You are on page 1of 10

5th International Conference and Exhibition on Mass Mining, Luleå Sweden 9-11 June 2008

Developing an optimised production forecast at Northparkes E48 mine


using MILP

D. Rahal GijimaAst, Australia


J. Dudley Rio Tinto, Australia
G. v. Hout Rio Tinto, UK

Abstract
Rio Tinto is planning to develop a number of large block caves in the coming years. It is recognised that
planning and optimisation software will be required to test production forecasts early in the development of
these projects. One such program is the MILP developed as part of the industry sponsored International
Caving Study.
The Northparkes Endeavour 48 (E48) optimisation study will allow evaluation of the MILP software for
further development and use in Rio Tinto. The MILP is being used to identify the production consequences of
changes in draw strategy, assumed drawpoint and materials handling productivities, and rates of secondary
breakage.
This paper describes the development of a subset of the parameters used by the MILP. The major outcomes
are a preliminary optimised life-of-mine production plan and the identification of areas where additional
work can refine the parameters used in the optimisation.

1 Introduction
It has long been recognised that production scheduling is an important part of operating a profitable mining
operation. Its importance has increased in recent years as the industry shifts to mining marginal reserves at
high production rates (West-Hansen et al., 1986; Chanda, 1990). Block caving is gaining favour as one of the
preferred methods for extracting massive, low grade deposits due to its low unit cost and high production
capacity.
Production scheduling in block caving is generally referred to as “draw control”. The objectives of draw
control are normally separated into short and long term scheduling (Diering, 2004). Short term scheduling
seeks to adapt to local mining conditions in an effort to achieve monthly targets. Long term scheduling seeks
to achieve strategic corporate goals through its definition of the monthly targets. It has been recognised that
long term production scheduling has a major impact on mining economics in addition to its importance in
establishing realistic production targets (Farahmand and Fine, 1986).
In practice, a realistic production schedule must achieve forecast production rates while obeying
geotechnical constraints. Examples of these geotechnical constraints include the minimum and maximum
drawpoint production rate and the maximum production difference between adjacent drawpoints. These
constraints combine to determine the draw strategy. The importance of establishing an effective draw control
system is reflected by the active development of cave scheduling packages (Diering, 2000; Guest et al. 2000;
Diering, 2004; Rahal and Smith, 2004; Raña et al. 2004; Rubio and Diering, 2004).
The work of Guest et al. was extended as part of the International Caving Study (Rahal et al., 2003; Rahal
and Smith, 2004) with the development of a MILP based production module as part of its Integrated Draw
Control System (IDCS). This paper presents the use of this MILP for the Northparkes, E48 mine
optimisation studies. It focuses on the data and parameters required to develop a long term production plan
rather than examining the optimised schedule in detail.
2 Optimisation Data
Input to the MILP can be separated into three categories: cave state, production targets, and system
constraints. The first of these, cave state, defines the physical mining environment. Examples include
drawpoint reserves, previous mining history and the connections within materials handling system. The
second category, production target, specifies both the desired production and the schedule intervals (days and
months per period). The final category, system constraints, includes drawpoint minimum and maximum
draw tonnage, the permissible relative draw rate difference between adjacent drawpoints, drawpoint
availability and the capacity of the materials handling system.
The E48 MILP study primarily focused on the effect of changing the system constraints. Changes to the
following constraints were included in the study: different maturity rule systems, minimum draw rate,
drawpoint availability, varying relative draw rate limits, and varying materials handling system capacity.
The scope of this paper prohibits a full description of these trials. The goal is to outline the methodology
used to determine the model parameters which apply to the E48 optimised schedule. The optimisation data is
presented as follows:
• E48 Block Cave, insitu cave reserves
• Production Schedule, the schedule resolution selected for the life-of-mine plan (months per period)
• Maturity Rules, development of a mm/day based system to mature all drawpoints in 9-12 months
• Relative Draw Rate Constraints, allowed draw variation between adjacent drawpoints
• Drawpoint Availability, based on differences in haulage distance
The parameters used in this study are being refined as the E48 study progresses. Methods for improving the
optimisation input data are suggested, where appropriate.

