Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/318250075

Integration of mathematical models for ore mining industry

Article in International Journal of Systems Science Operations & Logistics · July 2017
DOI: 10.1080/23302674.2017.1344330

CITATIONS READS

12 353

2 authors:

Shi-Qiang (Samuel) Liu Erhan Kozan


Fuzhou University Queensland University of Technology
54 PUBLICATIONS 1,030 CITATIONS 159 PUBLICATIONS 3,710 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Mining Methodology: this work was partially supported by the National Natural Science Foundation of China under Grant nos. 71871064 and 18BGL003. View project

Reducing Transport Costs Through the Automation of Schedule Generation View project

All content following this page was uploaded by Shi-Qiang (Samuel) Liu on 31 July 2021.

The user has requested enhancement of the downloaded file.


International Journal of Systems Science: Operations &
Logistics

ISSN: 2330-2674 (Print) 2330-2682 (Online) Journal homepage: https://www.tandfonline.com/loi/tsyb20

Integration of mathematical models for ore mining


industry

Shi Qiang Liu & Erhan Kozan

To cite this article: Shi Qiang Liu & Erhan Kozan (2019) Integration of mathematical models for
ore mining industry, International Journal of Systems Science: Operations & Logistics, 6:1, 55-68,
DOI: 10.1080/23302674.2017.1344330

To link to this article: https://doi.org/10.1080/23302674.2017.1344330

Published online: 05 Jul 2017.

Submit your article to this journal

Article views: 122

View related articles

View Crossmark data

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=tsyb20
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS
2019, VOL. 6, NO. 1, 55–568
https://doi.org/./..

Integration of mathematical models for ore mining industry


a b
Shi Qiang Liu and Erhan Kozan
a
School of Transportation and Logistics, Southwest Jiaotong University, Chengdu City, China; b School of Mathematical Sciences, Queensland
University of Technology, Brisbane, Australia

ABSTRACT ARTICLE HISTORY


This paper investigates the integration of planning and scheduling theory for open-pit ore mining Received  April 
industry under comprehensive mining constraints and attributes. Consequently, mine design plan- Accepted  June 
ning, mine block sequencing and mine production scheduling problems are modelled, solved, inter-
KEYWORDS
acted and integrated in a whole system. A series of interrelated mixed integer programming models Integrated mining
and hybrid solution approaches are developed. Based on a real-world case study, it is validated that optimisation; mine design
the integration of mathematical models is promising to significantly improve the overall efficiency of planning; mine block
mining management in an analytical way. sequencing; mine
production scheduling;
mixed integer programming

ABBREVIATIONS
MDP: mine design planning
at the strategic level, MBS:
mine block sequencing at
the tactical level, MPS: mine
production scheduling at the
operational level, MIP: mixed
integer programming

1. Introduction Lerchs–Grossmann approach. Underwood and Tolwin-


ski (1998) developed a dual simplex approach to solve a
Nowadays, mining activities take place all over the world
mathematical programming model of the MDP problem.
and have become a major source of a country’s natu-
Hochbaum and Chen (2000) presented a detailed study
ral wealth, especially for Australia. Modern mining is
of the so-called push-relabel network flow algorithm in
a complicated procedure that may sustain over many
comparison to the Lerchs–Grossmann approach. Topal
years and necessitate huge expenditure. This paper is
and Ramazan (2012) developed a mathematical formu-
concerned with the integration of different-level math-
lation model based on the network analysis to optimise
ematical programming models for optimising open-pit
the mine exploration plan at the strategic level.
resource extraction.

1.2. Mine block sequencing


1.1. Mine design planning
After the determination of the ultimate pit contour, the
In open-pit mining, the initial vital optimisation prob- next vital optimisation problem type is called mine block
lem type, known as mine design planning (MDP), aims sequencing (MBS). The purpose of MBS is to answer the
to find the optimal answer at the strategic level, that is, question at the tactical level, that is, which part of the
what to be mined or what is the ultimate pit contour orebody should be selected to be mined over mid-term
that yields the maximum net present value based on the periods. In the literature, the following notable papers
estimated geological information. The leading papers on dealt with MBS. Caccetta and Hill (2003) proposed a
MDP in the literature are as follows. As pioneers, Lerchs general mixed integer programming (MIP) model to
and Grossmann (1965) presented to the mining commu- maximise the profit in certain time periods over which
nity the methodology known as the Lerchs–Grossmann the blocks are sequenced. Boland, Dumitrescu, Froyland,
approach for MDP. Caccetta and Giannini (1988) pro- and Gleixner (2009) developed a column generation
posed several mathematical theorems to improve the approach for solving this type MIP model, in which

