An Application of Linear Programming for

Block Cave Draw Control

A R Guest1, G J Van Hout1, A Von Johannides1 and L F Scheepers2

ABSTRACT condition of the various areas, and to ensure that the draw control
procedures were been adhered to.
In the past various systems have been used to plan the extraction of ore
from block cave operations. These systems have varied from rather In the slusher drift type cave, ‘hang-up control’ was a major
sophisticated models with empirical ‘gravity flow’ rules applied to the problem. Not only was it a frequent problem, which is beneficial
very rudimentary models where only the recorded mined tonnage’s have from an operational point of view, but to bring down the hang-up
been accounted for against the available resource. The normal was also quite labour intensive. This resulted in the winch
requirement of these models has been to only plan the movement of
ground through the drawpoints. Little cognisance has been taken of the
operator simply ignoring the ‘problem’, in order to achieve the
other parameters that affect the total mining operations. The intended ‘Call’ that he was set, by readily scrapping ground from the
purpose of the linear program planning system is to not only include the closer drawpoints. In a typical slusher Cave at Kimberley Mines,
management, resource availability and geotechnical constraints, but to a Cave could have up to 17 drifts cutting across the kimberlite
also include the mining rules. The paper will discuss the various pipe, with each drift having up to 45 drawpoints. This resulted in
constraints that have been considered, and attempt to illustrate where the excess of 500 drawpoints being available (Figure 1).
system better serves the production system than previous ones. A brief
technical discussion on the linear programming system is included. With the introduction of the Draw Control system, other
methods then also had to be introduced to overcome these
‘operational shortcomings’. This was achieved by the
INTRODUCTION introduction of what became known as ‘strip mining’. This had
Any attempt to operate a Block Cave operation without an the affect of splitting the Cave essentially into three separate
effective Draw Control system would be irresponsible. Through working areas within the cave. A series of ten ‘open’ drawpoints
the proper use of an effective draw system, the fragmentation were separated by a series of eight to ten ‘closed’ drawpoints.
within a cave can be controlled, stress build up in areas can be This was repeated across the width of the pipe. During any given
alleviated, and even free water within the cave can to a certain month, a call would then be allocated to the open drawpoints.
degree be directed. However, miners being miners, it was soon realised that the
When De Beers introduced its first experimental cave in 1954 closed drawpoints, if pulled early in their ‘closed’ period, would
at Bultfontein Mine, Kimberley, a number of practical lessons yield finer ground. A method to overcome this ‘stealing’ was the
were quickly learnt (Gallagher and Loftus, 1960; 1961). In two introduction of a car tyre into the mouth of the drawpoint, held in
instances multi-tunnel collapses occurred, resulting from stress place with steel chains, as well as a lead meter seal, placed by the
build-up within the cave as a result of incorrect draw. During draw control officer.
subsequent years, a number of minor instances have also Another system that was also introduced to control the
occurred where, through draw control oversight, drifts have come ‘hang-up’ and to ensure that work was been carried out to bring
under high point loads, and have collapsed. This situation the hang-up down involved the ‘paint bombs’. This involved the
however is then only recovered by an intense adherence to a making up of poster paint bombs on the surface. No electric light
strictly controlled loading plan in adjacent drawpoints. bulb was ever discarded at Kimberley Mine, as the Draw Control
The first draw control system which was introduced in the officer required them all for that purpose. During his routine
Kimberley Mines consisted solely of a manual recording of the inspections to the cave, the throwing of one of these bombs duly
daily draw records coming up from the production drifts. In the marked any drawpoint that was ‘hung-up’. The only way that the
early days these records only recorded the ‘scraper’ tally from miner in charge of that section could get rid of the markings was
the drifts. It was soon realised that to control the loading within a to use explosives!
cave better, a more detailed record of where the ore was coming Having now reduced the Cave into two or three smaller
from was needed. This resulted in the first ‘winch indicator sections, it became easier to ensure that the drawpoints were
boards’ being devised, and installed on the scraper winches. being worked correctly. The ‘closed’ drawpoints were carefully
With this device, the winch operator was able to ‘calibrate’ the monitored during the time they were out of production, to ensure
indicator at the start of each shift, thus indicating from which that no excess point loading was taking place.
drawpoint within a drift the ground was being pulled. This added The strip control had a number of benefits. Firstly it allowed
to the knowledge of the draw records, and these records could the Cave to be divided into smaller sections. This allowed a
then be collated with the tonnage that was then taken by the train meaningful call to be placed on that area. Generally, a call for
to the ground passes. 80 per cent of the available drawpoints would be placed on an
Once this data was available, and being recorded and area. This meant that to achieve the allocated call almost all
graphically presented manually, it was then decided to use this drawpoints in the section had to be worked, allowing for some
data to proactively control the caving process. This resulted in drawpoints to be out of action for repairs, hang-up or just
the appointment of the first Draw Control Officer. consolidating prior to any action work. The other major benefit
Amongst his duties of collecting and presenting the data, was of the strip control was fragmentation. Whilst the ‘closed’
to make daily visits to the operating caves, to observe the drawpoints were out of action, loads were being transmitted to
this dead area. This resulted in secondary fragmentation of the
caved material during this period. As the different Caves in the
1. De Beers Corporate Headquarters, Johannesburg, South Africa. Kimberley area each had their own differing mixtures of soft and
2. Large Scale Linear Programming Solutions, Johannesburg, South hard kimberlites, the period that a particular drawpoint could stay
Africa. closed varied from cave to cave. Generally a single drawpoint
could remain closed for up to six months, but some caves had to

