Вы находитесь на странице: 1из 14

Journal of Mining Science, Vol. 47, No.

2, 2011

INTEGRATED OPEN PIT PUSHBACK SELECTION AND


PRODUCTION CAPACITY OPTIMIZATION

T. Elkington and R. Durham UDC 539.3


Strategic mine planning for open pits has now evolved to the simultaneous optimization of a number of
decisions and this leads to the consideration of additional solutions, with the potential to add value to
planning and reduce manual data transfer time. Unlike previous efforts, this paper outlines a method for
simultaneously optimizing mining and processing capacities, including intermediate and ultimate
pushback selection, with the determination of scheduling, cut-off grade, and stockpiling to maximize net
present value.

Mine optimization, mixed integer programming, open pit scheduling, cut-off grade, production rates

INTRODUCTION

Strategic mine planning for open pit mines typically involves solving the problems of what should be
mined, where it should be sent, and when this should be done. There have been two main broad
methodologies for optimizing these problems, and these methods have been broadly classified as “block
level resolution” and “aggregation” approaches. The “block level resolution” optimization approach was
proposed by Johnson [1], who developed a linear programming (LP) formulation to optimize the timing
of each block’s extraction. Gershon [2] found that this formulation produced infeasible mining sequences,
and that integer variables were necessary. Ramazan and Dimitrakopoulos [3] showed that when integer
constraints were enforced, realistically sized problems were not solvable within a reasonable
timeframe, even with a series of simplifications. Caccetta and Hill [4] outlined a branch and cut
procedure to improve the tractability of this optimization approach. The “aggregation” approach aims
to make the problem tractable, with the global problem being split into several smaller sub-problems,
which are often solved sequentially. These sub-problems include the optimization of the ultimate pit,
intermediate pushback selection, scheduling, cut-off, and sometimes, stockpile optimization.
Over time, approaches have been developed to optimize these decisions individually. Lerchs and
Grossman [5] applied graph theory to the open pit outline problem to maximize undiscounted cashflow.
It is common to determine pushbacks prior to scheduling to reduce the problem size and introduce
practical mining geometries. Parameterization of price, mining cost [6] and grade certainty [7] has been
used to create a series of nested pit shells within the ultimate outline to guide in pushback selection.
The “aggregation” approach to the open pit optimization problem accumulates blocks into larger
entities to reduce the size of the scheduling problem. Many methods use a “panel” resolution, where a
panel is a unique combination of bench and pushback. Other approaches include LP based pre-
processing steps [8, 9]. Typical constraints on an open pit schedule include mining and processing
capacity, product quality (blending), and sequencing constraints that ensure slope angles and practical
advance rates are enforced.