CONTACT Erhan Kozan e.kozan@qut.edu.au


©  Informa UK Limited, trading as Taylor & Francis Group
56 S. Q. LIU AND E. KOZAN

blocks are aggregated for reducing the problem size. comprehensive literature review (Kozan & Liu, 2011; Liu
Bley, Boland, Fricke, and Froyland (2010) improved & Kozan, 2012a; Newman, Rubio, & Weintraub, 2010)
this MIP formulation by adding inequalities derived by on the applications of operations research approaches to
combining the precedence and production constraints. mining industries, most of the research work in min-
Ramazan (2007) proposed a method to aggregate a sub- ing optimisation focused on the long-term MDP and
set of blocks as branched trees, which are able to reduce mid-term MBS problems. In Australia, mining compa-
the number of integer variables and the number of con- nies such as BHP Billiton and Rio Tinto are keen to
straints required within the MIP formulation. However, minimise the operational costs and maximise productiv-
many researchers indicated that solving the MIP model ity of mining equipment by the adoption of advanced
of MBS is computationally demanding for large-size operational-level MPS software. To meet the industrial
instances (e.g. up to one million blocks), thus leading requirements, Kozan and Liu (2012, 2016a, 2016b) and
to the development of numerous heuristic algorithms. Kozan, Liu, and Wolff (2013) proposed the operational-
Kumral and Dowd (2005) developed a simulated anneal- level MPS models for Australian ore mining industry
ing metaheuristic combined with Lagrangian relaxation. to optimise drilling, blasting and excavating operations
Ferland et al. (2007) modelled the MBS problem as a as well as the railing–stockpiling–shipping operations in
resource-constrained project scheduling problem, which mine supply chain scheduling.
was solved by a particle swam optimisation metaheuris-
tic algorithm. Myburgh and Deb (2010) reported an
application of evolutionary algorithm in which an initial 1.4. Contribution and innovation
feasible sequence of blocks represented as a chromo-
In this paper, the main contribution is to integrate a
some is iteratively improved by genetic operators such
series of extended MDP, MBS and MPS models for estab-
as crossover and mutation. Cullenbine, Wood, and New-
lishing a better overall mining management system. A
man (2011) developed a so-called sliding-time-window
real-world case study is reported to show the benefits of
heuristic for MBS under maximum production and
such a systematic integration. To the best of our knowl-
processing capacities. Chicoisne, Espinoza, Goycoolea,
edge, this study is pioneering to fill this gap in the liter-
Moreno, and Rubio (2012) developed a quick heuris-
ature. The interactive relationship among strategic-level
tic algorithm based on topological sorting techniques
MDP, tactical-level MBS and operational-level MPS is
for solving large-size MBS instances. Lamghari, Dim-
portrayed in Figure 1. As shown in Figure 1, the output of
itrakopoulos, and Ferland (2015) developed a hybrid
MDP at the strategic level is used as the input for MBS at
approach by integrating MIP and variable neighbour-
the tactical level; then the output of MBS is implemented
hood search for solving MBS. Mousavi, Kozan, and Liu
as the input for MPS at the operational level.
(2016a, 2016b) developed an extended model and new
The remainder of this paper is organised as follows.
solution approaches to optimise the short-term sequenc-
In Section 2, a series of extended/enhanced MDP, MBS
ing of mine blocks by incorporating more practical
and MPS models are developed and integrated in a whole
constraints such as blending, stockpiling and processing
management system. In Section 3, a real-world case study
requirements. Liu and Kozan (2016a) developed two
based on the real-world data is reported to show the ben-
cutting-edge graph-based algorithms without relying on
efits. Finally, we conclude the contribution and signifi-
any MIP solver to solve large-scale benchmark instances
cance of this research in the last section.
from MineLib (Espinoza, Goycoolea, Moreno, &
Newman, 2013).
2. Mathematical models
Currently, there are many challenges facing open-pit ore
1.3. Mine production scheduling
mining industry such as increasing mining costs, shrink-
After the selection of blocks over mid-term periods is ing mineable reserves, declining ore grades and enlight-
determined, the operational-level question, i.e. how/when ening environmental protection awareness (Wetherelt &
the mining equipment units (e.g. loaders, drills and exca- Van Der Wielen, 2011). It is a tendency that most open-
vators) are allocated to the selected blocks to perform the pit mines become larger and larger in scale. A large-
mining tasks at various processing stages (e.g. drilling, scale open-pit mine is becoming an increasingly com-
blasting and excavating) over short time intervals, will plex and interdependent system that should be optimised
be answered by short-term multi-resource multi-stage by careful coordination, management and harmonisa-
mine production scheduling (MPS). According to recent tion of its individual elements. In this section, to address
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS 57

