6 and Figure 5.7 are very encouraging since it demonstrates that
for the actual mine production from operations M l and M2, the reliability model does reproduce
the reliability obtained at these mines for a given set of production targets.

5.3 Calibration Based On Expected Tonnage per Period

A different way of validating the reliability model was to compare the expected tonnage
computed by the reliability model versus the actual tonnage mined in a given period of time. In
this case the draw point ROCOF curves, the draw point yield curves and the tonnage target per
period are input to the reliability model. The production targets in this case were set to be the
same as the actual tonnage mined per period. In this case the model should produce a reliability
equal to one for every period of the schedule since the production target has been set to be the
same as the actual tons. The expected tonnages are supposed to be greater than or equal to the
actual tons according to the definition of reliability introduced in Chapter 4. Figure 5.8 shows a
comparison between the actual production and the expected tonnage computed from the
reliability model. The correlation coefficient found between expected tonnage and actual tonnage
is 0.87 for 32 periods. There is one outlier (circled) which is the result of a very low production

101
caused by a problem related to an underground crusher. Disregarding this point the correlation
between actual tons and expected tons increases to 0.92.

Expected Tonnage

Figure 5.8 Comparison between expected tonnage versus actual tons mined from mine M2

Figure 5.8 Comparison between expected tonnage versus actual tons mined from mine M2shows
that the shape and the tonnage distribution for the production schedule is reproduced by the
expected tons computed by the reliability model which can be used to forecast mine production
capacity. In an application of the reliability model to a pre-feasibility study of a mine the actual
tons will be unknown. Then the reliability model will have to run until a reliability of 1.0 is
achieved for every period of the production schedule. The expected tonnage profile estimated by
the reliability model will become the best estimate of the production profile for this mine.

Although the calibration results have exhibited some scatter between actual reliability and
computed reliability at this stage of development it is considered that the correlation coefficients
are acceptable given the available input to the model. The model has shown to be a valid tool to
assess the tonnage distribution per period of a block cave mine and could be applied to an
industrial application.

102
6 DISCUSSION AND CONCLUSIONS
6.1 Discussion

A mine plan defines the production promise of a mine operation. Its main outcome, the
production schedule, is a bankable document that will support the ultimate decision of whether to
pursue the mining venture or not. The block cave mining method has gained tremendous
popularity in recent years due to its production capacity and low operating cost. Nevertheless, it
has been observed that the recently developed block cave mines have not been able to reach their
initial production targets. Even though there are several computer based applications that enable
mining engineers to compute a production schedule none of them integrates the fundamental
geotechnical models that sustain the process of caving the rock mass. The lack of geotechnical
models in the planning process leads to unrealistic production targets that do not take into
account the operational upsets such as hang ups, over sizes, wet muck and rock instability that
affect mining infrastructure. As a consequence the real value of the mine is hidden together with
the real potential of the mining system to deliver a given production plan.

The integration of geotechnical models in production schedules would require a reformulation of

the existing scheduling packages. However, even if the current geotechnical models were
integrated there would still uncertainty related to the parameters involved in those models and
their ability to represent the actual rock mass behavior would still exist. The other factor is the
ability of the current geotechnical models to simulate the actual caving behavior in highly
stressed and competent rock masses. Therefore, as discussed in Chapter 3, using the actual
performance of mining infrastructure as the main indicator that reflects the geotechnical events
that take place in a block caving operation seems to be the most appropriate approach based on
the current knowledge and technologies available to plan a block cave mine.

The concept of rate of occurrence of failure (ROCOF) has been successfully implemented to
represent the performance of mining infrastructure as part of the thesis presented in this
dissertation, in particular draw point ROCOF, related to their tendency to fail. The estimation of
draw point reliability has a function of the draw point ROCOF has lead to the conclusion that it
could subsume the effect of geotechnical effects on production performance. The reliability of a
draw point as a function of the Rate of Occurrence of Failure (ROCOF) was investigated and

103
tested with two set of operational databases. It was also found that a more detailed record of the
failures could perhaps enhance the estimation of draw point reliability by introducing time to
repair as a random variable. It was found that the draw point ROCOF curve represents well the
failure behaviour of a draw point during its operating life. It was observed that different draw
point ROCOF curves could represent the behavior of mining infrastructure to different rock
masses and stresses. Therefore for different geotechnical domains different draw point ROCOF
curves could be used to represent the effect of rock mass and stress level on failure performance
of different components of mining infrastructure. In this dissertation, two examples of draw point
ROCOF curves have been presented representing two extremes of geotechnical behavior in block
cave mining. This motivates novel research to develop a comprehensive database that could
support the choice associate to the shape of a draw point ROCOF curve for a given geotechnical
domain. The results of this proposed research could be used by mines facing feasibility studies or
operating mines that do not know the draw point failure behaviour of a given area of the mine.

The initial stage of the draw point ROCOF curve could represent the interaction of a draw point
and the undercut blasting activity since it could reflect the geotechnical events that tend to retard
the undercut blasting. At the moment undercutting usually is seen as an activity independent of
production. However it was shown that the traditional, advanced and pre undercutting methods
show a strong relationship between undercut and production level. Commissioning of draw point
to production will depend upon the undercut front position and its ability to progress in the short
term. Therefore, the burn in region of the curve could be affected by the performance of the
undercutting. If the undercutting is retarding the commissioning of a draw point the starting point
on the curve will be raised. Consequently, it may be possible to include the uncertainty of
undercut construction and stability as part o the production system.

In terms of the way that the ROCOF is estimated using draw point status, tactical problems that
induce draw point closure should not affect draw point status but rather draw point productivity.
There is a trend to think that draw points fail due to tactical decisions such that status is driven
by tonnage. However a draw point that is idle due to tactical reasons is still available to be mined
so this should not affect the shape of the draw point ROCOF curve.

104
Based on the reliability of the mining system components, the reliability of the mining system
was estimated as a measure of the ability of the mining system to produce a given production
target. The reliability of the block cave mining system was defined as the probability of the
system to achieve a given production target at a given time. A k-out-of-n model was constructed
to compute the reliability of the block cave mining system. It was found that the mining system
could be divided into: production areas, production feeders and production crosscuts. A
production crosscut is a k-out-of-n system composed of independent and non identical draw
points. At the same time a production feeder is a k-out-of-n system composed of independent and
non identical production crosscuts. Therefore the block cave reliability system consists of a dual
redundancy model at the draw point and the crosscut level. Every one of the k-out-of-n systems
is solved using a recursive algorithm. Finally the reliability model would produce a reliability
profile associated with a production schedule as an indicator of the ability to achieve a given
production target. The reliability profile would depend on the draw point reliabilities and the
general distribution of redundancy across the active layout.