School of Civil and Resource Engineering, University of Western Australia, Crawley, WA, Australia.
©
1062-7391/11/4702-0177 2011 Pleiades Publishing, Ltd.
177
Lane [10] provided mathematical rigour to the cut-off grade optimization problem. King [11]
implemented Lane’s theories for a fixed mining sequence with dynamic programming. Later
approaches have focused on the simultaneous optimization of multiple sub-problems, in order to
capture opportunities to add value through synergies between decisions. Combined schedule, cut-off
grade and stockpile optimization formulations, based on a Mixed Integer Programming (MIP)
approach, have been proposed [12 – 14]. Additionally, resource and market uncertainty has also been
built into open pit schedule optimization approaches [15 – 21]. Aspects that are not simultaneously
optimized with other sub-tasks are mining and processing capacities (referred to collectively in this
paper as production capacities), and the selection of pushbacks and ultimate pit outlines.
The optimization of production capacities is important because increased capacity must be
purchased at a large upfront cost, which may take many years to repay. Additionally, a reduction in
capacity does not often result in a significant recovery of that sunk cost. In the worst case scenario, this
cost is never repaid. Additionally, the selected production capacities will affect the selection of ultimate
and intermediate pit outlines, scheduling, and cut-off grade. Lane’s theory of cut-off grade found that
there were six possible optimal cut-off grade scenarios in any one period (mining limited, processing
limited, sales limited, and the three combinations of two of these limiting factors). The actual optimal
cut-off grade is driven by the grade distribution and the production capacities. The theory relies on
production capacities being pre-defined, and iterations of the planning process may be necessary to
determine the optimal production capacities. Alternatively, rules of thumb [22] that are based on
empirical relationship between processing capacity and the expected ore reserve may be used. An
interesting extension of Lane’s cut-off grade theory would be to incorporate production capacities as
decisions, with associated cost penalties.
Pushback selection is particularly important for mines that are limited by their vertical advance rate
or have limiting blending constraints. Pushbacks are often selected manually, typically from a nested
pit shell methodology based on the experience of the engineer and various empirical rules, which may
lead to suboptimal results. This paper utilizes an aggregation approach in a new formulation for
simultaneously optimizing production rates and pushback selection, along with scheduling, stockpiling,
and cut-off grades. The simultaneous optimization of all sub-problems allows relationships between all
decisions to be modeled; maximizing the opportunities to add value through mining in a more
strategically focused way. This optimization method also greatly reduces the number of analysis and
data transfer processes that are necessary in standard industry practice for open pit optimization.
METHODOLOGY
Definitions. Definitions are in reference to Fig. 1. A pit shell is the mining outline of an open pit
that maximizes undiscounted cashflows for a given set of slope constraints, revenues, and cost
parameters. Algorithms such as the Lerchs-Grossman [5] are typically used to automate this. Nested pit
shells (pits within pits) are usually generated by repeating the process while altering either the price or
cost parameters. This is achieved through utilizing the property of non-decreasing pit size within an
increase in price, or decrease in cost. The boundaries of these nested pit shells are candidates to be
intermediate pushbacks or the ultimate pit outline. A pushback is an incremental expansion of a pit
outline, often completed to reach material at a greater depth without violating slope constraints.
Pushbacks are often planned to blend the strip ratio throughout the life of the operation, with the
objective of reaching ore whilst minimizing the mining of waste that is directly attributable to it.
178
Fig. 1. Process schematic

A parcel is an aggregation of blocks which share common bench b and incremental pit shell h
properties (i.e. blocks in pit shell h but not in pit shell h – 1). A panel is the set of one or more parcels
allocated to the same pushback, located on the same bench. Each parcel within a panel is assumed to be
extracted at the same rate. Each block within each panel is allocated to a grade group g. It is assumed
that material in the same grade group, in the same panel, will be sent to the same destination(s) in the
same proportion(s). Importantly, the properties of the blocks within each grade group in each panel are
aggregated and averaged as appropriate. Thus, grade groups should include material with similar grade,
rock type, and metallurgical properties to achieve adequate data resolution. Smaller groups will closely
reflect the grade distribution in the resource, but this must be reconciled with a larger problem size.
Testing is often required to achieve an acceptable compromise between accuracy and efficiency. Grade
groups can be user defined, or automated with a clustering algorithm. The clustering process is most
useful in multi-element operations with multi-dimensional grade space.