Figure . Interactive relationship among MDP, MBS and MPS.

these challenges, a series of strategic-level (long-term) • Benches are divided into working benches and inac-
MDP, tactical-level (mid-term) MBS and operational- tive benches. The bench levels are the important
level (short-term) MPS models are developed and inte- coordination elements of mine geometry.
grated in a system to achieve the overall mining efficiency • A parcel is the set of blocks which are both on the
improvement. same bench and in the same nested pit shell. Usually,
each block within each parcel is allocated to the same
grade group.
2.1. Strategic-level MDP formulation • A panel is the set of one or more parcels in the same
A geometry layout of a typical open-pit mine is depicted pushback.
in Figure 2. The analysis of underlying connotations in • Ultimate pit limit is the contour of an open pit that
the geometry is important to understand the character- maximises the long-term net present value based on
istics of MDP problems. According to the SME mining a 2D (3D) block model.
handbook (Wetherelt & Van Der Wielen, 2011), one of • An aggregate (a mining job at the operational level) is
the main considerations in the design of open-pit mine a portion of a parcel or a subset of blocks on the same
geometry is to accommodate the mining equipment (e.g. bench and in the same grade group. In an aggre-
excavators and trucks) and to satisfy the slope constraints gate, it is usually assumed that the properties of all
between the blocks. blocks should be identical. Thus, the blocks in the
Based on Figure 2, the definitions of crucial elements same aggregate can be sent to the same destination
in open-pit mine geometry are presented as follows: at the same production rate.

• A block (unit) is the smallest unit in open-pit mine To design a pit, the entire volume of an orebody is sub-
geometry, e.g. with a width of 10 metres, length of divided into blocks and the values of blocks are estimated
10 metres and height of 15 metres. according to the geological and geostatistical information
• Nested pit shells (pits within pits) are usually cre- from drilling samples in exploration. Precedence rela-
ated by replicating the MDP process while adjusting tionships of blocks must be satisfied because they are the
the values of key parameters (e.g. either ore price or vital constraints to specify the slope requirements of the
mining cost) in a specific block model. pit and to guarantee feasibility requirements. For exam-
• A pushback is an incremental expansion of a pit out- ple, Figure 2(e) illustrates the precedence relationship of
line and regarded as an aggregation of several nested a 2D block model for mine design. In most real-world
pit shells. Pushbacks are often selected by blending cases, it is required that the pit slope cannot exceed 45
the strip ratio (the waste volume: the ore volume) degree, which means that a block (block 1) cannot be
throughout the mining phases. For example, a 5:1 mined until the one (block 2) located directly above it
strip ratio implies that excavating one cubic metre and its two immediate neighbours (blocks 3 and 4) have
of ore requires removing five cubic metres of waste been removed. Obviously, the 3D block model is much
during a mining phase. more complicated than the 2D block model because of
58 S. Q. LIU AND E. KOZAN

Figure . A sample open-pit mine geometry: (a–b) side views of an open-pit mine; (c) top view of an open-pit mine; (d) a bench; (e) slope
constraints in a D block model; (f) slope constraints in a D block model.

more complicated precedence relationships. Figure 2(f)


illustrates the precedence relations of a 3D block model, ε(zi −1) ×nx + xi ≥ εi ,
in which a block (block 5) cannot be extracted until the i = 1, . . . , nb|w(zi −1) ×nx + xi >  and (1.4)
blocks (i.e. blocks 6, 7, 8, 9 and 10) on the level directly 1 ≤ (zi − 1) × nx + xi ≤ nb
above and one of four blocks facing this block on the same ε(zi −1) ×nx + (xi −1) ≥ εi ,
level (i.e. one of four blocks directly under blocks 7, 8, 9 i = 1, . . . , nb|w(zi −1) ×nx + (xi −1) >  and (1.5)
and 10) have been excavated. 1 ≤ (zi − 1) × nx + (xi − 1) ≤ nb
Based on the above explanation, an enhanced long- ε(zi −1) ×nx + (xi +1) ≥ εi ,
term open-pit MDP model, which is directly imple- i = 1, . . . , nb|w(zi −1) ×nx + (xi +1) >  and (1.6)
mented in the integration process, is formulated as fol- 1 ≤ (zi − 1) × nx + (xi + 1) ≤ nb
lows.
In the model, the list of parameters include: nx is the
... An enhanced MDP model number of total x-coordination values in a matrix; nz is
Objective: the number of total z-coordination values in a matrix; nb
is the number of total blocks (nb = nx × nz ); xi is the
x-coordination value of block i; zi is the z-coordination

