Optimal Open Pit Short Term Planning Under Uncertainty and Blending Constraints
Optimal Open Pit Short Term Planning Under Uncertainty and Blending Constraints
Optimal Open Pit Short Term Planning Under Uncertainty and Blending Constraints
open‐pit short‐term planning under
uncertainty and blending constraints
Marcelo Vargas, Enrique Rubio and Nelson Morales
CUBE‐MINE Technology Ltda., Chile
ABSTRACT
Short‐term mine planning is a key point in the mining business, because it is when the profit
promise of long‐term plans is done. Despite that, most of the efforts in developing mathematical
models and techniques are spent in long‐term models whose output are based on many additional
information and simplified constraints. This means that the short‐term planner must deal with a lot
more variability in the data and meet constraints that were not taken into account when
constructing the long‐term plan. This often means that in order to satisfy the production goals,
imposed by the long‐term plan, the short‐term planner must either make changes in the
exploitation sequence, affecting future production, or incur in higher costs.
The aim of the study presented in this paper is to generate a tool considering such variables,
allowing for the generation of multiple decision scenarios done in a reasonable time, and providing
a guide for the mine planner, for short and medium term horizons. The focus of the model are the
blending constraints related to the geometallurgical attributes of the ore sent to the processing
plant, but it also considers the dynamics of stocks, geometrical constraints, plant and mine capacity,
among others, in order to determine a sequence of exploitation and processing such that the overall
ore production is maximised. It also includes the possibility to analyse mine operation design
selectivity using different loading and haulage equipments, according to the study of different basic
clustering blocks units.
The tool was implemented and applied at the BHP Billiton Spence mine, located in the north of
Chile, in order to obtain a plan for the quarter September‐November of 2008 and the year July 2009‐
July 2010. The results correspond to probability maps for the extraction period of each block,
graphical analysis of the reliability of production goal, and modelling of relations between
clustering unit sizes and the selectivity and functionality of the mine design.
INTRODUCTION
Mine planning is defined as the process of mining engineering in charge of transforming the mineral
resource into the best productive business, while a production schedule is a productive promise that
corresponds to a bankable document for investors [15]. Therefore, mine planning aims to determine
which part of the resource must be extracted in each period in order to maximise profits from mineral
processing, including decisions such as which part of the resource must be processed, the required
investment for the plant and hauling equipments requirements, among others.
Currently, mine planning process is made considering operational and metallurgical constraints,
building short and medium term mining plans through manual approaches of trial and error with
an important time investment to generate a result which no one absolutely knows if is the right
option which maximises the business in terms of recoverable metal, fleet utilisation and reserves
consumption among other items, since it is extremely difficult to include ore variability in the
production schedule using the available traditional software.
In consequence, the main decisions concerning Life of the Mine (LOM) are considered first, once
given these, further decisions are taken into account in relation to long term (10 to 30 years
depending on the deposit). Then, considering as fixed the results previously obtained, it is
necessary to develop plans for the medium term (usually up to five years) and so on till the daily
detail, going through the annual planning which compromises directly the operational budget. This
dismembered analysis structure leads to a massive scale mine exploitation, with large equipment
and high production rates, without taking into account that mineralogical variability of the body to
be exploited may require a greater selectivity degree in the extraction, ultimately leading to
significant discrepancies between long and short term (where it faces the real mineral variability)
incurring significant cost increases in order to achieve production goals involved.
RELATED WORK
Large open pit mine operation and administration is a huge and complex task, particularly for older
mines. Optimisation techniques can be applied successfully to resolution of important issues that
arise in mine planning and management. These applications include: ore body modelling and
reserves estimation; optimal pit design, determination of optimal production schedules,
determination of optimal operational design, determination of optimal mixtures, equipment
maintenance time determination, equipment replacement policies, and several others [3].
Mine planning process
In general terms, mine planning process can be divided into different levels according to the
characteristics of the decisions [1]:
• Strategic: Refers to exploitation methods choices, mining and processing capacity and overall
mineral reserve estimates. The main goal of strategic planning is to synchronise the market with
the available resources and the companyʹs mission statement.
• Tactic: It corresponds to processes specifications to make over the life of the mine, such as long
term production schedules and the use of equipment and processing plants inside the mine.
Tactic or conceptual planning determines how to achieve the target set earlier by strategic
planning. The result is the mining plan, which defines how and when resources are extracted in
each period of the mining business.
