Simulation Modelling of Mining Systems Massmin 2000

Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Simulation Modelling of Mining Systems

B E Hall1
ABSTRACT
This paper discusses the requirements for a successful simulation
modelling study, and the steps involved in the modelling process.
Advantages and disadvantages of simulation are described, and a case
study highlights how not to go about it. Two other case studies, for a
decline truck haulage and a block caving production operation, are used
to illustrate the types of variables whose impact may be investigated.
Typical results are presented to indicate the sorts of presentation styles
the author has found useful for conveying study results to project
sponsors. Standard spreadsheet charting techniques can be very useful in
this respect, but the possibilities are limited only by the imagination of
the study team.

INTRODUCTION
Simulations of mining operations have been carried out for many
years. In many cases, end-users of simulation projects have been
disappointed with the results obtained. Often this is because the
strengths and weaknesses of the tool have not been fully
understood, and the aims of the simulation project were not
clearly defined.
This paper discusses the requirements for a successful
modelling study of any kind, but with particular reference to
simulations. It considers when simulation is and is not the
appropriate tool to use, looking at some of its advantages and
disadvantages.
Examples of simulations of (a) block caving production
operations, from drawpoints to final delivery of ore, and (b)
operation of a truck haulage fleet in an underground mine, from
the source of the rock to its final destination, are described.
These illustrate the types of parameters which might be
investigated, and how results can be presented to facilitate
decision making based on sound engineering processes.

AN OVERVIEW OF MODELLING
There are many types of models that can be built, such as
physical models, computer spreadsheet models, and animated
discrete event Monte Carlo simulation models. The last of these
is the main topic of this paper. This description indicates that this
type of simulation takes account of both random variation and
changes in the system being modelled over time, and can be
animated to provide a visual appreciation of the behaviour of the
system.
All models have certain common features. They are
approximations of the real thing, and are built with a particular
purpose in mind. They should produce a good representation of
the particular aspects of reality relevant to their purpose, but are
unlikely to be good representations of other aspects of the
systems that have not been modelled.
The validity of a model is determined by the accuracy of its
outputs, not by the reality of its inner workings. A very accurate
representation of reality is a replica, not a model, and a lot more
time and effort than was necessary will have been expended to
obtain useful results. For example, in mining, linear elastic
models are often used to analyse rock stresses. Although in many
cases it is known that the ground does not in reality behave in a
1.

MAusIMM, Principal Mining Engineer, Australian Mining


Consultants Pty Ltd, Level 19, 114 William Street, Melbourne Vic
3000.

MassMin 2000

linear elastic way, the results are close enough not to alter the
decisions that would be made. Additional accuracy will not
change the outcome, so the added complexity of non-linear
calculations is not warranted. The crucial question to ask about a
model is not How accurately do its calculations mimic the real
interactions? but is rather If I feed in realistic input data, does
the model produce realistic outputs?
There is a problem here in identifying the boundaries and the
level of detail of the modelling project. If these are too wide or
too great, unnecessary work is done. But if they are not broad or
detailed enough, there is the risk of being lulled into a false sense
of security: work has been done and a result obtained, but a
feature not modelled may yet have an adverse impact. There is
often no easy way of identifying where the bounds should be.
For the sake of conservatism, it is this authors opinion that it is
better, and potentially less expensive in the long run, to err on the
side of doing too much than not enough (though one must be
careful to avoid paralysis by analysis).

Steps in the modelling process


All modelling projects have a number of process steps (Pegden,
Shannon and Sadowski, 1995). These apply to all types of
modelling, and are very relevant to discrete event simulations.
The following subsections briefly describe these steps, and the
reasons for them.

Identification of the problem


The model will generally be purpose built for the problem. The
problem to be solved must therefore be specified. This may be
vague at first, but become clearer as initial project scoping
proceeds.

Statement of project objectives


No model can do everything, or it will become too large and
cumbersome. The scope must therefore be defined to ensure that
all key issues are covered, and non-essential complications do
not distract from the key purposes. Together with the
identification of the problem, this is one of the most critical parts
of the project. It may require revisiting as the project progresses,
to maintain focus, and/or to formally revise the objectives.