nb
value of block i; i is the index of a block, which is deter-
Maximise : (wi − ci )εi , (1.1)
i=1
mined by zi × nx + xi ;  is a constant negative value; wi
sub ject to is the mining value of block i; and ci is the mining cost
εi ∈ {0, 1} , (1.2) of block i. The decision variable εi is binary, i.e. equals
i = 1, . . . , nb; one if block i is selected to be mined in the ultimate pit
limit, and zero otherwise. The objective function defined
εi = 0,
(1.3) in Equation (1.1) is to maximise the total net revenue of
i = 1, . . . , nb| wi ≤ ;
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS 59

blocks to be mined and thus to determine the ultimate pit once over time periods. Equation (2.3) satisfies the prece-
limit of an open-pit mine. Equation (1.2) defines that the dence relationship of blocks. Equation (2.4) requires that
decision variable εi is binary. Equation (1.3) requires that the minimum and maximum capacities (Rmin max
rt and Rrt )
a block is not considered if its block value is not greater of resource type r are satisfied in time period t, where
than a constant negative value. Formulae (1.4–1.6) satisfy gbr is the usage of resource type r by block b. Equation
the slope constraints which imply that if a block will be (2.5) declares that xbt is a binary variable, which equals
mined, then its pre-described immediate predecessors in one if block b will be mined by period t. Equation (2.6)
the matrix should be mined beforehand. defines that each block should be ready by the start of the
first period. Equation (2.7) declares that ybdt is a semi-
continuous variable between 0 and 1, which is the por-
2.2. Tactical-level MBS formulation tion of block b delivered to destination d in time period
After a set of blocks is selected by the strategic-level t. Equation (2.8) defines that if a block is selected to
MDP model, the purpose of tactical-level MBS model be mined in a period, then the entire block should be
aims at deciding in which periods these blocks should be extracted and delivered in portions to certain destina-
extracted under specific constraints. Commonly, in most tions.
MBS formulation models, the critical decision variable For other extended versions of MBS models with the
(xbt ) is defined in such a way that it equals one if a block consideration of additional constraints such as the grade
b is mined in a period t or zero otherwise. control and blending requirements, the inventory con-
One extended MBS model in the integration system is trol of stockpiles, the rehandling of materials from stock-
described as follows. piles to processing plants and the capacities of excavators,
please refer to our previous work (Liu and Kozan, 2016a;
... An extended MBS model Mousavi et al., 2016a, 2016b).
Objective:
  2.3. Operational-level MPS formulation
Maximise : ybdt − yb,d,t−1 pbdt , (2.1)
b∈B t∈T d∈D
The operational MPS model aims to optimally allocate
the critical mining equipment units at each operational
subject to stage and determine the timetable (e.g. starting, process-
ing and completion times) of mining equipment units
in the most efficient manner. In the MPS model, a min-
xb,t−1 ≤ xbt , ∀b ∈ B; t ∈ T (2.2) ing job is an aggregate of blocks selected in a short-term
period t, which was determined by the tactical-level MBS
xbt ≤ xb t , ∀b ∈ B; b ∈ b |b ⊂ B; t ∈ T (2.3) model. In a short-term MPS period, each mining job will
  be processed at several operational stages. Based on the
rt ≤
Rmin ybdt − yb,d,t−1 pbdt gbr ≤ Rmax
rt , above analysis and the fundamental job-shop scheduling
b∈B d∈D (2.4) formulation techniques (Pan, 1997), a basic MPS model
∀t ∈ T ; r ∈ R is developed as follows.
xbt ∈ {0, 1} ∀b ∈ B; t ∈ T (2.5)
... Basic MPS model
xb0 = 0, ∀b ∈ B (2.6) Objective:

ybdt ∈ [0, 1] ∀b ∈ B; t ∈ T (2.7)


Minimise max CiK , (3.1)
 i
xbt = ybdt , ∀b ∈ B; t ∈ T (2.8)
d∈D subject to

Defined by Equation (2.1), the objective function is



Lk
to maximise the total net present value of blocks to be xiklk = 1, i = 1, . . . , I; k = 1, . . . , K; (3.2)
mined over a set of periods denoted as T . pbdt is the net lk = 1
present value of block b delivered to destination d|d ∈ D,