The reliability center mine planning model provides a tool to compute production targets
integrating the potential infrastructure failures that may affect the viability of achieving a given
production schedule. Redundancy at the draw point and crosscut level has been added as a
planning variable to mitigate the geotechnical uncertainty. Therefore redundancy should be
carefully included in the mining system to maximize reliability at the minimum capital and
operating cost.

Several discussions related to how operators should deal with the redundancy contained in the
mining system have taken place during this research. The actual concept is that by introducing
redundancy in the medium and long term plans, logistical problems will be introduced to the
operation of the plan by having more resources than needed. Consider n-k redundant draw points
in a long term production plan. When the time comes to operate the plan there will be a need to
decide which draw points shall be mined, since there are going to be more draw points than
needed to fulfill the production target. If the reliability model has been properly implemented in
terms of draw point ROCOF curves and draw point yield this should not happen. Redundancy is
introduced in production planning to mitigate uncertainty. Therefore at the time the plan is

105
implemented there shall be just k draw points available since the rest will be unavailable as a
result of the geotechnical events. In the event that there is redundancy in the operation of the
mine an analysis such as the one shown in Chapter 4 needs to be conducted to plan the best
usage of redundancy in terms of delivering the medium and long term plans.

Fudge factors are widely used in the mining industry usually to reconcile the results of numerical
models to the actual performance of the mine. There could be the temptation to use the reliability
of a draw point as a fudge factor to adjust draw point productivity. It was demonstrated as part of
the thesis that this practice should be avoided since the actual productivity of a block cave
depends upon the amount of redundancy built in at the draw point and crosscut level rather than
the simple sum of the adjusted draw point productivity. The relationship between redundancy
and draw point reliability is non linear and varies over time. Consequently the expected tonnage
as a result of the reliability model should be understood as the best estimator of block cave
productivity since it is a function of the reliability model which subsumes the geotechnical model
that sustains the block cave mining system.

The approach to mine planning presented in this dissertation challenges the traditional approach
in which design and planning of a mine operation are seen as individual and independent
processes taking place as part of engineering a mine. The reliability approach to planning assists
mining engineers and geologists to estimate the probability of achieving a given production
schedule. To mitigate the periods of low reliability within the production schedule the mine
planners would introduce more infrastructure such as ore passes, draw points, production
crosscuts as means of redundancy to enhance the system reliability. Therefore the amount of
mining infrastructure to introduce in a plan would be seen as a planning decision rather than a
design decision. Thus when using the reliability approach to planning, mine design becomes part
of the mine planning process.

6.2 Conclusions

A n empirical approach which uses observations of the behaviour of mining infrastructure has
been employed to provide a guide for the planning of such infrastructure so as to attach
reliability to a production schedule. It is believed this is a better measure of future performance.

106
Specific conclusions are as follows:
• The low compliance shown by current production schedules can be attributed to the way
production schedules are computed and the way the fundamental models are integrated in
the process of computing these production plans
• The estimation of the rate of occurrence of failure (ROCOF) of mining infrastructure
based on operational status was shown to be appropriate to represent the geotechnical
events that affect the availability of mining infrastructure.
• The ROCOF of mining infrastructure has been shown to be time-dependent, behaving as
a mechanical component whose performance is well defined by the rate of occurrence of
failure of a draw point.
• The mining infrastructure exposed to different rock mass and stresses environments will
have different ROCOF curves associated with their likelihood of failure.
• The estimation of mining infrastructure reliability was successfully implemented as a
function of the ROCOF. The time to repair was showed to have a larger impact on the
low reliability components than on the high reliability ones.
• The reliability of mining infrastructure was shown to be highly correlated with the
geotechnical events that affect productivity. Consequently, the concept of using reliability
measures to subsume geotechnical events on mining infrastructure was shown.
• Block cave system reliability was defined as the ability of the mining system to achieve
at least a given production target at a given time within the production schedule.
• The production profile was redefined using the reliability profile which was computed as
a result of the block cave system reliability.
• Block cave system reliability was validated using two of the operational databases
obtained as part of the research. The expected tonnage as well as the system reliability
was shown to exceed a correlation coefficient over 0.85 when compared to actual
production performance.

• The reliability of a production schedule is an indicator of the technical uncertainty

contained in the mining system as a function of the balance between unknown rock mass
behavior and redundancy of mining infrastructure.

107
 • The definition of production capacity in a block cave mine goes beyond computing
accurately the draw point productivity. This capacity depends highly on the relationship
between draw point productivity, crosscut production capacity and the amount of
redundancy at the draw point and crosscut level.

6.3 Recommendations

It is important to construct a wide operational database that could store the distribution
parameters for different infrastructure components of a block cave mining system under different
rock mass and stress environments. This is seen as an important step forward in the development
of the theory presented in this thesis. It is envisioned that in the future there would be a set of
empirical charts that could support the estimate of the R O C O F shape depending upon the
geotechnical variables that sustain the mining system.

The reliability method should be used as a main tool to quantify the quality of the reserves
supported by a given production plan. The reliability model could be used to assist engineers to
classify different production schedules according to the international standards that relate mining
reserve estimation. Mining reserves must be classified according to technical uncertainty
involved not only in the resource estimation process but also as a result of the mining process.

A chart as shown in Figure 4.15 could be imported to a financial model to determine the extent
one could optimize reliability by lowering production targets particularly during the ramp up
period of the schedule. This is a fundamental question that needs to be answered in the process of
assessing the value of a block cave operation. Different levels of reliability of a production
schedule will need different amounts of "insurance" (Vergara, 2004) to compensate for the effect
of a potential failure of the original production promise. Thus the value of a block cave mine
must be derived as a function of the geotechnical uncertainty since the nature of the mining
system relies entirely on the rock mass. Consequently the reliability model will facilitate the
estimation of technical uncertainty to be fed into a financial model that could derive the true
value of a caving venture.