• Operating: They are daily made, for example, trucks address dispatching. Within the
operational planning, processes and operational rates resulting from the mining plan are
included. Here, the conceptual planning feedback takes place.
The strategic decision of determining the optimum final pit has been effectively treated using the
Lerchs‐Grossman algorithm [10] or the Picard flow networks method [13] (see also [2, 4]). These
methods are based on a block model which characterises a mineralised body. Moreover, the
problem of generating a production schedule can be defined specifically as the sequence in which
blocks should be removed from the mine with the objective of maximising the total profit, defined
according to the company strategic goals, subject to a series of physical and economic constraints.
This problem can be formulated using integer linear programming, however in real applications,
this formulation is too extended, both in terms of number of variables and number of constraints to
be solved with available MILP software. In consequence, it appears the necessity to divide the
problem, first establishing a final pit, and then sequencing the blocks contained in it.
A number of approaches for solving the sequencing and scheduling issue in the extraction of blocks
have appeared in the literature, including: heuristic [2, 8]; relaxation lagrangean; parametric
methods [5, 11, 15, 16], mixed integer linear programming [3, 4, 8]; and the application of artificial
intelligence algorithms such as genetic algorithms and neural networks [6]. Due to the complexity
and size of the problem, all these approaches suffer from one or more of the following limitations:
Is not possible to cover most of the constraints that arise, deliver sub‐optimal solutions without a
quality measure in most cases and they can only address problems of limited size. Smith [15] notes
that while long‐term planning is typically responsible for maximising the project value, short‐term
planning is commonly associated with a target based on production goals with maximum or
minimum limits of certain chemical constituents critical such as in mining coal with high sulfur
content, where the maximisation of sulfides production is considered. Smith [15] poses a mixed
integer programming model which is responsible for blocks scheduling extraction in the short term,
with the objective of maximising the production of material of interest, subject to certain constraints
on blending, while ensuring a simple scheme of horizontal and vertical constraints without
considering the presence of stocks. In the same line Camilo Morales [18], presents a model to
maximise the production of metal in a copper mine, subject to certain geometallurgical constraints.
Clustering
Aggregation mechanisms combined with optimal solution strategies have also begun to play a role
in block sequencing for the short‐term. Specifically, the authors try to reduce the model size by
combining blocks with similar properties. Ramazan [13] proposed an aggregation scheme that uses
linear programming to construct ʹfundamental treesʹ to reduce the number of blocks (entities) to be
sequenced. After performing this aggregation, the general problem of block sequencing can be
solved, but now with far fewer variables because of aggregation. The author applies his techniques
to a mine of about 40,000 blocks to be extracted for eight years. Boland et al. [19] uses binary
variables to meet precedences between blocks clusters, while continuous variables control the
amount of material removed of each cluster, as of each block in an aggregate. The authors
demonstrate their procedure using cases that contain up to 125 macro‐blocks aggregates in a model
of nearly 100,000 units. The times for a 1% solution close to the optimum, vary between thousands
and tens of thousands of seconds.
OPTIMISATION MODEL
The mathematical model that aims to deliver a guide for building short and medium term plans,
using a general outline of destinations and attributes for each block for copper mining will be
described as follows in this chapter. In relation to the destinations, it works on the basis that from a
copper mine, five main types of materials are extracted, which are: sulphides, oxides, mixed, sterile
and ROM. It also adds the chance of differentiate in campaigns the extracting of different minerals
lithologies and the stocks presence.
Clustering Tool
An obvious way to reduce the number of variables to solve the optimisation model is to select block
groups and add them in order to obtain a final model that now includes macro‐blocks.
Traditionally, this aggregation is done generating clusters of regular dimensions (see Figure 1), for
example, from blocks dimension of 10x10x10 to macro‐blocks of 30x30x30 where each macro‐block
containing 27 (33) of the original blocks, attributes as grades or pollutant contents are averaged to
obtain the final value for the macro‐block. This way of clustering blocks is useful and efficient for
porphyry ore bodies, where large volume of material with identical mineralogical characteristics
distributed uniformly but for ore bodies with chaotic nature in regard to the lithologies spatial
distribution, this might not be the right thing.