Concept. Typically, open pit strategic mine planning is a sequential process of generating a series
of nested pit shells, selecting an ultimate pit, choosing intermediate pushbacks, selecting production
capacities, scheduling, and cut-off and stockpile optimization. This new method proposes only two
steps (Fig. 1). First, nested pit shells are generated over a range of prices. These act as a geometrical
pre-processing step to reduce the practical options for pushback and ultimate pit outline selection. Step
two involves the selection of pushbacks from the available nested pit shells is built in an MIP
scheduling formulation, which incorporates cut-off grade and stockpile optimization implicitly.
Production capacities (mining and processing) are treated as decision variables within the formulation,
and can be increased over time at a cost.
Key Assumptions. The key assumption made in this formulation is that companies orientate their
strategy to maximize NPV. Optimized strategy may be different when acting upon another objective or
combination of objectives. This section outlines the key assumptions associated with optimization of
each aspect of the mine plan: defining areas where the method is most applicable.
179
Pushback Selection. Pushback selection is determined by allocating parcels in each pit shell to a
pushback. As nested pit shells are utilized for this particular method, the method exploits the property
that subsequent pushback outlines must be defined by pit shells of increasing size. In using this pre-
processing step, the progression of the pit shells, and consequently pushbacks, is largely defined by the
“greedy” nesting process. Therefore, the method outlined is most applicable for vertically orientated
deposits without significant blending constraints, for which there are limited options for mine
sequencing. However, the method could be adapted to other situations by applying customized
assumptions when linking pit shells to pushbacks. Additionally, the formulation does not allow for a
minimum mining width to be applied explicitly. As a result, practical pushbacks may not be produced.
However, a minimum and maximum tonnage in each pushback can aid in simulating this type of
constraint within the formulation.
Ultimate Pit Outline Selection. The ultimate pit outline is defined as the outline of the final
pushback. This is implied in the pushback selection process.
Schedule Optimization. The schedule optimization defines the portion of each panel that is mined
in each period. It is assumed that each block within a panel is extracted at the same rate. Progression of
mining within a panel is not optimized or planned. Additionally, the panel sequencing constraints are
applied strictly between time periods, but not within them. If time periods are too large (i.e. when there
are few panels mined in each time period), these two issues may combine to produce impractical
schedules. This must be managed with careful schedule verification and appropriate time period length
selection.
Cut-Off Grade Optimization. Cut-off grade optimization is not explicitly applied in this method.
The cut-off grade is implicitly defined by the material contained within the grade groups selected for
processing. In a single element operation with no blending constraints, the cut-off grade can be
estimated as the lowest grade block contained within material selected for each process. If these
conditions are not present, the definition of cut-off grade is multi-dimensional. The accuracy of the cut-
off grade estimation is dependent on the number of grade groups applied. An increased number of
grade groups will likely result in tighter bounds for each grade group and a more accurate estimation of
cut-off grade. However as discussed previously, this must be reconciled with an increased problem size
associated with more grade groups, and resultant increases in solution time. For simplicity, the method
defined in this paper applies to only one processing destination. However, the formulation can be
scaled to suit complex processing paths.
Stockpile Optimization. From a modeling perspective, stockpiles apply a delay between mining a
panel and when material within the panel is processed. It is assumed that material can be removed from
the stockpile in any period after it has been placed there. There are no explicit restrictions on the
sequence of stockpile depletion. This method assumes the mining operation is able to carry out such a
plan and have control over the movement of material on stockpiles.
Production Capacity Optimization. Production capacities are included in the formulation as free
decision variables. Each unit of mining or processing capacity must be purchased at a given unit cost. It
is assumed that the relationship between production capacity and cost is linear, which may not be
accurate in all cases. Production capacities are allowed to have staged increases over time to minimize
upfront capital commitment. It is assumed that capacity cannot be sold. Therefore each unit of
additional capacity must be utilized profitably in present value terms. The costs of capacity increases
are assumed to be incurred in the year prior to their implementation, to simulate the lead time required
to order and install new plants. A greater lead time could be modeled, if necessary.
180
MIP FORMULATION
Decision Variables
There are nine main types of decision variables used in this model. All variables are continuous
between 0 and 1 unless stated otherwise.