108
In this research the concept of redundancy has been used to enhance the reliability of a
production schedule. Nevertheless the research presented in this dissertation does not address the
optimal reliability of a production schedule. There is a clear correlation between redundancy and
production capacity. The production schedules computed using the traditional methods would
disregard the reliability of those schedules leading to an erroneous estimation of the mine value.
The optimal value of a mine should be supported by a financial tool such as real options that
allows the integration of production schedule reliability in order to represent the uncertainty
involved in the production planning process. The optimal production schedule should be such
that allows return to the share holders under a predefined level of uncertainty.

The fact that the raw data used to construct the draw point R O C O F curves and draw point yield
curves contain noise leads to model every one of the inputs of the reliability model as a random
variable with an underlying probability distribution. By integrating the probability distribution of
the inputs of the reliability model could lead to quantify the probability associated to a certain
level of reliability which opens a whole new area of research. It is recommended that further
research should be conducted on this area in order to quantify levels of certainty of different
reliability estimates.

109
REFERENCES
Aguayo A., Campos C , Mansilla M . , Sougarret J., Susaeta A., 2004. L H D versus Mechanized
Grizzly in III panel of Andina. In Proceedings, MassMin 2004, Santiago, pp. 415-420. Santiago,
(Ed: A Karzulovic and M Alfaro).

Araneda O. and Gaete S., 2004. Continuous Modeling for Caving Exploitation. In Proceedings,
MassMin 2004, Santiago, pp. 415-420. (Ed: A Karzulovic and M Alfaro).

Ascher, H . and Feingold, H . , 1984. Repairable Systems Reliability, Marcel Dekker Inc., New
York.

Barraza M . and Crorkan P., 2000. Esmeralda Mine Exploitation Project. In Proceedings,
MassMin 2000, Brisbane, pp. 267-278. The Australasian Institute of Mining and Metallurgy:
Melbourne.

Barraza M . , San Martin J. F., Montecino M . , 2004. Mine Technology and its Implementation and
Control, Reservas Norte - Sub 6, E l Teniente. In Proceedings, MassMin 2004, Santiago, pp.
681-685. (Ed: A Karzulovic and M Alfaro).

Barlow, R.E. Heidtmann, K.D., 1984. Computing k-out-of-n System Reliability, IEEE Trans, on
Reliability, Vol. R-33, Oct, 322 - 323.

Bartlett, P. J., 1992. The Design and Operation of Mechanized Cave at Premier Diamond Mine.
In Proceedings, MassMin 1992, Johannesburg, South African Institute of Mining Metallurgy:
Johannesburg

Barlett, P.J., Nesbitt K., 2000. Draw Control at Premier Mine. In Proceedings, MassMin 2000,
Brisbane. The Australasian Institute of Mining and Metallurgy: Melbourne.

Boland, P.J. and Proschan F., 1983. The Reliability of K out of N Systems. The Annals of
Probability, Vol. 11, pp. 760-764.

110
Brown, E T, 2003. Block Caving Geomechanics. JKMRC Monograph Series on Mining and
Mineral Processing. Vol 3, 515 p. Julius Kruttschnitt Mineral Centre, University of Queensland:
Brisbane.

Cacceta, L and Giannini, L M , 1988. The Generation of Minimum Search Patterns in the
Optimum Design of Open Pit Mines. In Proceedings, Australasian Institute of Mining and
Metallurgy, Vol. 5, N 293, pp. 57-61.

Carew T., 1992. The Casiar Mine Case Study. In Proceedings, MassMin 1992, Johannesburg,
Journal of the South African Institute of Mining Metallurgy: Johannesburg.

Chacon J., Gopfert H . , Ovalle A . , 2004. Thirty Years Evolution of Block Caving in Chile. In
Proceedings, MassMin 2004, Santiago, (Ed: A Karzulovic and M Alfaro).

Chanda E C K , 1990. A n Application of Integer Programming and Simulation to Production

Planning for a Stratiform Ore Body. Mining Science and Technology, Vol. 11, pp. 165-172

De Nicola, R., Fishwick, M . , 2000. A n Underground A i r Blast - Codelco Chile Division E l

Salvador. In Proceedings, MassMin 2000, Brisbane. The Australasian Institute of Mining and
Metallurgy: Melbourne.

Dessureault, S., Scoble, M . , Rubio, E, 2000. Simulating Block Cave Secondary Breakage - A n
Application of Information and Operations Management Tools in Mass Mining Systems. In
Proceedings, MassMin 2000, Brisbane, pp. 893-896. The Australasian Institute of Mining and
Metallurgy: Melbourne

deWolf, V . , 1981. Draw Control in Principle and Practice at Henderson Mine. In Design and
Operation of Block Caving and Sub Level Stoping Mines (Ed: D R Steward), pp. 729-735. Soc
M i n Engrs, A M E : New York.

Ill
Didyk M . and Vasquez, G., 1981. Draw Behaviour in E l Salvador Mine. In Design and
Operation of Block Caving and Sub Level Stoping Mines (Ed: D R Steward), pp. 737-743. Soc
M i n Engrs, A I M E : New York.

Dimitrakopoulos, R., Farrelly, C.T. and Godoy, M.C., 2002. Moving Forward from Traditional
Optimisation: Grade Uncertainty and Risk Effects in Open Pit Mine Design. Transactions of the
IMM, Section A Mining Industry, Vol. 111, pp. A82-88.

Diering T, 2000. PC-BC: A Block Cave Design and Draw Control System. In Proceedings,
MassMin 2000, Brisbane, pp301-335.. The Australasian Institute of Mining and Metallurgy:
Melbourne

Diering T, 2004. Computational Considerations for Production Scheduling of Block Cave Mines.
In Proceedings, MassMin 2004, Santiago, Chile. (Ed: A Karzulovic and M Alfaro).

Doepken, W. G, 1982. The Henderson Mine. In Underground Mining Methods Handbook (Ed:
W A Hustrulid), Vol. 1, pp. 990-997. Soc Min Engrs, A I M E : New York.

Dolipas R. S., 2000. Rock Mechanics as Applied in Philex Block Cave Operations. In
Proceedings, MassMin 2000, Brisbane. The Australasian Institute of Mining and Metallurgy:
Melbourne.

Dotson, J. C , 1966. Reliability Engineering and Its Application in Mining, Mining Congress
/.,pp. 68-78.