Collection and preparation of data


Input data, both in the description of the system and for the
numbers to be used for various parameters, must be accurate.
Real data should be used in the model as soon as possible for
reality checking as the model is being built. Non-availability of
data may compromise the validity of results. The old adage
Garbage In - Garbage Out is very relevant. If data is not
available, assumptions need to be clearly stated, and sensitivity
analyses may be necessary to ascertain how critical that data
might be, to determine whether more detailed investigation is
necessary.

Formulation and construction of the model


Generally it is preferable to construct a model as a group of
structured modules. This aids subsequent debugging and
modification.

Brisbane, Qld, 29 October - 2 November 2000

83

B E HALL

It is important that the potential ultimate model requirements


are known at the beginning of the project, even if all features are
not constructed initially. That way, allowance can be made in the
overall design for later incorporation of more detailed features. If
this is not done, at best there may be a loss of time and/or
efficiency as new items are grafted into something which was not
designed to accommodate them. At worst, it may be necessary to
start again from scratch, as the methodology used may not be
amenable to modelling the new features.

Verification and validation of the model


There are two distinct tasks to be done at this step. Verification
ensures that the model works as intended, for the full range of
input conditions. Validation tests that the model results are
realistic for known real world cases where possible. It is only
after this stage that it is possible to use the model with
confidence to make predictions about cases which do not yet
exist in reality. Animation can be a very useful debugging tool at
this stage of the process with simulations.

Modification and/or refinement of the model


Feedback loops at all stages of the modelling process help to
ensure that the model is constructed as efficiently as possible.
Early results may cause a rethink of model logic, or even of the
objectives of the study.

Using the model


After gaining confidence from the verification/validation phase,
the model can then be used to make predictions. However, it may
happen that, where a model is being developed to identify
problems with existing systems, the act of fully investigating and
describing the current system to enable it to be accurately
modelled leads to a realisation of where the problem lies. When a
modeller who is not fully conversant with the system starts
asking probing questions to gain understanding, those who know
the system may investigate certain aspects more thoroughly, and
discover something previously unknown. Animations of
simulations can also be very useful for understanding the
behaviour of a system. However, with an animated simulation, as
in real life, it is difficult or impossible to guarantee that a
problematical situation will arise while the animation is being
viewed, and that, if it does, it will be recognised as such.

Applying the results


It may appear self-evident, but it needs to be recognised that, if
there is no intention to use the results of a modelling project, it is
a waste of time and money doing the job in the first place. The
project sponsor needs to be in a position to influence
implementation of the study recommendations.

ADVANTAGES AND DISADVANTAGES OF


SIMULATION
Simulation has many benefits.
It can provide realistic estimates of the behaviour of a system. It
is ideal for evaluating the effects of changes in complex dynamic
and interrelated systems, such as changes to the number or types
of machines, processing rates, machine availability and
breakdown parameters, the process flow, allocation of operators
to various tasks, product mixes or blends, material handling
methods and stockpile sizes.
Simulation can provide results when simple analytical
solutions cannot be calculated, components in the overall system
interact in a complex way, and what is lost on the swings may

84

not necessarily be gained on the roundabouts. No matter how


complex the system, if it can be described, it can be modelled to
the requisite degree of accuracy. By utilising the stochastic
random behaviour features of the software, both average
performance and likely variability can be estimated. Experiments
can also be conducted without disrupting the real system, or
before the real system exists.
Animations make visualisation of the system easier. They can
often be crucial in selling the validity of the study to influential
sceptics. As noted above, they can also be very useful at various
stages of the modelling process. On the downside, brief viewing
of an animation has the potential to unduly influence an analyst
about the relative importance of an observed feature. To obtain
sufficient results to support rigorous engineering decisions, it is
generally necessary to do a large number of runs without the
animation. Output that can be analysed dispassionately must be
generated, tabulated, plotted and compared.
Despite its power there are a number of things that simulation
cannot do. For example, it cannot automatically optimise the
system. However, it will give answers to various what if?
questions which can point a team towards the optimum answer. It
cannot give accurate results from inaccurate data nor describe
system characteristics which have not been explicitly modelled.
Nor, importantly for clients expectations, can it provide fast easy
answers to complex questions. Simulation may, however, be the
only technique which is capable of producing the desired result
at all.