MassMin 2000 Brisbane, Qld, 29 October - 2 November 2000 461

A R GUEST, G J VAN HOUT, A VON JOHANNIDES and L F SCHEEPERS

FIG 1 - Plan of Wesselton Mine Block Cave Extraction Level.

move the strip faster. This strip movement was accomplished by A specific feature of the system was the introduction of a
opening one drawpoint on one side of the strip, and closing one ‘Control Horizon’. This in affect restricted the available tonnage
on the opposite side. (In caves that required faster strip to some defined percentage of the total reserves allocated to the
movement up to two or three could be opened or closed at a drawpoints. The benefit of this was an attempt to control the cave
time.) by not allowing the draw of any drawpoint to get excessively out
The result of the work being carried out in the closed strip was of control. No call would be allocated to any drawpoint that had
exceeded the Control Horizon.
evident once a drawpoint was brought back into production.
Bearing in mind that this new drawpoint would have now been The system calculated the calls for each drawpoint on an
out of action for up to six months. For the first three weeks this inverse weighting allocation, based on the number of months left
drawpoint would require a high degree of secondary blasting, to to deplete the reserves, at the previously calculated average
tonnage per month.
get the ground to move. However, for the next few months, the
ground would be well-fragmented and easy to work. Within the system data on Area Call, maximum production per
drift, average production per drawpoint, maximum production
per drawpoint, working drawpoints required were all required
EARLY COMPUTERISED DRAW CONTROL prior to setting up calls. The system would then ensure that none
SYSTEMS of these criteria were violated. The system would iterate the
allocated calls until a solution was achieved, or an error message
Once the main operational ‘features’ had been sorted out, and would be generated suggesting some remedial action. This type
computers started to make their appearance into industry, the first of message would become common on nearing the Control
computerised Draw Control System was written and introduced Horizon, and would indicate a lifting of the Control Horizon.
into Kimberley Mines, in 1967. The system basically followed No attempt was made to introduce sophisticated gravity flow
what the manual system had been doing, but obviously allowed rules into the draw, as the maximum tonnage extracted per
for a more efficient data handling, and graphical output. drawpoint per day rarely exceeded 30.