Dunbar, W. S., Dessureault S., Scoble M . , 1998. Modeling of Flexible Mining Systems, In
Proceedings Can Inst Min Metall, Annual General Meeting, Montreal, 8p.

Dunlop R and Gaete S, 1995. Seismicity at E l Teniente Mine: A Mining Process Approach. In
Proceedings, 4th International Symposium on Mine Planning and Equipment Selection. October
31 -November 03, Calgary, Canada.

112
Encina V., Fuentes C , Palma M . , 2004. Study of interaction between ellipses of extraction in a
physical model of granular material in 2-D. In Proceedings, MassMin 2004, pp. 141-149. (Ed: A
Karzulovic and M Alfaro).

Esterhuizen G. S., Rachmad L . , Potapov A . V . , Nordell L . K . , 2004. Investigation of Swell

Factor in a Block Cave Draw Column. In Proceedings, MassMin 2004, Santiago, pp 215-219
(Ed: A Karzulovic and M Alfaro).

Esterhuizen, G S, 1994. A program to predict block cave fragmentation. Technical reference and
user guide, version 2.1.

Febrian I., Yudanto W., Rubio E., 2004. Application of Convergence Monitoring to Manage
Induced Stress by Mining Activities at PT Freeport Indonesia Deep Ore Zone Mine. In
Proceedings MassMin 2004, pp. 269-272, Santiago, (Ed: A Karzulovic and M Alfaro).

Flores, G , Karzulovic, A and Brown, E T, 2004. Current Practices and Trends in Cave Mining.
In Proceedings, MassMin 2004, Santiago, (Ed: A Karzulovic and M Alfaro).

Flores, G, 2004. Geotechnical Challenges of the Transition from Open Pit to Underground
Mining at Chuquicamata Mine. In Proceedings, MassMin 2004, Santiago, (Ed: A Karzulovic and
M Alfaro)

Gershon, M , 1987. Heuristic Approaches for Mine Planning and Production Scheduling. Int J of
Min and Geol Eng, Vol. 5, pp. 1-13.

Glazer, S. and Hepworth, N . , 2004. Seismic Monitoring of Block Cave Crown Pillar - Palabora
Mining Company, R S A . In Proceedings, MassMin 2004, pp 565-569, Santiago, (Ed: A
Karzulovic and M Alfaro).

113
Guest A . R., Van Hout G. J., Von Johannides A . , Scheepers L . F., 2000. A n Application of
Linear Programming for Block Cave Draw Control. In Proceedings, MassMin 2000, Brisbane.
The Australasian Institute of Mining and Metallurgy: Melbourne.

Heslop T. G. and Laubscher, D. H . , 1981. Draw Control in Caving Operations on Southern

African Chrysotile Asbestos Mines. In Design and Operation of Block Caving and Sub Level
Stoping Mines (Ed: D R Steward), pp. 755-774. Soc M i n Engrs, A I M E : New York.

Hoyland, A., and Rausand, M . , 1994. System Reliability Theory: Models and Statistical Methods.
The Norwegian Institute of Technology. Wiley New York.

Hustrulid, W., 2000. Method Selection for Large-Scale Underground Mining. In Proceedings,
MassMin 2000, Brisbane, pp. 29-56. The Australasian Institute of Mining and Metallurgy:
Melbourne.

Kear, R. M . , 2000. The Use of Evaluation Surfaces to Assist in the Determination of Mine
Design Criteria. In Proceedings, MassMin 2000, Brisbane, pp. 29-56. The Australasian Institute
of Mining and Metallurgy: Melbourne.

Kajner, L . and Sparks G., 1992. Quantifying the Value of Flexibility When Conducting
Stochastic Mine Investment Analysis, Bulletin Can Inst Min Metall, 85, NO 964, pp. 68-71

Kaufmann, A . , Grouchko, D. & Cruon, R., 1977. Mathematical Models for the Study of the
Reliability of Systems. New York: Academic Press.

Kazakidis, V . N . and M . Scoble, 2002. Accounting for Ground-Related Problems in Planning

Mine Production Systems, Mineral Resources Engineering, Vol. 11, N 1, pp. 35-57.

Kazakidis, V . N . , and Scoble, M . , 2003. Planning for Flexibility In Underground Mine

Production Systems, S M E Publications Dept. prior to Nov. 30, 2003.

114
Krantz, D., and Scott, T., 1992. Hard Rock Mining: Method Selection Summary, SME Mining
Engineering Handbook 2nd edition.

Kraushacer, D. E., 1987. Long Term Planning of Coal Mines by Statistical Evaluation of
Databases in Existing Mines. In Proceedings APCOM 1987: International Symposium on the
Application of Computers and Operations Research in the Minerals Industries.

Kumar, U . and Granholm S., 1988. Reliability Technique - A Powerful Tool for Mine Operators.
Mineral Resources Engineering, Vol 1, No 1, pp. 13-28.

Kvapil R., 1965. Gravity Flow of Granular Materials in Hoppers and Bins. Part I and II, Int. J.
RockMech. Mining Sci. Vol 2, pp.35-41.

Lakner, A . A . , and Anderson, R. T. 1985. Reliability Engineering for Nuclear and Other High
Technology Systems. Elsevier Applied Science, London.

Laubscher, D. H . , 1990. A Geomechanics Classification System for the Rating of Rock Mass in
Mine Design. Journal of the South African Institute of Mining Metallurgy, V o l 90, No 10, pp
257-273.

Laubscher, D. H . , 1994. Cave Mining- The State of the Art, Journal of the South African
Institute of Mining Metallurgy, Vol. 94, pp. 279-292.

Lewis, R. S. and Clark, G. B., 1964. Elements of Mining, 3 edition, 768 p. John Wiley & Sons:
rd

New York

McClennan F. W., 1930. Mining Copper Co. Method of Mining Low-grade Ore Body, Vol 91 p
39.

115
McKinnon, S D and Lorig, L J, 1999. Considerations for Three Dimensional Modelling in
Analysis of Underground Excavations. In Distinct Element Modelling in Geomechanics (Ed: V
W Sharma, X R Saxena and R D Woods), pp. 145-166. Oxford and IBH Publishing: New Delhi.

Moss A . , Russel F., Jones C , 2004. Caving and Fragmentation at Palabora: Prediction to
Production. In Proceedings, MassMin 2004, Santiago, pp. 585-590. Santiago, (Ed: A Karzulovic
and M Alfaro).