PITFALLS TO AVOID WITH SIMULATION


Simulation may not be the appropriate tool
When a problem is identified, there are frequently many ways it
can be solved. At times, simulation may be used for problems
where it is not the most appropriate tool. This can result in an
end user unfamiliar with the strengths and weaknesses of
simulation becoming dissatisfied with it, and subsequently
rejecting its use for a problem where it is the ideal tool. At best,
the problem solving process may be slower than necessary if
simulation is applied inappropriately.
Two areas which this author believes are not the province of
simulation are the development and analysis of a long-term
mining plan, and detailed scheduling of mining activities over
time. The author believes that these two classes of problems are
perhaps best handled by spreadsheet and project management
software respectively. Stochastic add-ins for off-the-shelf
software permit evaluation on the impact of random events in
these types of analyses. Results of discrete event simulations of
the state of the mine at various times and with various constraints
applied may give invaluable input data with respect to production
capabilities and bottlenecks to use in the other analyses. But the
intricacies of determining optimum mine scheduling rules, for
instance, are at the cutting edge of research. They are not easily
incorporated into off-the-shelf simulation packages and models
developed using them.

Sufficient analysis may not be done


For valid conclusions to be drawn, a statistically valid sample
must be obtained. It is beyond the scope of this paper to discuss
this in any detail. However, to ensure that low probability events
or combinations of events are able to occur, and that their impact
on overall predictions is neither too great nor too small, the
simulation modeller must give careful consideration to both the
length of a simulation run and the number of replications done.

Brisbane, Qld, 29 October - 2 November 2000

MassMin 2000

SIMULATION MODELLING OF MINING SYSTEMS

It must be recognised that, because of the stochastic nature of


events both in real life and in a simulation model, results of
observations may not plot on a straight line or smooth curve such
as might be predicted. With a relatively simple simulation, it may
be that all random variations are averaged out, and ideal results
are obtained. But with a more complex system, it may be
impractical to achieve this. The more complex the model, the
more dangerous it may be to draw conclusions from a short run,
or from analysis of a single run. Longer runs, and for a number
of scenarios, may be required to interpolate trends. A regression
line may give a better estimate of the value of a dependent
variable than the actual observed value with a given value of an
independent variable. Similarly, with simulations, the trends,
rather than the results actually obtained for that case, may give
the best estimate of the most likely results for a given set of
circumstances.

Inadequate project control may compromise the study


A small case study is used to illustrate this potential pitfall.
Modifications to the treatment plant at a certain mine were
reducing significantly the amount of ore storage available. At the
same time, the mine was considering increasing its output. The
questions posed by senior management were apparently simple:
Can the mine produce at the proposed target rate, and is the
reduction in surge capacity a constraint?
A study team was established, comprising an experienced
modeller and a relatively junior mining engineer. The modeller
was skilled in the use of the simulation software package, but
was not familiar with the mine. The mining engineer was too
junior to seriously challenge what the modeller was building into
the model. He was also unaware of managements interest in
changing working rosters for both operating and maintenance
staff in both the mine and the treatment plant. Management had
not ensured that the need to evaluate such changes, should they
be necessary for the achievement of the goals, had been
communicated to the study team.
After several weeks of gathering data, building a model, and
generating results - all without further management involvement
- a draft report was prepared, concluding that the target
production rate was achieveable, and the surface surge capacity
was not a constraint. However, some comments in the report rang
alarm bells, and the author was called in to audit the study.
Investigations revealed that:

The study did not make use of random effects. All


calculations were fully deterministic.