462 Brisbane, Qld, 29 October - 2 November 2000 MassMin 2000

 AN APPLICATION OF LINEAR PROGRAMMING FOR BLOCK CAVE DRAW CONTROL

With the advent of the personal computer, graphical System Flow Description
presentations became a lot easier to generate. Also user
interaction with the computer program was possible, and the very
first version of Gemcom’s PCBC was written and introduced
onto Premier Mine (Owen and Guest, 1994). This for the first Actual production and other
time introduced Laubscher’s empirical gravity flow rules into the VMS relevant node data as measured
draw control system (Laubscher, 1994). Tracking data from by a Vehicle Monitoring System.
One of the major shortcomings of this system, was the lack of drawpoint to plant This data can be obtained per
production based input into the setting of calls. This system was shift or by accumulated time
run from a Draw Control base only. It was only in its latter years, periods.
when Production/Draw Control meetings were introduced on the
Determine resource Query of VMS data to
mines, that some degree of ‘production constraints’ were
movements summarise all the movements
introduced. These were then accounted for in the system during from a source (eg drawpoint) to
the allocation of the drawpoint weightings. destination (eg ore pass) per shift
With the introduction of a draw control system based on linear or accumulated time period.
programming, it is anticipated that by including and integrating
constraints from other disciplines like geology, mining and Drawbell interaction Apply the drawbell interaction
metallurgy into the system it will become more accepted as a and interface mixing and interfacing mixing algorithm
business planning tool and not perceived solely as a policeman. to simulate the caving process.
This has been done by not only accommodating for drift and Sequence is important in this
process.
drawpoint constraints, but also Load-Haul-Dump equipment, ore
pass, transfer level, train, as well as crusher availability into the After the drawbell interaction
allocation of calls. Mineral Resource
Database (MINRAS) and interface mixing the resource
must be updated, so that Linear
BRIEF DESCRIPTION OF A COMPLETE DRAW Program Draw Control System
uses the latest updated resource.
CONTROL SYSTEM
A successful and complete Draw Control System in a cave LP
mining operation not only consists of a long term scheduling Linear Program based long term
long term scheduler scheduler to run once the
tool, the linear programming. Other components to be integrated on the updated resource is updated and the
into the system are an accurate data collection system, an resource model constraint set has been modified
appropriate cave simulator, a correct mineral resource database and checked.
and a short term scheduling system.
Figure 2 shows a simplified system flow through these short term scheduler
different components making up the Draw Control System as new call based on Short-term scheduler to guide
envisaged by the authors. planned (LP) tons production in achieving long
It is not within the scope of this paper to describe in detail all and actual (VMS) term plan as produced by long
tons mined term scheduler
components within the draw control system. This will be covered
in a future paper.

INTRODUCTION INTO LINEAR PROGRAMMING Short-term production call

Production within
according to a Linear Program
BASED SYSTEMS total ore flow
based long term scheduler plan.
The search for the best, the maximum, the minimum, or, in
general, the optimum solution to a variety of problems has
entertained and intrigued man throughout the ages. Obviously, by
doing so companies are moving into a more efficient production FIG 2 - A complete Draw Control System flowchart.
scenario, based on sound mathematical principals, and for which
benefits have been clearly stated. Since the problem of linear
programming was first developed and applied in 1947 by George Programming models of real world large mining and
B Danzig, Marshall Wood and others in the US and UK military, metallurgical complexes can have matrix dimensions of up to
applications have been extended in virtually all areas of the hundreds of thousands of variables in as many constraints and in
economy, including many areas in the mining industry. order to solve these models, a mature matrix generator, OMNI
The optimisation of a linear function subject to linear and optimiser, CPLEX are required. By using mature software,
constraints is simple in mathematical structure but powerful in its management can be sure that the planning integrity is beyond
adaptability to a wide range of applications. The mathematical doubt due to superior data to matrix conversion methodologies
technique of Linear Programming is a search routine where the and optimisation techniques employed.
utility to be optimised is expressed as a linear equation and the Table 1 is an indication of the speed and size of problems that
domain in which the search takes place, is confined by boundary successfully have been solved with Linear and Mixed Integer
or constraint linear equations. Programming.
In the case of non-linear utilities or constraints, an extension of These performance figures are the output from a Silicon
Linear Programming, Mixed Integer Programming is fruitfully Graphics Origin 200, with 256 MB of RAM, 4 × R10 000 Mips
used for formulation purposes. The same applies to ‘GO – NO – processors.
GO’ modelling. Currently a Client-Server systems approach is adopted in an
Mixed Integer Programming is extensively used in the Draw attempt to minimise the cost of implementing expensive matrix
Control System. Generally, Mixed Integer Programming models generation and optimisation software at the sites. This however is
require more computing resources than Linear Programming one approach, and each site could actually have individual
models. Modern day Linear Programming and Mixed Integer set-ups.