2.1 E48 Block Cave


The E48 block cave is a proposed expansion of the Rio Tinto Northparkes Mine. Northparkes is a copper-
gold operation situated 350km west of Sydney in New South Wales (Figure 1). Historical production from
Northparkes includes two open-pit mines and two block caves (E26 Lifts 1 and 2). This third expansion of
the underground operations will be adjacent to the two E26 caves. Preliminary development of the E48 cave
has commenced with full production expected to begin in 2010.

Figure 1 Location of the Northparkes Mine (after Betts and Ross, 2005).

228
The current plan is for the E48 cave to produce 5.5 million tonnes per annum (Mtpa) through eight extraction
drives feeding a single gyratory crusher (Rio Tinto, 2006). The mine is currently reviewing the eight drive
layout with a view to expanding the design to include two additional extraction drives. An important aspect
of this review is to assess the productivity of the ten drive layout. Understanding the impact of the materials
handling system and associated drawpoint productivity levels on the life-of-mine (LOM) plan is critical to
making the right economic decision. As a result, the MILP study focused on modelling production from the
updated ten drive layout. A plan view of the cave with associated drawpoint tonnage is shown in Figure 2.
The drawpoint labelling convention is a combination of the extraction drive and drawpoint names (for
example ED04S01). Table 1 shows a summary of the basic parameters for the E48 cave. The footprint can be
characterised as having a high tonnage core oriented in the North-South direction. The legend shows colour
based on tonnage.

Figure 2 Distribution of insitu tonnage and maturity type within cave.

Table 1 Summary of the basic operational parameters for E48 block cave.

Parameter Value
Number Of Drawpoints 214
Extraction Drives 10
Drawpoint Mean Tonnes 199 kt
Drawpoint Min Tonnes 60 kt
Drawpoint Max Tonnes 408 kt

2.2 Production Schedule


The LOM production schedule for the E48 cave spans 96 months with a target production rate of 5.5 Mtpa.
The amalgamation of this time span into individual production periods will affect both the size (in memory
usage on a computer) and time required to solve the optimisation problem.
A series of trials were carried out to determine a reasonable compromise between period duration and
optimisation solution time. Schedules with more periods (and fewer months per period) have a higher
resolution but take longer to solve. The number of months per period increased throughout each schedule.
Figure 3 shows examples of this stepped increase for the 36 and 60 period trials. The 36 Period schedule has

229
three cycles of twelve periods: each having a period duration 1 month, 3 months, and 4 months. In
comparison, the 60 Period schedule is less granular with forty-two 1 month periods followed by eighteen 3
month periods. The figure also shows that both schedules generate schedules that span the same eight year
interval (circles).
The solution time for the 30, 36, 48 and 60 Period trials are graphed in Figure 4. These reference trials were
repeated as additional constraints were added to the production schedule. It is interesting to note that adding
the drawpoint minimum draw rate constraint to the production schedule decreased the time required to find
an optimum solution for three of the four schedules (diamonds). The 36 Period schedule was dropped from
the study after the baseline trials so there is no solution time data for the additional of the minimum draw rate
constraint.
The rapid solution time for the 60 Period trial indicated that it should be possible to optimise the full life-of-
mine schedule using single month periods. However the optimisation model would not load on the computer
with 1 Gigabyte of RAM. It is possible that a hardware upgrade later in the study will enable the solution of
a schedule based solely on single month periods.

Figure 3 An example of how months are agglomerated into production periods.

Figure 4 Effect of the number of periods on solution time for the baseline case and addition of
minimum draw rate constraints (circle and diamonds respectively).