mcbt Portion of bench b in pushback c to be mined in period t


whcbt Portion of pit shell h in bench b and pushback c to be mined in period t
p hbgt Portion of group g within pit shell h and bench b to be processed in period t
x hbgt Portion of group g within pit shell h and bench b to be added to the stockpile in period t
y hbgt Portion of group g within pit shell h and bench b to be reclaimed from the stockpile in period t
zhc [Binary] indicating whether pit shell h is allocated to pushback c (1) or not (0)
d cbt [Binary] indicating whether bench b in pushback c can be extracted in period t (1) or not (0)
mc t [Non-negative continuous] Added mining capacity in period t
pct [Non-negative continuous] Added processing capacity in period t

Coefficients
Whbt Discounted cost of mining bench b in pit shell h in period t

Phbgt Discounted net revenue (revenue minus processing cost) of processing group g within pit
shell h/bench b in period t
X hbgt Discounted cost of stockpiling group g within pit shell h/bench b in period t

Yhbgt Discounted net revenue (revenue minus processing cost and reclaiming cost) of reclaiming
from the stockpile and processing group g within pit shell h/bench b in period t
MCt Discounted cost of purchasing an additional unit of mining capacity in period t
PCt Discounted cost of purchasing an additional unit of processing capacity in period t
O hbg Ore tonnage in group g of pit shell h in bench b
G hbg Average available grade in group g of pit shell h in bench b
Qhb Total tonnage of pit shell h in bench b
S hbg Total recovered product in group g of pit shell h in bench b
Qcmin , Qcmax Minimum and maximum tonnage contained in pushback c
Bctmin, Bctmax Minimum and maximum bench advance for pushback c in period t
S tmin , S tmax Minimum and maximum selling limit in period t
Gtmin , Gtmax Minimum and maximum average grade limit in period t
Atmin , Atmax Minimum and maximum tonnage added to stockpile in period t
Rtmin , Rtmax Minimum and maximum tonnage reclaimed from stockpile in period t
Itmin , Itmax Minimum and maximum tonnage in stockpile at the end of period t
181
Objective Function
The objective of the optimization model is to maximize NPV (Equation 1). This incorporates the
discounted net revenue from material that is directly processed Phbgt and processed from the stockpile
Yhbgt , as well as costs from material added to a stockpile X hbgt and mining Whcbt . Also, the objective
function includes units of mining MCt and processing capacity PCt purchased over the life of the
operation:
 Phbgt p hbgt +  Yhbgt y hbgt −  X hbgt xhbgt −  W hbt whcbt −  MC t mc t −  PC t pc t . (1)
h ,b , g ,t h ,b , g ,t h ,b , g ,t h , c ,b ,t t t

Constraints
Quantity Constraints. These constraints relate to the capacity of mining, Equation (2), and
processing, Equation (3), in each period. These capacities are the accumulation of the units of capacity
added up to the given time period, and are consequently non-decreasing over time. Each additional unit
of capacity incurs a capital establishment cost, reflected in the objective function:
t
 Qhb whcbt ≤ 
h,c,b q =1
mcq ∀t , (2)

t
 Ohbg phbgt ≤  pcq ∀t . (3)
h ,b , g q =1

Product Constraints. These constraints relate to the minimum and maximum total recovered
product output, Equation (4), and average head grade feed for processing, Equation (5):
Stmin ≤  S hbg ( phbgt + yhbgt ) ≤ Stmax
h ,b , g
∀t , (4)

 Ohbg Gtmin ( p hbgt + y hbgt ) ≤  Ohbg Ghbg ( p hbgt + y hbgt ) ≤  Ohbg Gtmax ( p hbgt + y hbgt ) ∀t . (5)
h ,b , g h ,b , g h ,b , g

Pit Shell to Pushback Allocation Constraints. To avoid double counting, each pit shell is allocated
to a maximum of one pushback, Equation (6). Pit shells that are not allocated to a pushback are omitted
from the final mining outline. Additionally, material from each pit shell can only be mined from the
pushback it is allocated, and if a pit shell is allocated to a pushback, it must be fully extracted at some
stage, Equation (7). The “nested” nature of the pit shells used in this approach requires that a pit shell
may not be allocated to a lower numbered pushback than the pit shell directly preceding it,
Equation (8). Finally, all parcels allocated to the same pushback are mined at the same rate, Equation (9).

c z hc ≤ 1 ∀h , (6)

t whcbt − z hc = 0 ∀h, c, b , (7)


c
z ( h +1)c −  z hθ ≤ 0 ∀h, c , (8)
θ =1

whcbt − mcbt ≤ 0 ∀h, c, b, t . (9)


182
Pushback Constraints. Constraints may be applied to control the total tonnages contained within
each pushback, Equation (10). This may be used to achieve a minimum tonnage requirement for the
pushback. This constraint, when combined with Equation 8, can aid in identifying zhc variables that
must be zero. These variables may be eliminated in pre-processing the model before solving, potentially
improving solution efficiency. Additionally, the user can control the minimum and maximum number
of benches advanced in each pushback in every period with Equation (11).
Qcmin ≤ 
h,b,t
Qhb whcbt ≤ Qcmax ∀c , (10)

Qcmin ≤ 
h,b,t
Qhb whcbt ≤ Qcmax ∀c, t . (11)

Material Flow Constraints. It is important that material flow equilibrium is achieved throughout
the optimization from period to period. The amount of material processed directly and added to the
stockpile from a parcel progresses at the same rate as the mining of the parcel, Equation (12). Material
not directly processed or stockpiled is assumed to be sent to a universal waste dump. Secondly,
material cannot be removed from the stockpile until it has been placed there, Equation (13).
p hbgt + x hbgt −  whcbt ≤ 0 ∀h, b, g, t , (12)
c
t

q =1
( yhbgq − xhbgq ) ≤ 0 ∀h, b, g, t . (13)