MassMin 2000 Brisbane, Qld, 29 October - 2 November 2000 463

A R GUEST, G J VAN HOUT, A VON JOHANNIDES and L F SCHEEPERS

TABLE 1 • for many models like grade, size, density, revenue, costs and
contribution there can be large variability from block to
Examples of Linear and Mixed Integer Program matrices and block; and
solution times.
• ore blending constraints are required to maximise
Operation No of Variables Integers Solving plant-operating efficiencies.
constraints time
A typical Life of Mine Linear Programming model for strip
Alluvial LOM 34 284 35 785 5417 6.5 hours mining operations would include the following:
Front Cave 4649 4683 1176 110 sec
• An Objective function which maximises the total
Open Pit 11 141 9002 3030 1.5 hours contribution at various levels of costing like direct mining
Marine mining 16 754 13 362 4416 30 sec and treatment, direct mining and treatment and overheads,
etc.
• The Planning is broken up into multi-time periods, which are
user definable. It must be emphasised that the optimisation is
Figure 3 illustrates a typical configuration for a Draw not done individually for each time period but for all time
Control/Production Planning System of an operating periods simultaneously. In this way the optimisation is seen
environment at De Beers sites. over the Life of Mine, as well as adhering to the constraints
and calls of each individual time period.
• There is provision for multi-production areas (mines and
treatment), each carrying its own set of constraints and calls.
• Optimisation can be done on the total resource or parts
thereof
• A set of constraints, consisting of an upper and lower limit
for each parameter, per period, per production area. It can be
summarised as:
• Mining
• Percentage cubic metres per block to allow for specific
amounts or the total block itself
• Allow for a certain percentage of stripped reserves in an
area or subarea so that flexibility of mining ore due to
short-term production problems can be achieved.
• Allow for a minimum amount of ore to be left behind in
a block to avoid the problem of having a multitude of
remnants.
• Allow for certain amount of ore and overburden to cater
for various vehicle requirements and capacities.
• Allowance for certain ton-kilometres to avoid taking only
material in close proximity to treatment plants.
• Waste tonnage range to cater for the stripping fleet.
FIG 3 - A typical operating environment for Draw Control or Production • Ore tonnage range to cater for a particular mining fleet.
Planning System at De Beers sites. • Machine hours on all major production equipment.
• Allow for blocks to be mined in sequence to cater for
dragline deployment and logistics.
PRODUCTION PLANNING SYSTEMS BASED ON • Fix certain blocks to be mined to overcome practical
LINEAR PROGRAMMING problems like laying pipelines.
Before elaborating on the Draw Control System, it may be • Allow for mining precedence per block to cater for
worthwhile looking at other Production Planning Systems based different waste and ore horizons.
on Linear and Mixed Integer Programming. • Plant
Linear Programming based principles have successfully been • Allowance for certain volume (tons or cubic metres) per
used for the Life Of Mine and shorter term scheduling in open pit processing plant.
and strip mining environments, each with their own specific
• Blending of different rock types agreeable to plant
needs (Hoerger et al, 1999; Graham-Taylor, 1992; Scheepers and efficiency.
Wellbeloved, 1992).
• Allowable percentages of clay, wet, dry, soft, hard or
Life of Mine and short-term planning systems based on cemented material.
Linear/Mixed Integer Programming have been developed since
1974 for strip mining operations, for the following reasons: • Economic
• world-class deposits had to be mined responsibly under legal • Escalation or de-escalation of various levels of costing
constraints; per block.
• Revenue escalation or de-escalation per block
• the need to split large areas into manageable portions where
each subarea constitutes a mine with its own treatment • Contribution limits by looking at the difference between
facility; revenue and cost.
• Variable discount factor per period
• subareas are split into many mining blocks, which carry
relevant parameters for planning efficiencies, economic and • Required Profit to Revenue ratio per area or sub-area.
sustainable depletion strategies; • Exchange rate per period.