No allowance was made for the fact that the treatment plant
worked 24 hours a day, seven days a week, while the mine
had a weekend break in excess of 24 hours.

Though the report was being published with all the authority
of: A simulation study has determined that , the real
calculations in the model affecting the results could be fully
and accurately summarised as follows:
1.

Shaft availability
x hours per month
x hoisting rate
= tonnes hoisted per month,
> target production per month
target is achieveable

2.

(Hoisting rate - treatment rate)


x hours per month
= increase in stock in one month
< surge bin capacity
surge bin is not a constraint.

MassMin 2000

A complex model, built using a powerful tool, had done


nothing more than back of envelope calculations. The first
calculation was trivial. Management knew that the shaft itself
was not the constraint, but was aware that patterns of ore
production in the mine (affected by availability of ore and
equipment availabilities) and treatment in the mill (affected by
breakdowns and maintenance rosters) were the crucial issues, yet
these were not investigated at all in the study.
The second calculation was also trivial, but was compounded
by a serious logical flaw. The net rate of increase in stocks was
positive, and there would eventually come a time (after some two
months with the data used) when the surge bin would be filled,
and the hoisting rate would have to drop back to the treatment
rate. The analysis had an implicit assumption that somehow the
surge bin emptied itself at the end of each month.
Clearly there are a number of problems highlighted by this
example. If they had not become apparent in time, serious wrong
decisions could have been made. The study was subsequently
re-done, requiring a complete re-build of the model to enable it
to answer managements real questions. Certain workplace
changes were identified as necessary, and sold to the workforce
in part by using the simulation results.
The key lessons for a successful simulation study are:

the

sponsoring managers must


communication with the study team;

remain

in

regular

management must ensure that the problem to be solved and


the objectives of the study are clearly understood by all
concerned;

management must ensure that the model is capable of


answering all the questions they may wish to ask of it;

the team working on the study must consist of at least:


- a person familiar with the software, and able to
-

appreciate how real world situations may be translated


into model constructs; and
a person familiar with the system being modelled, to
ensure that all relevant real world behaviours are
accounted for by the model.

These two may at times be one and the same person. If not
they must both have such personal qualities and comparable
levels of seniority as are needed to ensure that each is able to
examine, understand and challenge the work of the other to
ensure the final result is reliable.

TYPICAL SIMULATION RESULTS


There are many types of systems which can be investigated, as
indicated in many papers on simulation over the years (Sturgul,
1997). In many of these, a substantial proportion of the paper is
devoted to reasons for the selection of the software, and/or a
detailed description of the mining system being studied. The
types of results obtained and presented often receive little
attention.
In this paper, the authors aim is to indicate to potential users
of simulation the types of things which might be investigated,
and a few ways in which the results might be presented. This is
done by way of two case studies. In each case, a description of
the system investigated is given in broad terms only, to provide
basic understanding. The types of variables included in the
model, and thus able to be changed for evaluation of their impact
on the system, are described. Some typical results from each
study are presented, showing a number of presentation formats
that this author finds useful.
It should be noted that, if something can be measured in real
life, then it can be measured and recorded in a simulation model.
The only limitations are disk space on the computers, the
imagination of the modelling team, and the time available to run
the models and draw useful conclusions.