Given the need for a mechanism able to make a smart clustering, using as a programming platform
Python language, a tool capable of receiving a block model and giving a set of macro‐blocks was
developed. These macro‐blocks, hence MRU (operational unit resources) are the result of clustering
blocks according to a function defined for convenience, being also possible to intercept different
types of aggregations, for example, in the case of ore bodies with high mineralogical variability, is
possible to generate macro‐blocks due to the interception of a regular clustering (30x30x30) with a
clustering by the attribute lithology (see Figure 2).
This method of aggregation generates a group of irregular macro‐blocks creating the necessity of
implementing tools that could work properly, particularly in the definition of vertical and
horizontal precedences and extraction progress control from the access.
Figure 1 Clustering traditional methodology
Oxxide Block
Sulfide Block
Sulfide Block
nerated clusterring tool
Figure 2 Gen
Implem
mentation off graph struccture
Given the new scennario, which must be cap pable of hand dling irregulaar macro‐bloccks (MRU), itt is
necessaary to connecct the spatiall arrangemennt of these MR RU through a graph, in which
w a nodee is
directly
y associated with a MRU U and the verrtices to the connections
c g
generated witthin a radius of
neighbbours of each MRU (node).
By hav
ving each lev
vel of the model
m charactterised by a connected graph,
g we caan calculate the
t
minimum distance ffrom road acccess to each M MRU, which aallows the determination o of the precedeent
MRU ffor each MRU U.
Fig
gure 3 Minimu
um path between two MRU
Input and model parameters
The op
ptimisation model
m requiress as input a block
b model in
i which each h block has a spatial locatiion
(coordiinates) and weell‐defined attributes, table 11 lists these atttributes and aa brief descripttion of them.
Each block
b has a tonnage ton((ū), destination dest(ū) є є (OX, SP, ROM,
R WST) and
a a set of A
geometallurgical atttributes for eaach block: g(ū ū, a) for a = [11, ..., A]. The same attributtes are requirred
for eith
her oxides an
nd sulphides stocks g(OX, a), g(SP, a) ffor a = [1, ..., A]. Finally, R ROM blocks aare
alwayss processed on n a different lline than in th
he case of oxid des and sulph hides.
Table 1 General model notation
Symbol Meaning
B Set of blocks
ū Block coordinates
T Time horizon (number of periods).
t Time period (t є {1,2,…,T})
OX, SP, ROM, WST Block destination (Oxides, sulfides, ROM or
waste line processing)
A Block model sets (e.g. geometallurgical)
a Block model Attribute
CU(ū) Copper content in the block
ton(ū) Block tonnage
The model requires maximum capacity of plant and mine for every period. On the other hand, for each
attribute, the average range that should be maintained for the processed blocks must be specified. If we
consider time periods 1, 2, ..., T, then the mining and processing capacities will be identified as P(t) and
M(t), and the maximum average value of each attribute as Ma for a = [1, ..., A]. Each time period is also
labelled as sulphur or oxide, depending on the campaign to which it corresponds.
The extraction control from an access is obtained by incorporating the TL parameter, which
corresponds to the number of periods to extract a MRU account after having removed the previous
MRU. In addition to that, the FMM parameter (mine movement factor) is set, which allow
saturating the hauling capacity.
Variables
Regarding the model blocks, two sets of variables are considered, one for mining and the other for
processing. The first set determines when a block is extracted from the mine. For a block U and
t = [1, ..., T], where:
1 ū block is extracted in 1. . t
mū,
0 other case
The second set determines when a particular block is processed:
1 ū block is processed in 1 … t
pū,
0 other case
Similarly are denoted as p(OX, t) and p(SP; t), the number of processed blocks from the stock of
oxides and sulphur respectively in period t.
To simplify the writing of the model, it is useful to introduce some auxiliary variables relating to
mine and process variables:
m ū, t m ū, t 1 si t 1
Δm ū, t
m ū, 1 si t 1
p ū, t p ū, t 1 si t 1
Δp ū, t
p ū, 1 si t 1
Δ variables have the advantage of having a rather simple interpretation, for example Δm(ū, t) is
equal to 1 where a block (ū) is mined in exactly the period t.
Objective function
Each block has a copper content CU (ū) as an attribute. We want to maximise the amount of metal
produced in the horizon of T time periods involved in the problem. Only sulphide and oxide blocks
are considered in the objective function because these are the main sources of copper.