464 Brisbane, Qld, 29 October - 2 November 2000 MassMin 2000

 AN APPLICATION OF LINEAR PROGRAMMING FOR BLOCK CAVE DRAW CONTROL

• General 2. Metallurgical constraints that will include plant capacities

• Allowance for a grade (carats per cubic metre or carats and rock type blends.
per hundred ton) range. 3. Economic constraints that will include cost and price
• Allowance for a particular size range. escalation or de-escalation, discounting, exchange rates and
• Allowance for a number of carats. Profit to Revenue ratio ratios.
• Allowance for a number of stones. 4. Geological constraints like grade, carats, stones, size and
• Allowance for a density range. density.
• User friendly front and back-end database systems are By using the combined constraint set it is possible to produce a
written around the Linear/Mixed Integer Programming better or optimised production mix resulting in better utilisation
component, allowing the mine planner flexibility and power and efficiency of production resources, even though only small
to mix and match multiple planning resources to multiple areas of flexibility exist because of the strict geotechnical rules
planning scenarios. and constraints.
A breakdown of the important aspects of the Linear and Mixed
THE LINEAR PROGRAMMING COMPONENT OF Integer Programming method and formulation for Draw Control
Systems is discussed below.
THE DRAW CONTROL SYSTEMS
As illustrated in the draw control system flowchart (Figure 2), Objective function
the long-term scheduler based on Linear Programming is not the
only, but nevertheless an important and essential component of a The objective function for the Linear and Mixed Integer Program
successful Draw Control System. The basic actions within this Draw Control System will be to maximise the NPV over Life of
specific component are shown in Figure 4. Mine and within each of the user defined multi-time periods. The
maximisation of NPV is closely associated with maximising ore
tons, as the ore tons generate revenue. The model should thus
minimise the mining of waste tons as they do not generate or
contribute towards revenue. This ensures that the Linear/Mixed
Integer Program not only achieves the best ore tonnage model
but the best ore tonnage model at the best profit. With the use of
discounting factors in the user defined periods the model will
always want to mine to a maximum earlier rather than later, as
this will optimise the NPV as stated by the objective function.

Resource input data

Vertical resource columns

The cave orebody is split up in vertical resource columns,
positioned above the drawpoints. The required input data for
these resource columns can be summarised as follows:
• opening resource in terms of ore tons at the start of the
planning horizon;
• estimate of the ore to waste mix in terms of a percentage
value;
FIG 4 - Iterative process within the Linear Programming based long-term • historical accumulated tons drawn per column;
scheduler.
• historical kimberlite tons drawn per column; and
• upper and lower capacity flowrates in terms of tons of ore per
During the initial examination into the use of Linear/Mixed period.
Integer Programming techniques to solve for an optimum
It is up to the geology and survey department to ensure that the
depletion strategy of a block cave it soon became apparent that
draw control officer running the Linear Programming application
the geotechnical rules and constraints would form the major and
has the correct data available at all times.
overriding control. These rules would limit areas of flexibility, as
one cannot select material to be mined from columns without
adhering to strict sequencing and relationship criteria between Stockpiles and dumps on surface
adjacent columns. Kimberlite orebodies are variable in nature Ore from resources on surface, eg stockpiles and dumps can be
and large variation in grade can occur within the pipe, which activated by the user to blend in with the ore streams from the
makes mining to a constant grade difficult and even more so if draw points to the plant.
geotechnical rules are adhered to which will limit areas for Both resource types contain opening volume (tons or cubic
accessibility. However it was felt that besides the meters), percentage kimberlite, percentage waste, grade, density,
all-encompassing geotechnical constraints it was still necessary stones, percentage concentrate, percentage rocktype(s).
to model the entire cycle of ore extraction and ore processing and
would thus also include maxima and minima constraints, per
period, for the following: Other resource data
1. Ore flow process, which will highlight the material that can The Linear Programming based long-term scheduler is also
pass from node to node. The nodes that have been modelled capable of incorporating the treatment of other resources such as
are tunnels, ore passes, haulage systems, underground sublevel cave blocks. If included into the program, input from
accumulation areas, shaft systems and treatment plant. these resources should be the same as for stockpiles and dumps.