Lk
where D is denoted as a set of destinations such as the yii klk ≤ 1, i, i = 1, . . . , I|i = i ; k = 1, . . . , K;
processing plants, stockpiles and waste dumps. Equation lk = 1
(2.2) ensures that each block is mined no more than (3.3)
60 S. Q. LIU AND E. KOZAN

mining jobs on each assigned mining equipment unit at



Lk
each stage. Note that the notation M is a large enough
yi iklk ≤ 1, i, i = 1, . . . , I|i = i ; k = 1, . . . , K;
constant value only used in disjunctive relationship con-
lk = 1
(3.4) straints. Equations (3.9–3.10) restrict the timing relation-
ship under the given processing route of each mining

Lk 
Lk job. Equations (3.11–3.12) declare the assignment and
yii klk + yi iklk ≤ 1, sequencing binary variables.
(3.5)
lk = 1 lk = 1
i, i = 1, . . . , I|i = i ; k = 1, . . . , K;
... Extended MPS models
yii klk + yi iklk ≥ xiklk + xi klk − 1, i, i = 1, . . . , I|i = i ; Moreover, we developed the following extended-version
k = 1, . . . , K; lk = 1, . . . , Lk ;
MPS models by adding or updating the parameters and
(3.6)
constraints in the basic MPS model step by step.
⎛ ⎞

Lk
Cik ≥ Ci k + pik + ⎝ yi iklk − 1⎠ M, E. Different ready times
(3.7)
lk = 1 Different ready times of mining jobs may be considered
i, i = 1, . . . , I|i = i ; k = 1, . . . , K;

in realistic mining scenarios. In a scenario, some min-
⎛ ⎞
Lk ing jobs are required to start in a certain period due to
Ci k ≥ Cik + pi k + ⎝ yii klk − 1⎠ M, the grade control or stockpile management requirements.
(3.8)
lk = 1 In another scenario, if two MPS horizons are merged,
 
i, i = 1, . . . , I|i = i ; k = 1, . . . , K; the ready times in the first-horizon mining jobs are dif-
Cik ≥ pik , i = 1, . . . , I; k = 1; (3.9) ferent from those in the second horizon. When differ-
ent ready times of mining jobs should be considered, the
Cik ≥ Ci,k−1 + pik , i = 1, . . . , I; k = 2, . . . , K; basic MPS model is extended by adding a new parameter
(3.10) ri and replacing Equation (3.9) with new Equation (3.13),
that is:
xiklk ∈ {0, 1}, New parameter:
i = 1, . . . , I; k = 1, . . . , K; lk = 1, . . . , Lk ;
(3.11) ri the specified ready time of mining job i.

New constraint:
yii klk ∈ {0, 1} , i, i = 1, . . . , I|i = i ;
(3.12)
k = 1, . . . , K; lk = 1, . . . , Lk ;
Cik ≥ ri + pik , i = 1, . . . , I; k = 1 . (3.13)

Here, i is the index of a mining job, i = 1, . . . , I; E. Different set-up and processing times
K is the number of operational stages; k is the index of In practice, the mining equipment unit needs to be
an operational stage, k = 1, . . . , K; Lk is the number of inspected or installed before the start of an operation. In
mining equipment units used at stage k; lk is the index of addition, the mining equipment units used at each stage
a mining equipment unit at stage k, lk = 1, . . . , Lk ; pik is may be non-identical with different operating capacities.
the processing time of mining job i at stage k as assuming In E2, different set-up times and processing times of dif-
that the mining equipment units at each stage are identi- ferent jobs at various operational stages are considered by
cal in the basic MPS model. There are three decision vari- adding the following new parameters and updating the
ables, namely, Cik that is the completion time of mining equations; the basic MPS model is extended by adding the
job i at stage k; xiklk that is the assignment binary vari- following new parameter and updating Equations (3.7–
able which equals one if the lk th equipment unit is allo- 3.10).
cated to mining job i at stage k; yii klk sequencing binary New parameters:
variable which equals one if job i precedes job i on the
lk th equipment unit at stage k. Equation (3.1) defines sk set-up time for a mining job at stage k.
the objective function of minimising the makespan, i.e. ik workload for a mining job i at stage k; for example,
the maximum completion time. Equations (3.2–3.6) sat- at drilling stage, the workload of a drilling equip-
isfy that the sequencing relationship between each pair ment unit is the total drilling length of this mining
of mining jobs on each assigned mining equipment unit job, measured in metres.
at each stage should be exclusive. Equations (3.7–3.8) θlk k operating capacity of a mining equipment unit lk at
define the disjunctive relationship between each pair of stage k. For example, at drilling stage, the operating
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS 61