: CU ū Δp ū, t
ū є OX,SP,ROM
Constrains
The blocks can be mined and processed only once:
m ū, t m ū, t 1 , 1 t T 1 (1)
p ū, t p ū, t 1 , 1 t T 1 (2)
Only mined blocks can be processed:
p ū, t m ū, t 1 , 1 t T (3)
Stocks are limited:
p OX, s N OX 4
p SP, s N SP 5
Capacities: Both the mine capacity and processing capacity at the plant must be respected. For any
t = [1...T]:
ton ū Δm ū, t ton OX p OX, t ton SP p SP, t M t 6
ū
Vertical precedences: Previous blocks to a given target block, should be removed, that is, for each
(ū1, ū2) є Γ (set of arcs in the precedence graph for a given angle of slope) and t = [1, ...,T].
m ū ,t m ū ,t 8
Horizontal Precedence: This restriction require that once a block is extracted in a period t, there is a
TL limit of periods to extract the next block in the horizontal arc of precedence (see Figure 4), ie., for
each (ū1, ū2) є Π (set of arcs in the precedence graph horizontal) and t = [1...T].
1 m ū ,t m ū ,t TL m ū ,t TL 1 9
Figure 4 Horizontal precedence diagram
Campaigns: During an oxide campaign, there is no chance for sulphide blocks to be processed:
For all t = [1, …, T] and ū such that dest (ū) = SP.
Δp ū, t 0 10
p SP, t 0 11
Similarly, during a sulphide campaign, then neither oxide blocks nor blocks from the oxide stock
are processed: For all t = [1, …,T] and ū such that dest (ū) = OX.
Δp ū, t 0 (12)
p OX, t 0 (13)
Blending: For each blending attribute a= [1, …,A] we have a maximum average of Ma so:
Transport capacity saturation: Because the graph structure and the definition of horizontal
precedences that ensure the geometric connectedness of the solution, it is possible to add a
constraint that ensures the saturation of the hauling capacity in the mine, for all t = [1,…,T]:
Heuristics to reduce computation time
A sequencing problem is divided into periods, from the first (1), to the horizon (T). In general, it is
observed for a universe of fixed blocks, the difficulty of solving a sequencing problem increases
dramatically with the number of periods T. It is natural then to search for decomposition resolution
schemes in order to maintain the T value bounded. A possible technique to use is a pattern of
ʹwindowsʹ, which resolves the problem of sequencing for the first Tʹ<T periods, then this firsts
periods are fixed as part of the solution and then the technique continues solving for the following
non fixed Tʹ periods in an iterative way until the entire horizon have been solved, as seen in the
Figure 5.
This mechanism was successfully applied in the optimisation model described above, achieving a
significant reduction in computation time, verifying that, for certain window sizes, feasible
solutions are obtained with a gap of no more than 5% with a computation time about 80% lower to
those for solving the problem without this scheme. This reduction in computation time can make
the leap to analyse uncertainty associated with mine planning process, incorporating into the
production schedule calculation, geostatistical simulations of the resource model.
1 2 3 4 5 6 7 … T
Resolution for T' periods It is assumed that in this period there is no activity
RESOLVED Resolution for T' periods Periods without activity
RESOLVED Resolution for T' periods without activity
RESOLVED Resolution for T' periods
Figure 5 Time‐Frame Methodology
RESULTS
The model was tested both for a short (Quarter) and medium (Forecast) time horizon, using data
from Spence mine, owned by BHP Billiton. For the first case, the data used are those for the quarter
July to September 2008 and for the second one from fiscal year July 2008 ‐ July 2009.
The issue presented in this study was solved using an integer programming model, whose
implementation was done using the AMPL programming language, and whose resolution was
made using the commercial software ILOG CPLEX, version 10.2.
Spence Mining Company
The Spence mine is a company 100% BHP Billiton owned. The mine operation is located near the
community of Sierra Gorda in Region II of Northern Chile, approximately 50 km southwest of
Calama and 150 km northeast of Antofagasta.
In geological terms, it corresponds to a supergene enriched and partially oxidized porphyry copper
deposit from Upper Paleocene age (~ 57 M Years), 100% covered by gravels with mineralisation
depth ranges from 80 to 100 metres below surface, where the supergene sulphide and the oxide
mineralisation are associated with the presence of atacamite and chalcocite respectively. Five types
of materials are drawn from the mine, which are: sulphides, oxides, mixed, ROM and waste, those
which are classified agree to characteristics of the block model.