MassMin 2000 Brisbane, Qld, 29 October - 2 November 2000 465

A R GUEST, G J VAN HOUT, A VON JOHANNIDES and L F SCHEEPERS

Constraints Metallurgical constraints

The Linear Programming scheme allows constraints to be set in • Treatment plant. The plant treating the ore produced from
terms of a minimum and maximum limit on any output per the cave or transported from the other resources can be
multi-time period. Setting these upper and lower bound included in the optimisation calculation. The required input
requirements needs careful attention and should be subject to for the plant is:
properly designed procedures, audited by consultants from the
relevant disciplines. • Upper and lower throughput capacities in terms of tons
ore per period.
Three major groups of constraints will be discussed in the
following sections. The first group is derived by strict Tolerated minimum and maximum ore to waste ratio in the
geotechnical rules imposed on the optimal solution. The next plantfeed expressed as a percentage.
group of constraints is dictated by mining limitations. The last The plant is defined as an ore process unit with an upper and
set of constraints includes straightforward requirements lower throughput capacity. All ore streams from the draw
prescribed by metallurgical requests. points via the haulage routes combined with the ore streams
from the surface resources are directed to the plant. The user
Geotechnical constraints has control over the maximum and minimum kimberlite to
waste ratios in the ore stream routed to the plant, ie drawn
• Column draw rates. The Linear Programming application from all the active draw points and drawn from surface
caters for a limited draw rate for each column within each resources. This control is available per time horizon.
time horizon. The column draw rates in the current Linear
Programming versions are expressed as tons per period.
Economic constraints
• Precedence of accumulated tons drawn. A set of
constraints is activated, which stipulates, that, a given column Escalation or de-escalation of various levels of costing per block
can only be drawn from, once a user defined tonnage has to take into account the perceived cost for each time period.
been drawn from its predecessors. Put differently, drawing Likewise for revenue, each period can be adjusted for a price
from a column is permitted, only if the accumulated ton increase or decrease. Associated with each time period is a
drawn from its predecessor adheres to a pre-specified discount factor use to estimate the present value of money for
minimum. This formulation generates one integer variable that particular time period. Further economic constraints are the
per column and its predecessor. required profit to revenue ratio and exchange rates to be used per
• Limits in differences of accumulated tons drawn between time period.
columns within time horizons. The model is constrained by
a set of equations to ensure that the accumulated tons drawn Geological constraints
from a column will be within a pre–specified range of Geological constraints like grade; carats, stones, size and density
accumulated tons of neighbouring columns. can be used per time period and can be useful in overall scenario
A further relationship is generated to nullify the constraint, planning.
once one of the columns has been depleted.
This set of constraints is dependent on its input from the rock INTERFACE BETWEEN USER AND LINEAR
mechanics and mine stability experts in order to limit the PROGRAM INPUT
development of ‘rat-holing’.
• Limits in ratios of tons drawn between columns within Owing to the complexity of Linear program input requirements a
time horizons. The model is constrained by a set of user-friendly front and back-end program was written to facilitate
equations to ensure that the tons drawn from a column, will the creation of the matrix. This interfaces with OMNI (matrix
be in a user specified ratios of tons drawn from neighbouring generator).
columns. A further relationship is generated to nullify the Currently two implementations of this interface are in various
constraint, once one of the columns has been depleted. This stages of commissioning and testing, one for a Front Cave, the
formulation generates two integer variables per predecessor other for a Panel Cave operation.
draw point. This set of constraints is dependent on its input When the user opens the front and back-end program a logon
from the rock mechanics and mine stability experts in order screen similar to Figure 5 appears.
to ensure a steady and acceptable angle within the cave back.

Mining constraints
• Ore flow constraints. The system has the ability to define
any number of relevant ore flow processes and the associated
capacity of these routes. The ore flow currently caters for:
• tunnels for the accumulated material from the
drawpoints;
• ore passes for accumulated material from various
tunnels;
• haulage for the accumulated material from ore passes;
• underground accumulation areas store material from
various haulage routes; and
• shaft systems for accumulated material from various
underground accumulation areas.
The user has control over the minimum and maximum flow
capacities of each ore flow route within each time horizon. FIG 5 - Opening screen when logging onto the draw control system.