Brisbane, Qld, 29 October - 2 November 2000

85

B E HALL

CASE STUDY 1 TRUCK HAULAGE IN A


DECLINE
Description of system
A study was undertaken on behalf of the Kanowna Belle Gold
Mine in Western Australia, 100 per cent owned by Delta Gold
Limited. The main objective of the project was to identify the
truck fleet requirements for achieving production targets over a
number of years as the average depth of production increased.
The study investigated the truck haulage system in some detail,
with simulation modelling taking into account such factors as
planned maintenance and breakdowns, shift working rosters
(times and durations of shift changes, meal breaks and vehicle
servicing), queuing and congestion in the decline passing bays
and at loaders, and the total duty requirements of the truck fleet
at various stages of the mine life. The impact of other vehicles in
the decline was also accounted for. The model was not developed
to allow significant decisions to be made about loaders and
loading operations, but simple modelling of loaders was used to
define the number and locations of truck loading points, taking
into account planned and breakdown maintenance of loaders and
their operational availability after allowing for meal breaks and
the like.
The main features of the trucking system were as follows. The
underground haulage occurred in a fairly typical decline,
commencing near the bottom of an existing open pit. Passing and
overtaking were possible only at defined passing bays. Mining
levels branched off the decline at various defined locations. For
each of the four major mining blocks defined at the mine,
representative levels were defined for production loading,
development loading and fill tipping. All material hauled to
surface was tipped either directly into the surface crusher, onto
the main run-of-mine stockpile near the crusher, onto a small ore
stockpile in the open pit, near the portal, or onto the waste dump.
Provision was also made for underground tipping at a possible
shaft location.

Simulations evaluated truck requirements to meet scheduled


tonnages of production and development ore and waste, and two
types of fill, for three specified years of the mines schedule.
Productivities were also assessed for the three major activities of
production, development, and backfill separately, for each
mining block. Two truck types, tippers and ejectors, were
modelled. All available ejector trucks were allocated to filling if
there was fill to be hauled. If not, ejector trucks were able to be
allocated to production or development tasks, as were all tipper
trucks. A simple truck despatching algorithm was implemented
to ensure that loaders and trucks were allocated to production
and development rock sources so that the simulated production
remained in line with the schedules. To maximise truck fleet
utilisation, queuing at one loader while another was idle was also
avoided.

Typical investigations and results


A number of production and equipment parameters were
measured in the model and output in a simple spreadsheet file,
which was subsequently processed to generate a variety of results
graphs. For each case run, the following types of plots were
typically generated, along with the detailed data tables which
they summarised.

Daily ore production from each stoping and development


source (Figure 1).

Cumulative frequency distribution of daily ore tonnages


(Figure 2), which summarises Figure 1.

Daily split of operating, maintenance and idle activities for


each type of truck and loader (Figure 3). These could also be
summarised into a pie chart if desired.

Histograms of the numbers of trucks in the decline or on


levels at any time (Figure 4).
Figures 1 - 3 show the quite significant fluctuations in daily
performance which can be expected in this particular system.
These result purely from the system characteristics, rather than
good or bad management practices. Figure 4 was useful for
identifying ventilation requirements.

Fig 1 - Daily ore tonnes trucked - year 3 targets.

86

Brisbane, Qld, 29 October - 2 November 2000

MassMin 2000

SIMULATION MODELLING OF MINING SYSTEMS

FIG 2 - Cumulative frequency distribution of daily ore tonnes trucked.

FIG 3 - Daily fleet availability and utilisation - year 3 targets.

MassMin 2000

Brisbane, Qld, 29 October - 2 November 2000

87

B E HALL

FIG 4 - Proportion of the time various numbers of trucks are in the decline.

For each mining block, cases were run with varying numbers
of trucks, for production, development, and filling separately. For
each product hauled, a number of plots were generated. For each
mining block, typical plots were:

Productivity (tonnes per hour and hours per tonne) as a


function of mining depth, at target production rates (Figure
8). The truck operating hours per tonne curves are essentially
linear with depth, indicating that the interference effects with
larger fleets are also linear with depth.

Tonnes hauled vs number of trucks in fleet, showing the daily


averages and a range of percentiles above and below the
average (Figure 5). The averages were generally relatively
linear until the output approached the amount of rock
actually generated by the model, when a levelling off
occurred. (It was specified for the study that ore should
always be available for trucking. The model therefore
generated rock to be trucked at some 200 per cent of the
target rates.)