capacity of a drilling equipment unit is measured in E. Side-cut requirements


metres/hour. In E4, the prefixed side-cut precedence relationships of
certain mining jobs on the same working bench are con-
Updated constraints: sidered by adding the following new parameters and con-
straints.
⎛ ⎞ New parameter:
ik ⎝ 
Lk
Cik ≥ Ci k + sk + + yi iklk − 1⎠ M, Si the subset of mining jobs that are side-cut predeces-
θlk k
lk = 1 sors of mining job i.
i, i = 1, . . . , I|i = i ; k = 1, . . . , K; lk = 1, . . . , Lk ;
(3.7b) New constraints:
⎛ ⎞
i k ⎝ 
Lk yi iklk ≥ xi klk + xiklk − 1,
Ci k ≥ Cik + sk + + yii klk − 1⎠ M, i = 1, . . . , I; i ∈ Si ; k = 1, . . . , K;
(3.16)
θlk k
lk = 1
xi klk = xiklk ,
i, i = 1, . . . , I|i = i ; k = 1, . . . , K; lk = 1, . . . , Lk ; (3.17)
i = 1, . . . , I; i ∈ Si ; k = 1, . . . , K;
(3.8b)
ik E. Diverse objective function
Cik ≥ ri + sk + , In E5, the following extended MPS models can evalu-
θlk k (3.9b)
i = 1, . . . , I; k = 1; lk = 1, . . . , Lk ; ate multiple objective functions to provide more analytic
ik decisions from different perspectives.
Cik ≥ Ci,k−1 + sk + , New parameters:
θlk k
i = 1, . . . , I; k = 2, . . . , K; lk = 1, . . . , Lk . di the due date of a mining job i.
(3.10b) wi delay cost rate if a mining job i is tardy.
ωklk fuel cost rate associated with the movement time
E. Sequence-dependent movement times of equipment unit lk at stage k.
In E3, sequence-dependent movement times of mining
New objective to minimise the number of tardy mining
equipment units are considered by adding the following
jobs:
new parameters and constraints.
New parameters: 
I
Minimise f (CiK , di ) | f (CiK , di ) = 1 if CiK > di ;
ηii distance between a pair of mining jobs i and i . i=1
v ii klk movement speed of a mining equipment unit lk else, f (CiK , di ) = 0. (3.18)
at stage k from a mining job u to mining job i in
sequence. New objective to minimise the total delay cost:

New constraints: 
I
Minimise wi Max {0, CiK − di } . (3.19)
i=1
ηii 
Lk
Ci k ≥ Cik + sk + yii klk + ik /θlk k New objective to minimise the total movement cost:
v ii klk
lk = 1
Lk
 
Lk
+ yii klk − 1 M, 
i, i = 1, . . . , I|i = i ; 
I 
I 
K 
Lk
lk = 1 yi iklk
 ηb b
lk = 1 Minimise .
v b bklk
k = 1, . . . , K; lk = 1, . . . , Lk ; i = 1|i=i i k = 1 lk = 1
(3.14) (3.20)

ηi i 
Lk
Cik ≥ Ci k + sk + yi iklk + i k /θlk k 2.4. Integrated MDP–MBS–MPS model
v i iklk
lk = 1 The motivation of proposing such an integrated MDP–
Lk
+ yi iklk − 1 M, 
i, i = 1, . . . , I|i = i ; MBS–MPS model is highlighted as follows. Based on
lk = 1 the feedback from our industrial partners, synchronisa-
k = 1, . . . , K; lk = 1, . . . , Lk . tion between determining the appropriate sizes of aggre-
(3.15) gates (mining jobs at the operational level) and allocating
62 S. Q. LIU AND E. KOZAN

the mining equipment units at each operational stage to Ruk the maximum allowable number of mining oper-
them is one of the critical decisions between mine plan- ations to be executed by equipment unit u at stage
ning and scheduling engineers. In execution, a tentative k.
multi-stage mine production timetable may be not grat- ξuk the ready time of equipment unit u at stage k.
ified for immediate use, because some inputs (e.g. the γ k workload of one block on bench γ at stage k.
number of blocks in each aggregate) need to be adjusted γ k the weighting factor of one block on bench γ at
in an interplayed way for several rounds between mine stage k according to its material property.
planning and scheduling engineers. In this scenario, to τuk the operating capacity of equipment unit u at stage
answer such a question: ‘how to achieve the appropriate k.
balance between the mining workload and the mining ςγ uk the set-up time of a mining operation on bench γ
resources?’, an innovative integrated MDP–MBS–MPS by equipment unit u at stage k.
model is developed. dγ the due date of the blocks on bench γ , which is
determined by the ending time of a time period in
New variables: MBS.