466 Brisbane, Qld, 29 October - 2 November 2000 MassMin 2000

 AN APPLICATION OF LINEAR PROGRAMMING FOR BLOCK CAVE DRAW CONTROL

The front-end program is able to cater for different levels of As the Linear Program uses multi-time horizons and optimises
users and has an auditing facility to log any changes made during over the all periods, the user needs to define the number of
a session. periods by adding a period number, a description and how many
Depending on the user profile, the program will assign specific days are within that period. The user also has to set the minimum
rights to the user. A draw control officer for instance will only be and maximum draw rates, the status of the drawpoint (open or
allowed to run the simulations, whilst the geotechnical engineer closed) for each drawpoint, for each period.
may change any of the critical parameters. Vast experience and data from existing or ceased cave mines
If an authorised user changes crucial rules in the system, an allow De Beers personnel to assess what draw rates are
acceptable at the operation for which the Draw Control System is
audit table is created and the user is prompted to give a
devised.
justification why these changes were made. This information and
the user name are stored in the audit table, along with all changes At Premier Mine, maturity rules are in place to forecast at
made. what rate the drawpoints can produce, depending on the
drawpoint maturity (Bartlett, 1998). The maturity of a drawpoint
When the user opens a working scenario in the system, a is expressed in terms of months in production or tons produced
screen similar to Figure 6 appears. from that drawpoint, whatever the lowest value.
At Koffiefontein Mine, the column draw rate is expressed in
millimetres draw down per day. As the column draw rates in the
Linear Programming are required in terms of tons per period the
front to back end program must convert the user input in terms of
millimetres per day to tons per period.
A default maximum draw rate value is assigned to all
drawpoints during all multi-time periods. The user may change
these default values but these may not exceed a limit value kept
in a secure database. All users have read access to this database
but only a few authorised users may change these values. This
ensures that the system cannot be abused, on purpose or by
accident and management can at all times verify who made the
changes and what justifications were given.
When the user selects the ‘calls’ tab, a screen similar to Figure
8 appears. Here the user decides what the total production call is
per period that will be treated by the plant and how this is split
between the different resources.

FIG 6 - General information associated with a typical simulation

scenario.

Apart from the title, project name, date, name of the creator
and the comments the user may add, essential information
required here is which resource (Figure 3) has been used and
what accumulated mined data has to be loaded from the Vehicle
Monitoring System.
When the user selects on the next tab in the scenario window, a
screen similar to Figure 7 appears.

FIG 8 - Production calls per period.

As can be seen, the metallurgical constraints are not only put

in as a maximum limit to its capacity in tons but it also allows a
certain percentage waste of the total production mixture.
When the user clicks the ‘constraints’ tab, a screen similar to
Figure 9 appears.
For each period, a generic set of constraints is applied in terms
of geotechnical and mining related parameters.
Mining constraints could dictate that the accumulated mined
profile in a certain section should have an overall slope angle of
25 degrees. This angle ensures that there is enough time to
construct and open new drawpoints without obstructing the
global production.
Another constraint could be that there is a relation between
two neighbouring drawpoints depleting the same ore column.
The ratio between the production of the top drawpoint and the
FIG 7 - Period definition, drawpoint status and planned draw rates for a production of the bottom drawpoint could be designed at 40 to 60
scenario. (equals 66 per cent).

MassMin 2000 Brisbane, Qld, 29 October - 2 November 2000 467

A R GUEST, G J VAN HOUT, A VON JOHANNIDES and L F SCHEEPERS

When comparing Linear and Mixed Integer Programming

results to manual spreadsheet type systems there have been
up to 20 per cent improvements.
5. Ability to store the desired plan with its constraint set,
making the management, control and auditing task feasible.
6. The system is flexible and generic, allowing for many
different planning scenarios and the ability to pick and
choose from different resource combinations.
7. For feasible outcomes one always knows that the optimum
result for a given constraint set is being presented.
8. With the advent of new technology (hardware and software)
and the incorporation of new mathematical algorithms (eg
Gomory cutting plane integer algorithm), the run time of
Linear Programming based systems is decreasing rapidly.