Moyano A . F. and Vienne J. C , 1993. Remote Operation of L H D and Pick Hammers in Rock
Burst Conditions, Innovative Mine Design for the 21 st
Century. In Proceedings of the
International Congress on Mine Design, Kingston, Ontario, A A Balkema, Rotterdam, pp 579-
591.

Owen, K . C. and Guest, A . R., 1994. Underground Mining of Kimberlite Pipes. In Proceedings,
XV CMMI Congress (ed: H W Glen), 1, pp 207-218. South African Institute of Mining
Metallurgy: Johannesburg.

Prasetyo R., Flint D. C , Setyoko T. B., Samosir E., 2004. Use of the Modular Dispatch System
to Control Production Operations at the D O Z Block Cave Mine. In Proceedings, MassMin 2004,
Santiago, pp. 696-701. Santiago, (Ed: A Karzulovic and M Alfaro).

Peele, R., 1941. Mining Engineers' Handbook, 3 edition. John Wiley & Sons: New York.
rd

Rahal, D., Smith, M . , van Hout, G., and von Johannedis, A . , 2003. The Use of Mixed Integer
Linear Programming for Long-Term Scheduling in Block Caving Mines. In Proceedings,
APCOM 2003: 31st International Symposium on the Application of Computers and Operations
Research in the Minerals Industries, May 14-16, 2003, Cape Town, South Africa.

Ramani, V . R., Bhattacherjee A . , Pawlikowski R. J., 1989. Reliability, Maintenance, And

Availability Analysis of Longwall Mining Systems. Mineral Resource Engineering, Vol 2, No 1,
pp. 3 - 1 7 .

116
Rech W. and Keskimaki, K . W., Stewart D. S., 2000. A n Update on Cave Development and
Draw Control at the Henderson. In Proceedings, MassMin 2000, Brisbane, pp. 495-505. The
Australasian Institute of Mining and Metallurgy: Melbourne.

Riddle, J., 1976. A Dynamic Programming Solution of a Block - Caving Mine Layout. In
Proceedings APCOM 1976: International Symposium on the Application of Computers and
Operations Research in the Minerals Industries.

Rojas E., Molina R., Bonani A., Constanzo H., 2000. The Pre-Undercut Caving Method at the E l
Teniente Mine, Codelco Chile. In Proceedings, MassMin 2000, Brisbane, pp. 261-266. The
Australasian Institute of Mining and Metallurgy: Melbourne.

Ross J. G. et al., 1934. Block Caving at King Asbestos Mine, Quebec, Trans Can Inst Min &
Met, Vol 37 p 209. Ibid, Vol 39 (1936) pp. 441.

Rubio, E , Caceres, C, Scoble, M . , 2004. Towards A n Integrated Approach To Block Cave

Planning. In Proceedings, MassMin 2004, Santiago, pp. 128-134. Santiago, (Ed: A Karzulovic
and M Alfaro).

Rubio, E , Diering, T., 2004. Block Cave Production Planning Using Operations Research Tools.
In Proceedings, MassMin 2004, pp. 141-149. (Ed: A Karzulovic and M Alfaro).

Rubio, E , Dunbar W S, Scoble M , 2001. Scheduling in Block Caving Operations Using

Operations Research Methods. In Proceedings Annual General Meeting, Can. Inst. Min &
Metall.: Montreal, Quebec, Canada, May 2001.

Russel, F. M . , 1987. Application of A PC-Based Network Analysis Program to Mine scheduling.

In Proceedings APCOM 1987: International Symposium on the Application of Computers and
Operations Research in the Minerals Industries.

117
Rushdi, A . M . , 1987. Efficient Computation of k-out-of-n System Reliability, Reliability
Engineering, Vol 17, 1987 Feb, pp. 157-163

Samis, M . and R. Poulin, 1998. Valuing Management of Flexibility: A Basis to Compare the
Standard D C F and M A P Valuation Framework. Bulletin Can. Inst. Min. Metall, Vol. 91, No
1019, pp.69-74.

Singh, A . and Skibniewski, M . J., 1991. Development of Flexible Production Systems for Strip
Mining. Mining Science and Technology, Vol.13, pp. 75-88.

Smith M L , 1999. The Influence of Deposit Uncertainty on Mine Production Scheduling,

International Journal of Surface Mining, Vol. 13, pp.173-178

Summers, J., 2000. Analysis and Management of Mining Risk. In Proceedings, MassMin 2000,
Brisbane, pp. 63-82. The Australasian Institute of Mining and Metallurgy: Melbourne.

Tobie, R. L . and Julin, D. E. 1982. Block Caving: General Description. Underground Mining
Methods Handbook, pp.967-972. S M E Mining Engineering.

Tota, E. W., 1997. Northparkes Mines - New Performance Benchmarks for Underground
Mining. In Proceedings 1997 AusIMM Annual Conference, pp. 57-64 (The Australasian Institute
of Mining and Metallurgy: Melbourne).

Trigeorgis, L., 1990. A Real Options Application in Natural Resource Investments. Advances in
Future and Options Research, Vol 4, ppl53-164.

van Hout G., Allen S., Breed M . , Singleton J., 2004. Status of Draw Control Practice and Waste
Management at Cullinan Diamond Mine. In Proceedings, MassMin 2004, pp. 491-497. (Ed: A
Karzulovic and M Alfaro).

118
Verdugo R., Ubilla J., 2004. Geotechnical Analysis of Gravity Flow During Block Caving. In
Proceedings, MassMin 2004, pp. 141-149. (Ed: A Karzulovic and M Alfaro).

Vergara, L , 2004. Personal communication. E l Teniente Codelco - Chile

Wilke, F L , Mueller, K , and Wright, E, 1994. Ultimate Pit and Production Scheduling
Optimization. In Proceedings, APCOM1994, pp 499-505 (IMM: London).

Xiaotian S., 1989. Caving Process Simulation and Optimal Mining Sequence At Tong Xuang Y u
Mine, China. In Proceedings APCOM 1989: International Symposium on the Application of
Computers and Operations research in the Minerals Industries.