Truck fleet availability and utilisation components vs number


of trucks (Figure 6). The non-working times increase slightly
with increasing fleet size while output is increasing, but
when the maximum rock output is approached, these idle
times increase more rapidly.
Figure 5, like Figure 1, indicates the level of normal daily
production variability to be expected. It can also indicate the size
of fleet required to be able to produce at least the daily target for
a specified proportion of the time. The results on which Figure 6
is based can indicate realistic availability and utilisation targets
for maintenance and operating sections, taking account of both
the size of the fleet and the duty required of it.
Comparisons for all mining blocks were generated, such as:

Production and productivity curves for each block, for


various fleet sizes (Figure 7). Production curves for Blocks A
and B, relatively close to surface, show significant flattening
of the curve as the rate of ore generation used is approached.
Deeper Blocks C and D exhibit a more gentle curvilinear
behaviour as the increasing truck fleet results in some
increasing interference.

88

Figure 9 shows the effect of changing passing bay spacing in a


truck decline. The Kanowna Belle model had the facility to
evaluate this behaviour, but this was not done for that study. This
figure was generated for another similar study. It can be seen
that, as the spacing of passing pays decreases, the truck
productivity approaches that of a one-way loop haulage, at
significantly less cost.
For simulations of the total materials movements, two distinct
modes of filling operation were identified. When cemented fill
was to be run, the operators would target nominally full-time
operation of the cement slurry plant and fill loading system with
no work breaks. At other times, dry rock fill would be run at its
average required daily rate with normal work rosters. This
indicated two possible operating strategies, which were both
assessed:
1.

operate a truck fleet which is large enough to produce the


average daily ore target, even when cemented fill is being
hauled at the maximum rate, aiming to do scheduled
maintenance when cemented fill is not required and the
total trucking demand is lower; and

2.

operate a truck fleet so that the overall daily average target


is met by producing more than the target when cemented
fill is not required, and less when cemented fill is required.

Study conclusions
The main conclusions of the study were as follows:

Brisbane, Qld, 29 October - 2 November 2000

MassMin 2000

SIMULATION MODELLING OF MINING SYSTEMS

FIG 5 - Daily ore tonnes trucked - Block A - production ore only.

FIG 6 - Truck fleet availability and utilisation - Block A - production ore only.

MassMin 2000

Brisbane, Qld, 29 October - 2 November 2000

89

B E HALL

FIG 7 - Daily ore tonnes trucked - production ore only.

FIG 8 - Truck productivity vs haulage depth - production ore only.

90

Brisbane, Qld, 29 October - 2 November 2000

MassMin 2000

SIMULATION MODELLING OF MINING SYSTEMS

FIG 9 - Trucking capacity as a function of fleet size and passing bay spacing.

1.

If variations in ore production were acceptable depending


on the filling duty at any time, a fleet of a certain number of
trucks was required.

and downstream conveying of the DOZ ore. The model was built
as a standalone model, but could also be incorporated into an
overall site ore flow model developed independently for PTFI.

2.

If it was required that daily production should average the


target even while filling with cemented fill, up to two
additional trucks would be required in later years.

Description of system

3.

Significant daily fluctuations in output were likely in any


case.

An extraction level with:


- up to 30 extraction

4.

Over-trucking and consequent significant queuing idle


time was necessary to meet cemented fill targets, which
effectively required the fill loading stations to work
non-stop.

5.

There was a reducing contribution made by additional


trucks as the rock generation limit was approached.
Therefore, if the development, drilling and blasting
capability were not significantly in excess of the daily
targets, additional trucks would be required to provide
additional instantaneous trucking capacity, to ensure that
average targets were met despite there being no ore to haul
at times. (This effect was not quantified for this particular
study.)

CASE STUDY 2 BLOCK CAVING PRODUCTION


OPERATIONS
PT Freeport Indonesia (PTFI) commissioned a simulation of the
underground block caving production system for the planned
DOZ operation in the Ertsberg Mine in Irian Jaya, Indonesia.
The simulation modelled the mucking and secondary blasting
activities on the block caving extraction level, trucking
operations on the truck haulage level, and simplified crushing

MassMin 2000

The mining system modelled consisted of:

panel cross-cuts, aligned


approximately north east - south west;
an orepass dump near the centre of each panel,
effectively breaking each panel into northern and
southern sections, and thereby allowing two loaders to
operate in each panel simultaneously; and
three additional dumps off the northern perimeter drive,
servicing the northern sections of a number of the central
panels.