230
2.3 Production Ramp-up
Maturity rules (also referred to as production rate curves; Diering, 2000) regulate the maximum drawpoint
draw rate based on the depletion of reserves above a drawpoint. As the cave is initiated, the maximum draw
rate must balance production rate with cave propagation rate to ensure that a large airgap does not form
above the broken rock mass. The impact of draw rate on fragmentation must also be considered as it has
been suggested that there is a relationship between draw rate and secondary fragmentation within the cave
(Laubscher, 2000).The changes in draw rate are normally classified as ramp-up to full production, steady
state production and ramp-down to drawpoint closure. The ramp-up duration is often quoted in terms of
either time (months) or percent draw (both height and tonnes). For the E48 study the stated goal was to
mature all drawpoints after mining for 9 to 12 months. The following two sections show the effect of
applying a global daily draw rate (Time Based Drawpoint Ramp-up) and the application of the maturity rule
systems within the MILP (Production Based Drawpoint Ramp-up).

2.3.1 Time Based Drawpoint Ramp-up


The E48 Pre-feasibility study used a time based global ramp-up regime as shown in Table 2. All drawpoints
were mined at a fixed production rate for each quarter in the first year. One possible handicap of using this
system with the wide range of draw column tonnages (60 to 408 kt, Table 1) in the E48 cave is shown in
Figure 5. At the end of the first year over 20% of the draw columns have been depleted by a third. The draw
columns with low insitu tonnage will close much earlier than the columns with higher tonnages.
The contrast between the time-based (month) and the depletion-based (maturity) systems can be illustrated
by selecting a depletion percentage for drawpoint maturity. Figure 6 shows a histogram of the time required
for drawpoints to reach maturity if the depletion threshold is 7%. (A threshold of 7% was selected because it
ensures that all drawpoints reach full production in the first twelve months). It can be seen that the majority
of the drawpoints reach maturity well before the target of 9 to 12 months. Increasing the maturity threshold
shifts the histogram to the right as all drawpoints take longer to mature if the maturity threshold is increased.
The E48 MILP study has chosen to apply a maturity system based on depletion status (x-axis) and vertical
draw rate in mm per day (y-axis). Five maturity profiles were used to ensure that all drawpoints reached full
draw after 9 to 12 months production. Drawpoints with a low tonnage were constrained by “slower” maturity
rules while high tonnage drawpoints were governed by the more aggressive, “steep” maturity profiles.

Table 2 Time based drawpoint ramp-up

Month t/d mm/d


1-3 50 69
4-6 70 96
7-9 90 123
10-12 120 164
12+ 200 274

231
Figure 5 Plot of drawpoint depletion after first year of production using original ramp-up.

Figure 6 Number of drawpoints reaching full maturity per period (depletion threshold 7%).

2.3.2 Depletion Based Drawpoint Ramp-up


The MILP model allows maturity profiles to be assigned on a drawpoint-by-drawpoint basis if required.
However it is more common to assign different maturity profiles to groups of drawpoints depending on insitu
reserves and/or local geology. The mechanism driving ramp-up variability in the E48 operation is the large
variation in column heights (hence highly variable insitu reserves) as shown in Table 1 and Figure 2.
Preliminary trials using the MILP indicate that five maturity classes (Figure 7) can be used to ensure that all
drawpoints reach full production in 9 to 12 months. The legend in Figure 7 shows the depletion level where
full maturity is reached (i.e. maximum draw rate of 200 t/d). The effect of these maturity rule profiles on
drawpoint ramp-up duration can be seen in Figure 8. All drawpoints reach maturity within the target interval.
The use of this differential ramp-up smoothes depletion rates across the cave by holding back production in
low tonnage drawpoints. This weakens the effect that the maximum drawpoint production rate has on
ensuring even draw. (The original drawpoint ramp-up scheme ensured even draw by restricting all
drawpoints to the same production rate.) However, even draw is maintained by applying the relative draw
rate constraints described in the next section.

232
Figure 7 Plot of different ramp-up (maturity) profiles for drawpoints within the E48 cave.

Figure 8 Number of drawpoints reaching full maturity per period (five depletion based
maturity types).