The operation is conducted through an open pit mine with leaching and SX‐EW (Solvent
extraction/Electro winning) production facilities, with a mining rate ex‐pit up to 260,000 tonnes/day
and an ore treatment rate of 50,000 tonnes/day for an approximate nominal annual production
capacity of 200,000 tonnes copper cathode. Finally, the cathodes are carried to either ports of
Antofagasta or Mejillones for shipping to customers.
Geometallurgical constraints
The processing plant imposes certain geo‐metallurgical constraints to obtain final copper cathode
with maximum recovery, ensuring maximum efficiency step in the process. The constraints that must
be met are:
Table 2 Geometallurgical Constraints
Attribute Upper limit
CO3 0.8 kg/t
Clacid 0.8 %
Blend 25 %
F_Geomet 20 %
Quarter Spence Application
It was worked with the block model for the period between July and September 2008, which has a
total of 9,976 blocks of 10 x 10 x 15 metres size. Model exploitation was roughly divided into two
weeks period (see Table 4), with certain exceptions such as when there is a change of campaign or
when days of the last week are less than the 14 required. The mathematical model considers the use
of two stocks, sulphides and oxides respectively, and its objective function is to maximise the
tonnage of copper recovered.
Table 33 Resolution parameters
LIM
MITS
Mine d CO3 F_Geomeet Blend
Plant Clacid
Perio
od Days Camppaign
(Kt) (Kt) (Kg/t)) (%) (%) (%)
1 14 Sulphides 3,080 700 0.8 1.2 20 25
2 14 Sulphides 3,080 700 0.8 1.2 20 25
3 14 Sulphides 3,080 700 0.8 1.2 20 25
4 14 Sulphides 3,080 700 0.8 1.2 20 25
5 14 Sulphides 3,080 700 0.8 1.2 20 25
6 7 Sulphides 1,540 350 0.8 1.2 20 25
7 14 Oxidees 3,080 700 ‐ 1.2 20 25
8 1 Oxidees 220 50 ‐ 1.2 20 25
Depending on minee and plant caapacities preseented in Tablle 4, saturation n of the transsport capacity y of
95% iss imposed (Eqquation 15). Finally, a slo
ope angle of 50% (Equatio
on 8) and a value of TL =
= 2
(Equattion 9) are esttablished, the latter param meter as explained above co orresponds to o the maximu um
time to ock on a path of minimum distance, afteer having extrracted its prev
o extract a blo viously.
Graph
h 1 Copper reccovered by leveel of aggregatio
on
Access
Access
Figure 6 Geometric results for Phase 2‐Bank 1580 with C40 (left) and C70 (right) clustering
CONCLUSIONS
Implementing a graph structure allows to define a pattern of vertical and horizontal precedence
consistent with the logical order of blocks removal in the operation of an open pit mine. On the
other hand, this restrictive scheme helps to reduce the range of feasible solutions, which combined
with the use of an time‐frame algorithm for ciplex resolution and the possibility of generate MRU,
can both reduce the calculation times for a total of 14,000 blocks in about 97%, allowing to solve the
problem in a personal computer at a time between 0.5 and 10 minutes depending on the level
of aggregation.
Adding the concept of operating exploitation units (MRU) that feeds the optimisation model, not
only allows to reduce significantly the model computation time and to address problems involving
a larger number of variables, but also provides a tool for incorporating minimum operating areas
for mining equipment, allowing analysis of the decrease or increase of the recovery associated with
selectivity degree that can be achieved with a certain profile of mining equipment. The decrease in
computation time and the use of MRU, allow address the problem of extract scheduling for a
greater number of blocks (100,000 blocks).
FUTURE WORK
The implementation of the optimisation model described in this paper implies an important step to
tackle the problem of uncertainty associated with the block sequencing problem for both the short
and medium term, incorporating blending and operational constraints, this thanks to the drastic
decrease in the calculation times. In this regard an important step goes in hand with the ability to
add intelligent way, with the development of the tool described in this document. The challenges to
address in the future are:
Improving resolution techniques: While in the model in its current stage, it is possible to deal with
a universe of 100,000 blocks, in particular for those involved in the medium‐term plan of Spence
mine, still need to make significant improvements in algorithms resolution if it is wanted to address
at some point the long‐term problem.
To study changes in the objective function: The implementing a graph structure allows to
incorporate the item of distances traveled to extract certain volume of blocks, this allows to aspire
some changings in the objective function that incorporate the cost item in the model.