A truck haulage level with:


- a one-way loop haulage, outbound empty in the north,
-

and inbound full in the south;


truck loading chutes in cross-cuts connecting the inbound
and outbound haulage drives;
an additional haulage loop to the north to access the three
northern ore passes; and
a central dumping area, with east and west accesses from
the haulage, and three truck dumping locations into the
dump pocket above the crusher.

Crushing and conveying facilities consisting of dump pocket,


crusher, surge bin, and a generic conveyer drawing from the
surge bin, representing the entire ore handling system
downstream of the surge bin.

Brisbane, Qld, 29 October - 2 November 2000

91

B E HALL

The main features that were taken into account in the model
were:

prioritisation of ore passes, taking account of ore pass

On the extraction level:

truck loading chute hang-ups and failures.

a number of different loader and hang-up drill types;


loader movements between drawpoints and dumps, between
cross-cuts, and to and from the workshop;

delays due to planned work breaks, breakdowns and


scheduled maintenance, and full and blocked dumps;

prioritisation of drawpoints and cross-cuts, taking account of


drawpoint daily draw limits, maximum time between
successive drawings, and maintenance of even draw;

three types of drawpoint hang-ups;


temporary sterilisation of drawpoints caused by hang-ups in
adjacent drawpoints;

drilling and blasting of hang-ups;


different hang-up, sterilisation and blasting parameters for
different muck types;

blasting outside scheduled times as an option; and


drawpoint failures and repairs.
On the truck haulage level:
a number of different truck types;
truck movements between chutes and dumps, including the
effect of broken down trucks in the haulage;

delays due to planned work breaks, breakdowns and


scheduled
maintenance,
full/blocked
dumps,
empty/inaccessible ore passes, and queuing at chutes and
dumps;

contents, and numbers of trucks already allocated to passes;


and

Crushing and conveying system:


planned maintenance shutdowns and unplanned breakdowns;
existence of surge capacity at certain locations, and
availability of ore to draw or space to dump in surge bins;
and

maximum processing rates for each item of plant.


Typical investigations and results
The model was delivered to PTFI, who subsequently performed
all their own analyses. To facilitate this, all inputs to the model
were via text files. These were generated by Microsoft Excel
spreadsheets for ease of use by PTFI personnel. The model also
wrote all its output to a spreadsheet file. A template Excel
spreadsheet was provided for processing the results from each
model run into a useable format.
Sample results have been generated by the author,
independently of PTFIs studies. Figures 10 and 11 show two
different views of production with varying numbers of loaders
and trucks. The figures show clearly the regions where the
operating constraint is either the number of loaders, the number
of trucks, or neither of these, but rather the existing state of the
mine.
The flat but irregular horizontal region of the surface in Figure
10 is a good example of the situation, described above, where
even long simulation runs may not result in a smoothing of all
the random effects. This is the region where the state of the
mine rather than the size of the equipment fleets, is the

Fig 10 - Daily production as a function of loader and truck fleet.

92

Brisbane, Qld, 29 October - 2 November 2000

MassMin 2000

SIMULATION MODELLING OF MINING SYSTEMS

FIG 11 - Contours of daily production as a function of loader and truck fleet.