Sequencing Constraints. In order to optimize the sequencing of panels, a series of constraints are
used to ensure that practical scheduling requirements are met. Equation (14) only enables mining in a
panel when it becomes available. This availability is dependent on the bench directly above in the same
pushback, Equation (15), as well as the same bench in the previous pushback, Equation (16), being
completed. Equation (15) assumes that bench numbering increases with depth.
t
 mcbα − d cbt ≤ 0
α =1
∀c, b, t , (14)
t
d cbt −  mc (b −1)α ≤ 0 ∀c, b, t , (15)
α =1
t
d cbt −  m( c−1)bα ≤ 0 ∀c, b, t . (16)
α =1

Stockpile Constraints. The stockpile constraints are used to control the addition, Equation (17), and
reclamation, Equation (18), of material from the stockpile over time, as well as imposing overall size
constraints, Equation (19):
Atmin ≤  Ohbg x hbgt ≤ Atmax ∀t , (17)
h ,b , g

Rtmin ≤  Ohbg y hbgt


h ,b , g
≤ Rtmax ∀t , (18)

t
I tmin ≤   O hbg ( x hbgt
h ,b , g q =1
− y hbgt ) ≤ I tmax ∀t . (19)

183
CASE STUDY
A hypothetical copper orebody with approximately 160 000 blocks was used to demonstrate the
proposed methodology. To begin, a pit optimization was completed with the Whittle strategic mine
planning package. Overall 31 nested pit shells were generated using a parameterized copper price,
ranging from 30 to 200 percent of the base case price and resulting in 454 parcels. The configuration
analyzed involves ore being fed to a flotation circuit and the resultant concentrate is sold. Some of the
key parameters and assumptions include: three pushbacks containing a minimum of 80 Mt of total
material (selection of these pushbacks is determined in the optimization), maximum advance of six
benches per year per pushback, unlimited stockpiling capacity (at a cost), a discount rate of 10 %, a
capital cost of $25/ tpa of processing capacity, and a capital cost of $5/ tpa of mining capacity.
Base Case. An optimization completed at the base case price of $3,300/t Cu yielded the production
schedule shown in Fig. 2. The NPV of this scenario is $1,031 million (inclusive of all capital and
operating costs). There are some interesting aspects to these results.
Both mining and processing capacities are fully established in the first year, and indicates that there
is no benefit to staging the installation of capacity over time for the given parameters. This may not be
the case if only low grade material was available in Year 1. The optimization was able to identify that
the revenue generated by having a high cut-off grade in Year 1 definitely compensated for the lesser
discounting of capital costs associated with full capacity installation in Year 1. Mining and processing
capacities are both fully utilized in each of the first eight years of the operation. This reflects the need for
every additional unit of capacity to be justified by utilizing it to bring forward revenues. This result was
achieved without the inclusion of time costs, which may be applied to encourage a similar result [14].
The stockpiling of material, rather than increasing capacity to absorb additional ore tonnes, indicates
that the operation is optimized with a processing limited configuration in these periods. The integrated
optimization is able to resolve the times later in the project when the project may become mining
limited (e.g. Year 7), and provide stockpiled material to fill any gap in processing tonnes.
Cut-Off Grade and Stockpile Analysis. The base case described above included the ability to raise
the cut-off grade above the marginal cut-off grade (the grade at which a block’s revenue exceeds its
cost of processing) and stockpile material. In practice, many operations do not consider stockpiling,
often for practical or ideological reasons. Also, many operations work to a simple marginal cut-off
grade, due to its ease of calculation, justification and communication. The use of the proposed method
of optimizing production capacities allows for the study of the impact of these common practices on the
mining and processing capacities that the operation is able to economically sustain. Three strategies
were considered for comparison: 1) stockpile optimization base case (Scenario SP OPT), 2) marginal
cut-off grade with no stockpiles (COG MARG), and 3) optimized cut-off grade with no stockpiles
(COG OPT). The production schedules for COG MARG and COG OPT are shown in the Appendix.
It is worth noting the difference in NPV returned by each of these strategies to the same resource
and constraint set. The predicted NPV of the COG OPT scenario is $995 million, and the COG MARG
scenario is $917 million. This represents a decrease in NPV of 3.5 % and 11.1 % from the SP OPT
scenario, respectively. These differences in NPV must be considered when deciding optimal mining
practice, i.e. does a simpler operation justify the sacrifice of significant value? Fig. 3 shows that all
these scenarios considered sustain a similar processing capacity of 25 – 28 Mtpa. The presence of the
stockpile in SP OPT reported the lowest optimal processing rate. COG MARG is encouraged to have a
184
higher processing capacity, bat higher financial risk, to enable high grade material to be reached earlier
when discounting effects are not as significant. Both SP OPT and COG OPT cases are able to increase
cut-off grade (Fig. 3b) and mine faster (Fig. 4) in order to access high grade material earlier. Additionally,
SP OPT is able to operate at full processing capacity for the life of the mine, whereas the other cases
both had processing shortfalls in some periods, and consequently revenue shortfalls, which may result
in liquidity issues.
Figure 3b shows very similar mining capacity and utilization for both COG OPT and SP OPT. The
difference in the cut-off grade between these two scenarios (Fig. 4) is accounted for by the difference in
processing capacity. The COG MARG scenario was the only case to increase either mining or
processing capacity over the life of the operation. The mining capacity was increased from 42 Mtpa to
50 Mtpa in Year 4. Purchasing the additional capacity earlier in the life of the operation would have
cost more in present value terms and would not have been fully utilized. Note the extent to which
mining and processing capacity is fully utilized for all scenarios in most periods (particularly earlier
periods). This analysis has highlighted the impact of operational mining practices on the need to
purchase expensive, risky capital equipment and plants. The most flexible method (SP OPT), which
allows stockpiling and cut-off grade to vary implicitly, produces the highest NPV with the lowest
capital expenditure associated with processing capacity. Thus, the downside risk associated with future
project closure is minimized. The combination of higher value for lower operational risk implies that
those operations currently employing a strict fixed marginal cut-off grade strategy without stockpiling
may produce suboptimal project outcomes.