119
APPENDICES

A Proposed Operational Database in Block Cave Mines

The operational database stores a collection of production and monitoring data that is recorded as
the operation of the mine takes place. Usually this database is composed of different component
such as the production module, equipment module, and geologic module. The production module
stores tonnages drawn from the draw points per day, the ore passes and haulage system activity,
secondary breakage activity, draw point status. The equipment module stores the location of
equipment every hour, the equipment availability, maintainability indexes. The geologic module
stores draw point sampling, geologic mappings, geotechnical monitoring. In this research the
most relevant component of the operational database consists of the production module that
contains tonnages records and draw point status.

The production module of the operational database will be used for two different purposes in this
research; the first purpose would be to compute a set of key indicators related to caving
performance such as, draw points, ore passes, cross cuts, the second purpose would be to
compute the rate of failure occurrence of the mining components.

The key caving performance indicators have been defined in order to summarize the status of the
caving. It is proposed that this performances would be used to compare a given block cave
against its initial feasibility study or other block caves in order to explain production drops or
any special unexpected behaviour happening in the operation. A summarized list of the key
performance indicators recommended to be derived out of the production database is presented
as follows:
• Production performances, this involves curves of tons versus draw points, production
crosscuts, ore passes, crushers, trains, etc. Also these indicators are usually measured in a
time range such as shift, day, week, month or year, then they are linked to cave maturity
(height of draw), geology, structural patterns, etc.
• Fragmentation, this involves curves that could represent the evolution in time of hang ups
and oversize. Usually these indicators will be cross checked against production
performances.

120
 • Operational status, store different kinds of status associated to the draw points such as
hang up, oversize, not used, wet muck. Often is found that just the change on status is
stored rather than the daily draw point status. Also the crosscut, ore passes, crushers and
trains status is often missing from the production database. This piece of data is
fundamental to compute the reliability of the mining components.

• Mean time to failure and mean time to repair are both indicators derived from the
reliability system theory that can be used to analyse the performance of secondary
breakage activities. Nevertheless the most important application of these indicators is the
relationship between mean time to failure and the geotechnical factors triggering the
failures.
• Draw point priority is a historical record of the sequence how the draw points were
drawn. This is important to back analyse the relationship between this sequence and the
efficiency of secondary breakage.

A representation of the workflow to derive mining components reliability from the operational
database is presented in Figure A . l .

System Reliability Model

Expected
Key Caving Productivity of the
Performance Indicators Components
- Production performance - Tons/ dpt
- Fragmentation - Tons/ Xcut
Operational - Operational status - Tons/ OP
Database - Mean time to failure
Components
- Mean time to repair
Reliability
- Draw point priority
- dpt
-Xcut
-OP

Figure A.l The reconciliation model

121
It is interesting to note that most of the mines visited did not have a comprehensive design of
their operations database including tonnage records, draw point status, hang up frequency,
convergence, stress monitoring, equipment information, etc. The operations database was
evolving as new data was collected. Also it was noted that some of the important data such as
explosive consumption and undercut records were kept in spreadsheets and often that
information was corrupted and consequently useless. Usually there were no records regarding
repair efficiencies such as time and resources allocated to infrastructure repair or even cross cut
rework. Almost all the operations visited had a draw point status classification for however this
classification was usually useless or difficult to interpret. At some operations for example the
draw point status was recorded when there was a change on it. At some other operations the draw
point status was measured using the Modular Mine dispatch system which did not record
accurately enough the factors happening at the draw points, mainly because the secondary
blasting crews and equipment were not connected to the dispatch system. The Tamrock
automated dispatch system also did not record the status of the draw point accurately since they
were looking at the L H D mechanical behavior rather than the mining system. Therefore there is a
fair amount of work to be done in order to establish a comprehensive, dynamic and a centralized
database to store monitoring data.

122
B Recursive Algorithm for the Reliability of a k-out-of-n System
The recursive algorithm to compute a k-out-of-n system begins by computing the intermediate

reliability entries R {ij)

e which are defined as follows:

R {i,j)=qjR {iJ-l)+PjRe{i-l>J-i)
e e 0<i<n,0<j<n (B.l)

The boundary conditions applied to B . l are shown below

R {0,0) = l;R {-lj)
e e = R {j + hj) = 0
e 0<j<n (B.2)

The derivation of the recursive algorithm considers the generation function presented by Barlow
and Heidtmann (1984)

g (z) =
n I! k + P z) =
t fl K (i, n)z l
(B.3)
i=l i=0

The above generating function can be rewritten as the following recursive relationship

SjM) = fi{q, + p z)^R {iJ-l)z

i e
i
(B.4)
i=l /=o
From B.3, gj (z) = + p z)g _
j j l (z) and the right hand side of B.3 can be rewritten as follows:

M=fe+/^)ix(u-iy (B.5)
1=0 /=0

By introducing the terms inside the total sum of Equation B.5 the following equation can be
derived:

i R {ij)z'=£[q R {iJ-l)+PjR {i-lj-l)y

i e J g t (B.6)
i=0 ;=0

R {i>j) is derived by comparing both sides of Equation B.6. Rushdi (1987) directly linked the
e

intermediate k-out-of-n probabilities, R (i,n), with the total reliability as follows:

e

/?(*,n) = l X M (B.7)

To better understand the above formulation an analytical derivation of the recursive algorithm
will be explored in the following sections. Consider the definitions shown below.

p = reliability of component i
t

q , = 1 - p = unreliability of component i
l t

123
First a system of 2 components functioning out of 3 available will be computed as follows:

R (2,3) = p p q
e x 2 3 + p q p +q p p
i 2 3 l 2 3 (B.8)

By grouping the terms Equation B.8 can be written as follows:

R
e( 3)={Pi 2+<llP2)P3
2
a
+
PlPi 3 a
- ( -)
B 9

Compute the exact probabilities of having 1 and 2 components functioning out of 2 available as
shown below

R {l2)=p q +q p
e l 2 l 2 (B.9)

R {2,2)= p
e Pl 2 (B.10)

By substituting Equations B . l l and B.10 in B.9, the exact probability of 2 components

functioning out of 3 available can be rewritten as shown below

R fa) e = R (l,2)pe 3 + R {2,2)q e 3 (B.ll)

Equation B.12 suggests a recursive algorithm to compute the exact probability of having i
components functioning out of j available with i< j:

R(i, j) = R {i, j - l)qj + R (/ - 1 , j - l)

e e Pj (B.12)

Mathematical induction will be used to show the above recursive algorithm. First the algorithm
will be tested for j =1

R {l,l) = R {l,0h + K(0,0)

e e Pl (B.13)

By definition R (l,0) = 0 and R (l,i}= p , which is the trivial solution when there is only one
e e x

component available with reliability p . If the probability of having i components functioning

{

out of n available, R (i,n),

e is known the induction will be used to construct the R (i,n + l) e

scenario. The j = n +1 scenario means that an extra component is added to the system as means

of redundancy. Therefore this component n+l could be either functioning or broken.