constraining factor. Investigation of these irregularities indicated


that, although the long run average number of unserviceable
drawpoints was consistent in all cases, the pattern of location of
these over time affected the production rate to varying degrees.
Runs of sufficient length to account for all possible variations in
drawpoint failure patterns would be impractically long. It is
therefore necessary to estimate the expected maximum possible
production from the mine in a particular state as the average of
those cases which are judged to be limited by the state of the
mine.
Figures 10 and 11 were generated with a secondary blasting
drill fleet too large to be a constraint. Having identified a limited
set of suitable truck and loader fleet sizes, similar plots could be
generated for various numbers of drills. This would then assist in
deriving the balance of trucks, loaders and drills best able to
meet production targets. Also, using fleet sizes and production
rates as shown in Figure 10, together with productivities and
utilisation statistics from the model (similar to those in Figure 6)
and applying various operating and capital costs and revenue
assumptions, the optimum economic operating point could be
determined.
The PTFI data specified different characteristics for different
ore types. In particular, various categories of wet muck would
hang up in drawpoints less frequently than similar dry ore, but
would require longer periods of delays to allow stabilisation of
the hang-up before secondary blasting. Figure 12 shows the
decline in maximum production rate as the proportion of
drawpoints with wet muck increases. The values shown for each
data point are the estimated average for the irregular mine-state
limited cases in each scenario, as described above, with the
ranges of the irregularities indicated by error bars. It can be
seen that there is initially a significant effect as the number of
wet drawpoints increases, but this effect reduces as the
proportion of wet drawpoints increases further.

MassMin 2000

A number of other parameters were reported by the model.


Loader and truck activities were reported as 8 and 13 categories
respectively. Similarly, 20 different drawpoint states were
identified and reported, and summarised into five main
categories. Records were also kept and plotted for truck chute
status and drawpoint and roadway repairs. These were all
presented in plots similar to Figure 3 for showing the various
effects over time, and as pie charts to summarise the results for
each run. Stacked bar charts, similar to Figure 6, were also used
to show trends across a number of runs. As noted above, the
information able to be recorded and reported is virtually
unlimited.
Sample results were also generated to test the unscheduled
blasting feature of the model. For the situation assumed, it could
be seen (Figure 13) that the number of unscheduled blasts is
small until there are excess loaders available. However despite
this increase in the amount of unscheduled blasting, the overall
benefits are negligible. This would suggest that factors such as
maintaining an even draw eventually limit the overall production
rate, despite the apparent ability to blast more frequently and
muck higher tonnages in the short-term.
The model constructed permitted the effect of a large number
of parameters to be assessed. The author is aware that PTFI has
continued to use the model to investigate a large number of
possible mining scenarios.

CONCLUSION
Simulation is a powerful tool for the mining engineer. When used
in appropriate applications it is able to provide insights into
system behaviour in a way that few, if any, other techniques can.
However, if improperly used, serious mistakes can arise. This is
perhaps exacerbated by the fact that results are not always
intuitive, and can rarely be checked by random audits of reported
results only, or by a person not familiar with the intricacies of the
particular software package used.

Brisbane, Qld, 29 October - 2 November 2000

93

B E HALL

FIG 12 - Block caving production potential as affected by drawpoints with wet muck.

FIG 13 - Impact of unscheduled blasting on production.

94

Brisbane, Qld, 29 October - 2 November 2000

MassMin 2000

SIMULATION MODELLING OF MINING SYSTEMS

With a complex system, a large number of runs may be


necessary to adequately investigate the response of the system to
variations in a number of parameters, both individually and
simultaneously.
Necessary skills in the simulation team are the abilities to
assess what cases actually need to be evaluated, and to present
the results in ways which can be readily understood. Spreadsheet
(and other) plotting and graphing software can assist with this.
The information reported by the model as it runs must meet the
needs of the project sponsors, but is generally only limited by the
imaginations of the modelling team.

The managements of Delta Gold Limited and PT Freeport


Indonesia are thanked for their permission to describe the models
and results pertaining to their respective operations.

REFERENCES
Pegden, C D, Shannon, R E and Sadowski, R P, 1995. Introduction to
simulation using SIMAN, 2nd edn, pp 8-24 (McGraw Hill: New
York).
Sturgul, J R, 1997. Annotated bibliography of mine system simulation, in
Mine Simulation - Proceedings of the First International Symposium
on Mine Simulation. (Balkema : Rotterdam).

ACKNOWLEDGEMENTS
The author wishes to thank the management of Australian
Mining Consultants Pty Ltd for permission to prepare and
present this paper, and also the companys secretarial staff for
assistance in its preparation.

MassMin 2000

Brisbane, Qld, 29 October - 2 November 2000

95

You might also like