ACKNOWLEDGEMENTS
The permission of the Director, Operations, De Beers
FIG 9 - The constraints table in the front to back end program. Consolidated Mines Limited, to present this paper is gratefully
acknowledged.
The authors acknowledge the work of Mr A Langbridge and
The user supplies the information in specified units (degrees or his team at Technical Systems, De Beers Consolidated Mines
ratio’s or tons) and the front end will translate these into tons, as Limited for the successful development and implementation of
required by the Linear Programming. the front-end program.
Exceptions to the general applied rules may be defined in a
separate window (see ‘Matrix Constraints’ button in Figure 9). REFERENCES
The variance between planned and actual production should be
kept within acceptable limits as determined by the constraints Bartlett, P J, 1998. Unpublished internal report. Premier Mine, De Beers
Consolidated Mines.
discussed under the heading ‘Constraints’ and illustrated in
Figure 9. However, if the actual production is in far excess of CPLEX 6.60, 2000. Registered trademark of ILOG, Paris, France.
what was planned by the long-term scheduler, some of the rules Gallagher, W S and Loftus, W K B, 1960. Block caving practice in De
may have to be violated to ensure that the system finds a solution Beers Consolidated Mines Limited. Mine Managers Association of
South Africa, Papers and Discussions.
for the next long term plan and rectifies the overall situation. The
Gallagher, W S and Loftus, W K B, 1961. Yielding arches for support of
strength of the Draw Control System described here is the ability
block cave scraper drifts. Ibid. Mine Managers Association of South
to provide a solution (by adjusting the rules), even if the Africa, Papers and Discussions, pp 439-458.
production is completely off track as what was planned by the PCBC, 2000. GEMCOM Software International Inc, Vancouver Canada.
Linear Programming application.
Graham-Taylor, T, 1992. Production Scheduling using Linear and Integer
Programming, in Proceedings, 1992 AusIMM Annual Conference, pp
CONCLUSIONS 159-162 (The Australasian Institute of Mining and Metallurgy:
Melbourne).
The major benefits of the Linear/Mixed Integer Programming Hoerger, S, Bachmann, J, Criss, K and Shortridge, E, 1999. Long term
system to previous systems have been established: mine and process scheduling at Newmont’s Nevada operations, in
APCOM 28th Symposium, Computer Aplications and Operations
1. Draw control is now an integral part of the overall business
Research to the Mining Industry, at the Colorado School of Mines.
cycle, which includes technical input from geology, mining,
Laubscher, D H, 1994. Cave mining, the state of the art, J S Afr Inst Min
metallurgy and finance. Metall, October, pp 279-293.
2. The Linear and Mixed Integer Programming approach OMNI 2.52, 2000. Registered Trademark of Haverly Systems Inc,
optimises over Life of Mine as well as multi-time periods Denville, USA.
within Life of Mine. In this way the model is always aware Owen, K C and Guest, A R, 1994. Underground mining of kimberlite
of the long term implications as well as solving short-term pipes, in XVth Congress, Johannesburg, SAIMM, Vol 1, pp 207-218
constraints, eg Linear Programming is used to decide when Peele, R, 1941. Practice at De Beers Diamond Mine, Kimberley, South
it is the best time to start support works in certain tunnels. Africa, (Eds: L Parkinson and H T Dickson) 3rd Edition, Ch 10, pp
This is only possible since an optimisation is done over a 392-398.
multi-time period. Scheepers, L and Wellbeloved, D, 1992.Optimisation of Integrated
Mining and Metallurgical Complexes by means of Linear
3. Fast turn around time regarding output, which means that Programming and Case Study. Survival Strategies for the
many options, can be run and analysed, giving the planner Metallurgical Industry, Johannesburg, South Africa (SAIMM).
and management flexibility regarding different options and Wagner, P A, 1914. The diamond fields of Southern Africa (Struik:
confidence in the outcomes. Johannesburg).
Williams, G, 1902. The diamond mines of South Africa – some account of
4. All actual production data (history) is stored allowing a their rise and development.
re-depletion and re-planning phase to take place, which can
be compared to the current or other depletion strategies

468 Brisbane, Qld, 29 October - 2 November 2000 MassMin 2000