The probability of the subsystem composed of n+l components in which the n+l component is

broken and there are / components functioning can be written as R (i,n)q .

e n+l

The probability of the subsystem composed of n+l components in which the n+lst component is

functioning can be written as R (i-l,n)p where

e n+l R (i-l,n)
e is the probability of having M

124
components functioning out of n available since the n+l component is functioning. Then the
probability of having i components out of n+l available can be written as

R {i,n +1) = R {i,n)q

e e n+l + R {i - l,n)p
e a+l (B.14)

Equation B.15 is the same as applying the recursive algorithm shown in Equation B.13 to the
problem of finding the exact probability of having / components functioning out of n+l
available.

The table below shows how the recursive algorithm works in matrix form.

0 1 2 3 4

1 q1 q1q2 q1q2q3 q1q2q3q4

0

q1q2p3+(p1q2+q1p2) q1q2q3p4+(q1q2p3+[
P1 p1q2+q1p2
q3 1q2+q1p2]q3)q4
P

1
[p1p2q3+(q1p2+p1q2)

•
p1p2q3+(q1p2+p1q2)
0 p1p2 P3]q4+[q1q2p3+(p1q2
3
2
P
+q1p2)q3]p4

p1 p2p3q4+[p1 p2q3+(
0 p1p2p3
q1p2+p1q2)p3]p4
3

0 0 0 p1p2p3p4
4

Figure B.l Intermediate reliability calculation

The above recursive algorithm was programmed to test whether or not it would be suitable to be
used to compute the reliability of a crosscut and consequently the block cave mining system. The
individual draw points reliabilities presented in Table B . l were used to test the initial algorithm.
The resulting intermediate entries reliabilities are shown in Table B.2.

125
 Table B.l Draw point reliabilities to compute the entries R (i, j) E

Dpt Ri
1 0.82
2 0.79
3 0.77
4 0.76
5 0.75
6 0.73
7 0.73
8 0.72
9 0.70
10 0.67

Table B.2 Intermediate entry reliabilities R (i, j) E

H (lj)
e 0 1 2 3 4 5 6 7 8 9 10
0 1.00 0.18 0.04 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00
1 0.00 0.82 0.32 0.10 0.03 0.01 0.00 0.00 0.00 0.00 0.00
2 0.00 0.00 0.65 0.39 0.17 0.06 0.02 0.01 0.00 0.00 0.00
3 0.00 0.00 0.00 0.50 0.42 0.23 0.11 0.05 0.02 0.01 0.00
4 0.00 0.00 0.00 0.00 0.38 0.41 0.28 0.15 0.08 0.04 0.02
5 0.00 0.00 0.00 0.00 0.00 0.29 0.38 0.30 0.20 0.11 0.06
6 0.00 0.00 0.00 0.00 0.00 0.00 0.21 0.33 0.31 0.23 0.15
7 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.15 0.28 0.30 0 25
8 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.11 0.23 0 28
9 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.08 0.18
10 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0b

Finally to compute the k-out-of-n system reliability, the rows k to n of column n are summed up.
For example for the system 5-out-qf-10 the system reliability is computed by adding up the
shaded rows of Table B.2 resulting in a reliability of 97.9%.

126
 /

C Proposed Production Scheduler Workflow

A flow chart of the proposed procedure to compute the system reliability embedded in the
production schedule is shown in Figure C . l .

Production Scheduler Reliability Model

Inputs Procedures

Draw Point Draw Point Read

List [
Yield Parameters

Tonnage
j
i Bathtub
ZE:
Dp_rel
Matrix i Curves

Status
j
Draw Point
::::x::::
Pb_tons
Matrix Database
::::x;;::
k_dp
Production
Targets
Rel_CX
Control
Parameters
k cx

I
System Reliability

Figure C.l Process flow for the proposed scheduler including reliability

The components of the proposed system to compute block cave production schedules reliability
are separated into two categories components on the production scheduler side and components
on the reliability model

127
Draw point list corresponds to the list of the draw points including their names,
crosscuts at which they belong, default draw rate, priority based on the draw strategy.
Figure C.2 shows an example of the draw point list generated on the production scheduler
Draw PoinlXC
P13-01E P13
P13-01W P13
P13-02E P13
P13-02W P13
P13-03E P13
P13-03W P13
P13-04E P13
P13-04W P13

Figure C.2 Draw point list used in the reliability model

Status matrix corresponds to the detailed draw point/period report of status as an output
of a production schedule simulation. Every row of this report represents a draw point and
every column a period of the production schedule. Figure C.3 shows an example of the
status matrix generated as an output of the production scheduler.
Draw PointMay,03 Jun,03 Jul,03 Aug,03 Sep, 03 Oct,03 Nov,03 Dec,03
L01W01 iC C C C A A A A
L01W02 :C C C C C A A A
L01W03 C C C C A A A A
L01W04 !C C C c C A A A
L01W05 C C C c C c A A
L01W06 iC c" c c A A A A
L01W07 iC c c c C A A A |
L01W08 !C c c c A C A A ' '
L02E01 C c c A A A A A
L02W01 C c c A A A A A
L02E02 C A c A A A A A
L02W02 ]A C c A A A A A
L02E03 C C c A A A A A
L02W03 C c c A A A A A
L02E04 C 'c c" A A A A A
L02W04 A c c A A A A A
L02E05 A c c A A A C A
L02W05 C c c A A A A A
L02E06 C c c~ A A A A A
L02W06 A c c A A A A A
L02E07 |C c c A A A C A '

Figure C.3 Status matrix used in the reliability model

Tonnage matrix corresponds to the detailed draw point/ period report of tonnage as an
output of production schedule simulation. Every row of this report represents a draw
point and every column a period of the production schedule. Figure C.4 shows an
example of the tonnage matrix generated as an output of the production scheduler.