2.4 Relative Draw Rate Constraints


The relative draw rate (RDR) constraints are a fundamental part of ensuring that even draw is maintained
across the cave. It does this by limiting the difference in draw tonnage between adjacent drawpoints. The
benefits of maintaining even draw are twofold: it ensures that weight from the cave does not damage pillars
by preventing point loading, and it minimizes dilution by prohibiting isolated draw within the cave.
Even draw does not require all drawpoints to produce at the same rate. Typical relative draw limits for a
proportional Height-Of-Draw strategy range between two and four times the production of neighbouring
drawpoints. In the E48 MILP study, three ratios of relative draw were tested: 0.5 to 2.0, 0.375 to 2.67 and
0.25 and 4.0. These bound pairs reflect tight, intermediate and maximum binding limits respectively.
Preliminary trials indicated that the large column height differences needed to be recognised when
developing the E48 draw strategy. The best production results were achieved by enforcing a tight bind to all
drawpoint pairs for the first twelve months (0.5 to 2.0). After the first year, the relative draw rate binding was
varied as shown in Figure 9. The numbers associated with each draw column correspond to the maturity
profile types shown in Figure 7. The faint lines represent the tightest relationship pairs. The lightest are the

233
relative draw rate constraints with an intermediate bind. Finally, the maximum RDR constraints are indicated
by the darkest lines between drawpoints.
The RDR constraint type was assigned based on the difference in the maturity rule type assigned to the
drawpoint pairs. If the pair shared the same maturity type (roughly the same tonnage), the tight bind was
applied. If the drawpoint pair were of adjacent maturity types (for example 7% and 10% maturity depletion
values), then the intermediate RDR values were used. Finally, if the drawpoint pair had a large difference in
maturity type, the relationship was loosened to the maximum of 0.25 to 4.0.
The relative draw rate limitations for the E48 mine were based on rules of thumb and practical guidelines
from previous MILP studies. Rio Tinto is undertaking REBOP modelling in an attempt to quantify the effect
of different RDR constraint levels on material flow within the cave.

Figure 9 Schematic of different maturity types and different relative draw rate bounds.

2.5 Drawpoint Availability


Drawpoint availability can have a significant effect on both the ability to achieve a production target and to
maintain the ideal cave depletion profile. The three main factors affecting drawpoint availability considered
in this study were haulage distance, secondary breakage, and LHD interaction in the southern extraction
drives (ED07 to ED10). The E48 extraction layout and distribution of relative drawpoint availability within
the cave are shown in Figure 10.

Figure 10 The distribution of drawpoint availability within the E48 cave.

234
The effect of haulage distance on drawpoint availability was estimated using mineHAUL. The relative
capacity modifier for each drawpoint was calculated as a ratio of drawpoint haulage distance to the shortest
haulage distance. This resulted in drawpoints closer to the tip having a higher availability (total capacity).
The effect of secondary breakage on drawpoint availability was investigated by using the fracture frequency
domains (FFD) within the cave to estimate oversize and hang-up frequency. The draw columns above each
drawpoint were separated into 50m slices and classified according to their FFD. The preliminary secondary
breakage analysis indicated that differences in the FFD were not enough to cause a significant difference in
drawpoint availability.
The effect of LHD interaction has not been addressed to date. As part of continuing E48 optimisation studies
Arena will be used to investigate the effect of both secondary breakage and LHD interaction on drawpoint
availability. The results of the Arena study are expected to supersede the drawpoint availability estimates
currently used within the MILP.

3 Optimised Production Schedule


The optimisation constraints described above are among the most important of those that limited cave
production. Additional constraints on the production system include materials handling system capacity,
drawpoint minimum production rate, and LHD run-in (availability during the first year of production). A
series of thirty optimisation runs have been carried out to date as part of the MILP scheduling project.
Figure 11 shows the current optimised LOM schedule for the E48 cave.

Figure 11 An optimised LOM production schedule for the E48 mine.

It can be seen that the cave ramps up to its target production during the first seven months. It maintains this
rate for most of the cave life. The step change in production occurs because the number of months per period
increases from one to three in Period 43 (Figure 3, 60 Periods). The average monthly production for the three
month periods is shown as the dashed line for comparison to the single month periods.
The drop in production towards the end of the schedule (Periods 56 to 60) results from the MILP balancing
requested production with maintaining a smooth cave shape to reduce dilution from overlying waste. The
current objective function rewards maintaining production less than maintaining cave shape in the later
periods of the schedule as geotechnical considerations take precedence in the current formulation.