Fig. 2. Base case production schedule

Fig. 3. Processing rate schedule, by cut-off and stockpile practice (a); combined mining and reclaiming rate
schedule, by cut-off and stockpile practice (b): 1 — SP OPT; 2— COG MARG; 3 — COG OPT
185
Fig. 4. Cut-off grade versus cumulative tonnes mined profile, by cut-off and stockpile practice: 1 — SP OPT;
2 — COG MARG; 3 — COG OPT

The term “optimization” can be misleading, considering uncertainty from many sources will often
make optimal solutions sub-optimal over time. A significant source of uncertainty is commodity price.
The influence of this parameter on strategy selection must be well understood to avoid suboptimal
project outcomes. To explore this, optimizations were completed at a number of different scenarios to
gain a greater understanding of the sensitivity of each of the major planning decisions to changes in
price. Five scenarios were considered:
• The base case price (Scenario A);
• 20 % below the base case price (L);
• 20 % above the base case price (H);
• The price falls from the H case to the A case over 5 years (HA);
• The price increases from the L case to the A case over 5 years (LA).
The final two scenarios (HA and LA) were selected to reflect possible planning when assuming
that conditions will improve or deteriorate over time. This method could be applied to many more
possible price paths, to gain greater understanding of the relationship between price and mining
strategy. The optimized production schedules for each of these scenarios are shown in the Appendix. A
direct impact of a price change on planning is to change the size of the final pit. It would be expected
that the size of the final pit would be proportional to the price at the end of the mine life. This is
supported by the results of the A and HA scenarios having the same final pit size. However, the LA
scenario resulted in a 10 smaller final pit size, perhaps due to its lower processing rate than the HA and
A cases (Fig. 6), and resultant increase discounting implications.

Fig. 5. Pushback sizes, by price scenario


186
Fig. 6. Low price scenario (L) production schedule (a); high price scenario (H) production schedule (b);
high to average price scenario (HA) production schedule (c); low to average price scenario (LA)
production schedule (d)