In this integrated model, the following decision variables ... An integrated model
are defined: Objective:
θγ the number of aggregated blocks of the θth min- 
Minimise (max{0, Cθγ K − dγ })θγ ,
ing job on bench γ . (4.1)
θ∈γ γ ∈ϒ
φθγ a binary variable that equals one if the θth min-
ing job on bench γ is created. subject to
xruk a binary variable that equals one if the rth mining
operation is processed by the equipment unit u at θγ γ k γ k
Qruk ≥ Qr−1,uk + ςγ uk +
stage k.  τuk 
+M zrukγ θ + yr−1, ukγ + xr−1,uk − 3 , (4.2)
yrukγ a binary variable that equals one if the rth mining
operation by the equipment unit u at stage k is ∀ r|r > 1, u, k, γ , θ;
performed on bench γ . θγ γ k γ k
Qr u k − ςγ u k − ≥ Cθγ k
zrukγ θ a binary variable that equals one if the rth mining  τu k 
+M zrukγ θ + zr u k γ θ + xr −1,u k − 3 , (4.3)
operation by the equipment unit u at stage k is
performed for the θth mining job on bench γ . ∀r, u, k, r |r > 1, u , k , γ , θ;
Cθγ k the completion time of the θth mining job on θγ γ k γ k
Qruk − ςγ uk − ≥ ξuk
bench γ at stage k; and Qruk is the completion τuk
+ M (zrukγ θ + yrukγ − 2), (4.4)
time of the rth mining operation by the equip-
ment unit u at stage k. ∀r = 1, u, k, γ , θ;
θγ γ k γ k
New parameters: Qr u k − ςγ u k − ≥ Cθγ k
 τu k 
+M zrukγ θ + zr u k γ θ − 2 , (4.5)
The list of new parameters is defined as follows for the ∀r, u, k, r = 1, u , k , γ , θ;
integrated model:
Qruk ≥ Cθγ k + M zrukγ θ − M, ∀r, u, k, γ , θ; (4.6)
K the set of operational stages, each of which is
indexed by k|k ∈ K. Qruk ≤ Cθγ k − M zrukγ θ + M ∀r, u, k, γ , θ; (4.7)
Uk the set of available equipment units at stage k,

indexed by u| u ∈ Uk . zrukγ θ = yrukγ , ∀r, u, k, γ ; (4.8)
ϒ the set of working benches in an operational hori- θ∈γ
zon, indexed by γ |γ ∈ ϒ. 
yrukγ = xruk , ∀r, u, k ; (4.9)
γ the minimum allowable number of blocks in a
γ ∈ϒ
mining job on bench γ . 
 θγ = γ , ∀γ ;
γ the maximum allowable number of blocks in a (4.10)
θ∈γ
mining job on bench γ .
γ the number of all available blocks on bench γ . xr+1,uk ≤ xruk , ∀r|r < Ruk , u, k; (4.11)
γ the maximum allowable number of mining jobs
on bench γ , indexed by θ|θ ∈ γ . φθγ ≤ φθ−1,γ , ∀γ , θ|θ > 1; (4.12)
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS 63

Table . (a) Data of mining jobs in an MPS case study.


Mining Tonnes Volume Surface Blocks Drilling Drilling Blasting Excavating
jobs (t) (m ) (m ) (units) metres (m) time (hours) time (hours) time (hours)

J , . , .   . . . .


J ,, . , . ,  . . . .
J ,, . , . ,  . . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  . . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  . . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  . . . .
J ,, . , . ,  ,. . . .
J ,, . , . ,  . . . .