235
4 Conclusions
As part of the continuing optimisation of the E48 mine plan, a MILP optimisation model is being used to
examine the impact of different production constraints on total cave capacity. The strength of using the
MILP lies in its ability to generate realistic production schedules that require little manual manipulation. This
paper gives an overview of the process by which realistic constraint parameters are being determined for
inclusion in a LOM production schedule.
It was found that the relative draw rate (RDR) limit, drawpoint availability and the materials handling
system all have the potential to affect production rate. The RDR limits were based on empirical rules of
thumb and previous MILP experience. The present drawpoint availability and capacity limits on the
materials handling system warrant additional refinement before the end of the project. These input
parameters will be refined by using both REBOP and Arena. REBOP will be used to quantify the effect of
the current 0.25 to 4.0 and 0.5 to 2.0 relative draw limits on material flow. Arena will be used to include both
secondary breakage and LHD interaction within the LOM schedule. It is also expected that a hardware
upgrade will allow a complete LOM schedule to be developed based on single month periods.

Acknowledgements
The authors wish to thank Rio Tinto management and in particular Craig Stegman (General Manager at
Northparkes Mine) for their permission to publish this work. The authors also acknowledge the mine
personnel at Northparkes Mines for their cooperation in the E48 MILP Study.

References
Betts, M. and Ross I. (2005) ‘The Design, Installation and Commissioning of the Northparkes Mines’ Lift 2 Ground
Handling System, Hoist & Haul’, Proceedings (AusIMM) Perth, 33.
Chanda, E.C.K. (1990) ‘An Application of Integer Programming and Simulation to Production Planning for a Stratiform
Orebody’, Mining Science And Technology, 11(2), 165-172.
Diering, T. (2000) ‘PC-BC, A Block Cave Design and Draw Control System’, MassMin 2000, Brisbane, 469-484.
Diering, T. (2004) ‘Combining long term scheduling and daily draw control for block cave mines’, MassMin 2004,
Santiago Chile, 486-490.
Farahmand, D. and Fine, I. (1986) ‘A Practical Procedure for Underground Development and Production Scheduling
Using a Microcomputer’, 19th Application of Computers and Operations Research in the Mineral Industry,
Ramani,R.V. Society of Mining Engineers of the American Institute of Mining, Metallurgical and Petroleum
Engineers, Inc., Littleton, Colorado, 907-911.
Guest, A.R, van Hout, G., von Johannides, A. and Scheepers, L.F. (2000) ‘An Application of Linear Programming for
Block Cave Draw Control’, MassMin 2000, Brisbane, 461-468.
Laubscher, D.H.L. (2000), ‘A Practical Manual On Block Caving’, International Caving Study, October 2000, Brisbane,
Section 11.
Rahal, D., Smith M., van Hout, G. and von Johannides, A (2003) ‘ The use of mixed integer linear programming for
long-term scheduling in block caving mines’, 31st Application of Computers and Operations Research in the
Minerals Industries, South African Institute of Mining and Metallurgy, Johannesburg, 123-131.
Rahal, D. and Smith M. (2004) ‘A draw control system for scheduling production in block caving’, MassMin 2004,
Santiago Chile, 479-485.
Raña, F., Telias, M. and Vicuña, Mario (2004) ‘Controlled draw in block/panel caving’, MassMin 2004, Santiago Chile,
474-478.
Rio Tinto Internal Report (2006) ‘E48 Pre-feasibility Study, Northparkes Mines’.
Rubio, E. and Diering, T. (2004) ‘Block cave production scheduling using operation research tools’, MassMin 2004,
Santiago Chile, 141-149.
West-Hansen, J., Sarin, S.C. and Topuz, E. (1986) ‘Long-Term Production Scheduling in Underground Coal Mines –
An Application of Sequencing Theory’, 19th Application of Computers and Operations Research in the Mineral
Industry, Ramani,R.V. Society of Mining Engineers of the American Institute of Mining, Metallurgical and
Petroleum Engineers, Inc., Littleton, Colorado, 185-195.

236

You might also like