Build model with variables destinations: In its current stage, the model requires each block to
have preset destination, this naturally affects the maximum value achieved by the objective
function, an important challenge is to incorporate this item as a variable, so that the destiny of each
block is the result of optimisation.
REFERENCES
Alexandra M. Newman, Enrique Rubio, Rodrigo Caro, Andres Weintraub (2007), A Review of Operations
Research in Mine Planning, Workshop on Operations Research in Mining. Pp. 1-13. [1]
Caccetta, L. and Giannini, L.M. (1986), Optimization Techniques for the Open Pit Limit Problem, Proc. Australas.
Inst. Min. Metall. 291p., pp.57‐63. [2]
Caccetta, L., Giannini, L.M. and Kelsey, P. (1998), Application of Optimization Techniques in Open Pit Mining,
Proceedings of the Fourth International Conference on Optimization Techniques and Applications (ICOTA’98)
(L. Caccetta et al. Editors.), Vol. 1, pp. 414‐422. (CurtinUniversity of Technology: Perth, Australia). [3]
Caccetta, L., Kelsey, P. and Giannini, L.M. (1998), Open Pit Mine Production Scheduling, in A.J. Basu, N. Stockton
and D. Spottiswood (Eds.), Computer Applications in the Minerals Industries International
Symposium (3rd Regional APCOM), Austral. Inst. Min. Metall Publication Series 5 p., pp. 65‐72. [4]
Dagdelen, K. and Johnson, T.B. (1986), Optimum Open Pit Mine Production Scheduling by Lagrangian
Parameterization, in: Proc. 19th APCOM Symposium of the Society of Mining Engineers (AIME)
pp. 127‐142. [5]
Denby, B., Schofield, D. and Bradford, S. (1991), Neural Network Applications in Mining Engineering, Department
of Mineral Resources Engineering Magazine, University of Nottingham, pp. 13‐23. [6]
Francois‐Bongarcon, D.M. and Guibal, D. (1984), Parametization of Optimal Designs of an Open Pit Beginning of a
New Phase of Research, Trans. SME, AIME 274, pp.1801‐1805. [7]
Gershon, M. (1983), Mine Scheduling Optimization with Mixed Integer Programming, Mining Engineering 35 p.,
pp. 351‐354. [8]
Gershon, M. (1987), Heuristic Approaches for Mine Planning and Production Scheduling, Int. Journal of Mining and
Geological Engineering 5 p., pp.1‐13. [9]
Lerchs, H. and Grossmann, I.F. (1965), Optimum Design of Open Pit Mines, Canad. Inst. Mining Bull. 58 p.,
pp. 47‐54 [10]
Matheron, G. (1975), Le Parametrage des Contours Optimaux, Technical Report No. 403, Centre de Geostatistiques,
Fontainebleau, France. [11]
Picard, J.C. (1976), Maximum Closure of a Graph and Applications to Combinatorial Problems, Management Sc. 22 p.,
pp. 1268‐1272. [12]
Ramazan, S., (2007). The new fundamental tree algorithm for production scheduling of open pit mines. European
Journal of Operational Research, Vol. 177, No. 2, pp. 1153‐1166. [13]
Rubio, E., (2006). Block cave mine infrastructure reliability applied to production planning. Thesis (Ph. D.).
Vancouver, Canada. The University of British Columbia. The Faculty of Graduate Studies (Mining
Engineering). [14]
Smith, M. L., (1998). Optimizing short‐term production schedules in surface mining: Integrating mine modeling
software with AMPL/CPLEX. International Journal of Surface Mining, Reclamation and Environment,
12 p., pp.149‐155. [15]
Whittle, J. (1993), Four‐D User Manual, Whittle Programming Pty Ltd., Melbourne, Australia. [16]
Whittle, J. (1998), Four‐X User Manual, Whittle Programming Pty Ltd.,Melbourne, Australia. [17]
Morales C. and Rubio E. (2010), Development of a mathematical programming model to support the planning of short‐
term mining, Proc. of the 34th Intern. Symp. on Application of Computers and Operations research in
the Mineral Industry (APCOM), Vancouber, Canada. [18]
Boland, N., Dumitrescu, I., Froyland, G., and Gleixner, A., (2009). LP‐based disaggregation approaches to solving
the open pit mining production scheduling problem with block processing selectivity. Computers and
Operations Research, Vol. 36, No. 4, pp. 1064‐1089. [19]