Fig. 7. Processing rate schedule, by price scenario

The processing rate comparison between the constant price scenarios shows that optimal
processing rates are highly sensitive to changes in price, with a 6.2 Mtpa (25 %) increase associated
with a 20 % increase in price from the base case, and a 9.3 Mtpa (37 %) decrease associated with a
20 % decrease in price from the base case. As a result of such a significant change in processing
capacity, the mine life of these three constant price scenarios is negatively related to price, even though
reserves are positively related to price. The scenarios that started with the lower price (L and LA) were
the only scenarios to indicate that processing capacity should be staged. The full capacity was
established in Year 2 for each of these cases. For each scenario, the full processing capacity is utilized
in all periods until the end of the mine life. Interestingly, the HA scenario reported a higher processing
rate than the H scenario, even though this case expected prices to drop in the future. This result stems
187
from the need to fully utilize periods of high price, when those periods do not last for the whole mine
life. The processing rate can be slightly lower when it is assumed that the high price applies at all
times. The LA scenario reported a higher optimal processing capacity than the L case, but lower than
the A case, to reflect the benefit of deferring ore tonnes until a more advantageous price outcome can
be gained. The mining rate tends to follow a similar trend to the processing rate (Fig. 7), in relative
terms. Figure 8b shows that the ratio between the mining rate and processing rate is similar (between
2.0 and 2.5) for all cases. A higher price assumption tends to lead to a slightly higher ratio. A
characteristic of all scenarios is the full utilization of mining capacity for the initial periods. The mining
rate only drops beneath the capacity in periods of stockpile reclamation at the end of the mine life. This
reflects the need for any additional upfront capital cost from increased mining capacity to be justified.
Unlike the processing capacity, the full mining capacity is established in the first period in all cases.
This reflects the need to mine through overburden to reach high grade ore as early as possible.
All price scenarios demonstrate similar stockpile utilization patterns (Fig. 9): an initial build-up on
stocks, followed by a minor depletion at the commencement of the third pushback, subsequent build-up,
and depletion at the end of the mine life. The maximum stockpile size is similar for all scenarios other
than the L case, indicating a relative insensitivity of optimal stockpiling strategy to a change in price.
The direct processing cut-off grade (Fig. 10) shows that all price scenarios reported a declining cut-
off grade strategy as being optimal, as suggested by Lane [10]. Remarkably, the trends and magnitudes
of all cases are quite similar. This indicates a relative insensitivity of optimal cut-off grade to price
changes when optimized in an integrated manner. The implication is that the optimized cut-off grade
and stockpiling strategy at a single price is likely to represent a near-optimal strategy for a wide range
of prices. However, optimal production rate is very sensitive to price and perhaps represents the most
important risk and return decision at the greenfield planning stage.

Fig. 8. Mining/reclaiming rate schedule, by price scenario (a); mining to processing ratio schedule, by price
scenario (b)

Fig. 9. Stockpile size versus cumulative tonnes mined profile, by price scenario
188
Fig. 10. Direct processing cut-off grade versus cumulative tonnes mined profile, by price scenario

Computational Experience. It is well known that the biggest issue with the implementation of MIP
formulations to large-scale problems is tractability. The original base case scenario, without the
application of problem size reduction strategies, had 76 157 variables, 59 990 constraints, and 797
binaries. This problem size was reduced to 34 461 variables, 46 007 constraints, and 693 binaries after
minimum and maximum pushback sizes were utilized to remove impractical pit shell to pushback
combinations. The reduced problem was solved in just less than 13 hours to within 1% of optimality
with the CPLEX v11 solver. To suit commercial applications, further reductions in problem size are
necessary. Unfortunately, the move of production capacities from constraints to decision variables
makes it difficult to define the eventual life of the operation and the time window in which panels can
be mined. Consequently, efficiency algorithms, such as those defined by Topal [23], could not be
applied. Future research will be conducted to improve the tractability of this method.
CONCLUSIONS
This paper demonstrated a new MIP method of simultaneously optimizing mining and processing
capacities, intermediate pushback, and ultimate pit outline selection in conjunction with optimized
scheduling, cut-off grade and stockpiling. The case study demonstrated how the cut-off grade and
stockpiling practices of the operation affects the optimal production capacities, as well as NPV.
Additionally, a parametric price study showed the relative insensitivity of optimal cut-off grade and
stockpiling, as well as the mining to processing rate ratio, to changes in price. The main strategic
aspects affected by price were the ultimate pit outline and production capacities. The ability to
simultaneously optimize multiple strategic elements enables a greater solution space to be considered,
and will likely lead to higher value solutions, with the potential to understand and possibly minimize
downside risk. Future work will focus on incorporating uncertainty, both market and geological, into
this planning framework. The ultimate aim is to enable more informed, robust strategic decision-
making for mining operations.