128
 Draw Point Name Oct,02 Nov, 02 Dec, 02 Jan,03 Feb, 03 Mar,03 Apr, 03
L01W01 0 0 0 0 0 0 0 0!
L01W02 0 0 0 6 0 0 0 0
L01W03 0 0 0 0 0 0 0
L01W04 0 0 0 0 0 0 0 J3
0
L01W05 0 0 0 0 " 0 0 0 0!
L01W06 0 0 0 0 0 0 0 0
L01W07 0 0 0 0 6 0 0 0
L01W08 0 0 0 0 0 0 0 0
L02E01 0 0 0 0 0 0r 5 0
L02W01 0 421.3821 0 386.6383 0 0 0 0
L02E02 0 0 0 322.1986 37.13389 0 0 6
L02W02 0 144.157 2886.905 4216.199 0 244.8 ^73^2407 47.51491
L02E03 0 0 0 211.7305 0 0 10.19063 0
L02W03 0 44.35601 1931.679 2899.787 0 1020 0 6
L02E04 0 0 0 4354 284 0 0 0 6
L02W04 0 1053.455 774 7942 386.6383 0 367.2 81.52501 38.01193
L02E05 0 0 0 478.695 0 0 0 114.0358'
L02W05 170.7322 898.2092 2504.814 193.3191 0 1060.8 30.57188
' °]
L02E06 0 0 0 0 0 10.2 0 0|

Figure C.4 Tonnage matrix used in the reliability model

There are two categories of components on the reliability model of the proposed system: input
data and procedures to actually compute the system reliability.
• Draw point yield corresponds to the curve that defines the maximum productivity of a
draw point as function of its maturity, measured in cumulate tonnage drawn to date.
Several draw point yield curves can be input in the reliability model to represent the swell
factor and fragmentation characteristics of different sectors of the mine. Figure C.5 shows
an example of a yield curve
Dpt Maturity Yield
0 7,500
2,493 7,500
8,774 7,500
14,082 7,500
19,416 7,500
24,939 7,500
30,546 7,500
36,060 ' 7i500
41,605 7,500
46,960 *~ 7,500
52,753 7,500
58,084 7,500
63,275 7,500
109,732 7,500
200,000 7,500

Figure C.5 Draw point yield curve used in the reliability model

129
• Draw point ROCOF curve corresponds with the evolution of the R O C O F over time for
a given draw point or group of draw points. Different areas of the mine may have
different ROCOF curves as it has been shown in previous section of this dissertation.
Therefore the system allows several ROCOF curves to be input as part of the initial set
up. Figure C.6 shows an example of ROCOF curve used as part of the reliability model.
Dpt Maturity Fail rate
0 0.59
10000 0.59
20000 0.53
30000 0.50
40000 0.48
50000 0.47
60000 0.46
70000 0.45"
80000 0.44
90000 0.43
100000 0.43
110000 0.42
120000 6.42"
1300001
0.41
1400001
0.41
150000 0.40

Figure C.6 Draw point ROCOF curve used in the reliability model

• Draw point database corresponds with the integration of the draw point list generated as
part of the production scheduler, cumulative tonnage before starting the run per draw
point, draw point ROCOF curve, draw point yield curve per draw point, default ROCOF
per draw point. Figure C.7 shows an example of how the draw point information is
integrated to make it available for the reliability model.
Draw Poin XC l CumTJan02 Priority ROCOF Status Tons Bathtub Yield |
P13-01E P13 ]_ 1 0 1
• 0 7 1 278 0.16 A 8000 1 T]
P13-01W P13 | 91,177 279 0.18 A 8000 1 1 j
P13-02E P13 j 109,168 280 0.15 A 8000 1 i;
P13-02W P13 j_9j^541 281
1
0.18 A r _
8000 1 1!
P13-03E P13 |____4 282 0.20 8000 1 il
P13-03W P13 I 125,169 283 0.24 A 8000 1 1|
P13-04E P13 | 157,887 284 0.37 C 8000^ 1
P13-04W P13 j_ 145,910 285 0.40 C 8000 1 1
1

Figure C.7 Draw point database used as part of the reliability model

130
 • Production targets matrix corresponds to the total tonnage target per period and the
production per crosscut as percentage of the total tonnage per period. Figure C.8 shows
an example of the production targets imported in the reliability model.
Period j Target P13 P14 P15 P16 P17
1 1.085,000 0.03 0.02 * '. o o T 0.02 0.03
2 1,147,000 0.03 0.02 0.01 0.02 0.03
3| 1,209,000 0.03 6.02 0.01 0.02 0.03
4| 1,271,000 0.03 0.02 0.01 0.02 6.03
5! 1,333,000 0.03 0.02 0.01 0.02 ' 0.03
61 1,395,000 0.03 0.02 o.or 0.02 0.03
n 1,457,000 0.03 I bT02 0.01 0.02 6.03
8l ' 1,519,000 0.03 l_ 0 0 1
0.02 0.03

Figure C.8 Tonnage targets per period used in the reliability model

• List of Control parameters corresponds with the settings to achieve even draw
performance such as the minimum number of draw points per crosscut per call, periods to
compute reliability, nominal tonnage production per crosscut, number of draw point
ROCOF curves, number of draw point yield curves, switch to use or not production per
crosscut instead of the nominal production per crosscut, switch to run past historical
tonnages or forward estimation.

• Read parameters corresponds with the procedure to actually read all the input
information and store it in different arrays.

• Dp_rel corresponds with the procedure to compute draw point reliability based on
ROCOF

• Pb_tons corresponds with the procedure to compute crosscut production target based on
proportions

• k_dp corresponds with the procedure to compute the number of draw points per C X to
achieve crosscut production target

• Rel_CX corresponds with the procedure to compute crosscut reliability

• k_CX corresponds with the procedure to compute the number of crosscuts to achieve
total production target

• R_Sys corresponds with the procedure to compute the system reliability

The procedure to compute crosscut reliability and system reliability is the same since both
systems are k-out-of-n non identical components. It is important to note that as a result of

131
Rel_CX and R_Sys procedures the reliability as well as the expected tonnage of the system is
computed. For the general case of crosscut reliability the Rel_CX procedure simulates a number
of different scenarios of i draw points working out of the n available. In fact there is an input
parameter called samples, which corresponds with the number of random numbers used to
compute the most likely crosscut productivity.

132