REFERENCES
1. T. Johnson, “Optimum open pit production scheduling,” PhD Thesis, University of California, Berkeley,
USA (1968).
2. M. Gershon, “Optimal mine production scheduling: Evaluation of large-scale mathematical programming
approaches,” Int. J. Mining Engineering, 1 (1983).
189
3. S. Ramazan and R. Dimitrakopoulos, “Recent applications of operations research in open pit mining,” SME
Transactions, 314 (2004).
4. L. Caccetta and S. Hill, “An application of branch and cut to open pit mine scheduling,” J. Global
Optimization, 27 (2003).
5. H. Lerchs and I. Grossman, “Optimum design of open pit mines,” Transactions of the CIM, 68 (1965).
6. J. Whittle and L. Rozman, “Open pit design in the 90’s,” in: Proceedings of Mining Industry Optimisation
Conference, Australasian Institute of Mining and Metallurgy, Sydney (1991).
7. S. Ramazan and R. Dimitrakopoulos, “Stochastic optimization of long term production scheduling for open
pit mines with a new integer programming formulation,” in: Orebody Modelling and Strategic Mine
Planning, Spectrum Series, 14, 2nd Edition The Australasian Institute of Mining and Metallurgy (2007).
8. P. Stone, G. Froyland, M. Menabde, B. Law, R. Pasyar, and P. Monkhouse, “Blasor - Blended iron ore mine
planning optimisation at Yandi,” in: Orebody Modelling and Strategic Mine Planning, Spectrum Series, 14,
2nd Edition The Australasian Institute of Mining and Metallurgy (2007).
9. S. Ramazan, “The new fundamental tree algorithm for production scheduling of open pit mines,” Eur. J.
Operational Research, 177 (2007).
10. K. Lane, The Economic Definition of Ore: Cut-Off Grades in Theory and Practice, Mining Journal Books,
London (1988).
11. B. King, “Optimal mine scheduling policies,” PhD Thesis, University of London, London (2000).
12. S. Hoerger, J. Bachmann, K. Criss, and E. Shortridge, “Long term mine and process scheduling at
Newmont's Nevada Operations,” in: Twenty-Eighth International Symposium on the Application of
Computers and Operations Research in the Mineral Industry, Colorado School of Mines, Golden (1999).
13. M. Menabde, G. Froyland, P. Stone, and G. Yeates, “Mining schedule optimisation for conditionally
simulated orebodies,” in: Orebody Modelling and Strategic Mine Planning, Spectrum Series, 14,
2nd Edition, Australasian Institute of Mining and Metallurgy, Melbourne (2007).
14. T. Elkington and R. Durham, “Open pit optimisation - Modelling time and opportunity costs,” Transactions
of the Institution of Mining and Metallurgy, Section A: Mining Technology, 118 (2009).
15. R. Dimitrakopoulos, C. Farrelly, and M. Godoy, “Moving forward from traditional optimization: Grade
uncertainty and risk effects in open-pit design,” Transactions of the Institution of Mining and Metallurgy,
Section A: Mining Technology, 111 (2002).
16. R. Dimitrakopoulos and S. Ramazan, “Uncertainty-based production scheduling in open pit mining,” SME
Transactions, 316 (2004).
17. M. Godoy and R. Dimitrakopoulos, “Managing risk and waste mining in long-term production scheduling of
open-pit mines,” SME Transactions, 316 (2004).
18. S. Ramazan and R. Dimitrakopoulos, “Traditional and new MIP models for production scheduling with in-
situ grade variability,” Int. J. Surface Mining, Reclamation and Environment, 18 (2004).
19. A. Leite and R. Dimitrakopoulos, “Stochastic optimisation model for open pit mine planning: Application
and risk analysis at copper deposit,” Transactions of the Institution of Mining and Metallurgy, Section A:
Mining Technology, 116 (2007).
20. A. Boucher and R. Dimitrakopoulos, “Block-support simulation of multiple correlated variables,”
Mathematical Geosciences, 41 (2009).
21. R. Dimitrakopoulos and S. Abdel Sabour, “Evaluating mine plans under uncertainty: Can the real options
make a difference?” Resources Policy, 32 (2007).
22. H. Taylor, Valuation and Feasibility Studies, Mineral Industry Costs, J. Hoskins and W. Green (Eds.),
Northwest Mining Association, Washington (1978).
23. E. Topal, “Early start and late start algorithms to improve the solution time for long-term underground mine
production scheduling,” J. Southern African Institute of Mining and Metallurgy, 108 (2008).

190

Вам также может понравиться