3.1. Data analysis



θγ ≤ γ φθγ , ∀γ , θ; (4.13)
Due to confidentiality agreement, parts of the case study
data are given in Table 1 (a)–(c) and other parts of data
max(θγ − γ, 0) ≥ φθγ , ∀γ , θ; (4.14)
are briefly explained as follows. In this case study, 18 min-
 ing jobs will be scheduled in an expected 18-week MPS
zrukγ θ = φθγ , ∀k, γ , θ; (4.15) horizon. Each mining job will be processed consecutively
u∈Uk r∈Ruk
at three main stages, that is, drilling stage, blasting stage
φθγ , xruk , yrukγ , zrukγ θ ∈ {0, 1} ∀r, u, k, γ , θ. (4.16) and excavating stage. The critical resource at the drilling
stage is the drill equipment with two available units in this
Defined by Equation (4.1), the objective function aims case study. At the blasting stage, the critical resource is
to minimise the total weighted tardiness. Equations (4.1– the mobile processing unit with two units. At the exca-
4.7) define the complex timing relationship of the cre- vating stage, the excavators are the critical resources dur-
ated mining jobs. Equations (4.8–4.11) describe the com- ing the excavation process. In this case study, we first use
plex exclusive relationship in the creation of mining jobs. five excavators that have the same production rate of 2100
Equations (4.12–4.15) determine the size of the created tons per hour on average. The independent set-up time
mining jobs within the allowable range. Equation (4.16) of an excavator is about 1.5 hours on average. Note that
declares four binary variables. For more details on multi- all time elements are measured in hours. Associated with
resource multi-stage scheduling, please refer to Liu and each mining job, there are the specific (x, y, z) coordinate
Kozan (2009, 2011, 2012b, 2016b, 2017). values which are used to roughly determine the move-
ment distance between each pair of mining jobs at each
stage. Table 1(b) presents the operating capacity of each
3. Computational results mining equipment unit at each operational stage, respec-
All of the above proposed mathematical models were tively, measured in metres/hour, square metres/hour and
coded in Visual C# language and solved using IBM ILOG- cubic metres/hour at the drilling, blasting and excavat-
CPLEX 12.4 .NET package (IBM, 2012) which includes ing stages. Table 1(c) shows the release dates of each min-
dynamic link libraries and interfaces and thus is able to be ing equipment unit at each operational stage, measured
compiled under Microsoft Visual Studio 2012. Extensive in hours.
computational experiments were performed on a desktop
computer with Intel Core i7 CPU at 3.4 GHz and 8 GB
3.2. An initial solution
RAM under 64-bit Windows 7 operating system. In this
section, an insight-provoking case study based on real-life Extensive computational experiment with sensitivity
mining data is conducted to validate the proposed mod- analysis has been performed for identifying the key val-
els and illustrate how to significantly improve the perfor- ues, rationality and effects for the purpose of making
mance of open-pit mining process. quantitative decisions on how to improve the short-term
64 S. Q. LIU AND E. KOZAN

Table . (b) Operating capacities of mining equipment units.


Stage index Stage name No. of equipment units Measurement Unit  Unit  Unit  Unit 

 Drilling  metres/hour   NA NA
 Blasting  metres /hour   NA NA
 Excavating  metres /hour    NA

NA: this equipment unit is non-available in this operational horizon.

open-pit mine production efficiency in industrial prac- 3.3. An optimised solution


tice. After running the CPLEX optimiser to solve this
When the limit of CPLEX running time is set as 1800 s, a
MPS case with the time limit of 60 s, the timetable of an
high-quality MPS schedule in this case study is obtained
initial random solution is obtained and shown in Table 2.
and given in Table 3. In comparison to the initial
In the first column (Mining Jobs), the indexes of mining
schedule, the application of such a high-quality MPS
jobs are indicated. The second column (iThUnit) gives the
schedule can result in the efficiency improvement by
index of the resource unit (i.e. the drill equipment) used
(4279.6−2953.4)/4279.6 × 100 = 30.99%. In addition,
at the drilling stage for a mining job. The starting time
it implies that the proposed scheduling methodology is
(ETime), processing time (PTime) and completion time
promising to mining management staff in the determina-
(CTime) of a mining job at the drilling stage are shown
tion of the optimised timetable of short-term mining pro-
respectively in the third, the fourth and the fifth columns.
duction in a mathematical way instead of onerous man-
The other column names are kept in the same fashion for
ual input in practice at this mine site. More importantly,
the blasting and excavating stages. The maximum com-
with the application of the proposed MPS methodology,
pletion of this initial random solution is 4279.6 hours,
mining industry practitioners are able to ensure that the
which is highlighted in bold.
mining equipment units are always deployed in a more

Table . (c) Release dates of mining equipment units.


Stage index Stage name No. of equipment units Measurement Unit  Unit  Unit  Unit 

 Drilling  Hours   NA NA
 Blasting  Hours   NA NA
 Excavating  Hours    NA

Table . An initial solution of the MPS case study.


Drilling stage Blasting stage Excavating stage
ETime PTime CTime ETime PTime CTime ETime PTime CTime
Mining jobs iThUnit (hours) (hours) (hours) iThUnit (hours) (hours) (hours) iThUnit (hours) (hours) (hours)

J  . . .  . . .  . . .


J  . . .  . . .  . . 4279.6
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS 65

Table . A high-quality solution of the MPS case study.


Drilling stage Blasting stage Excavating stage
ETime PTime CTime ETime PTime CTime ETime PTime CTime
Mining jobs iThUnit (hours) (hours) (hours) iThUnit (hours) (hours) (hours) iThUnit (hours) (hours) (hours)

J  . . .  . . .  . . .


J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . 2953.4
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .
J  . . .  . . .  . . .

effective manner and rapidly respond to changes at the across an increasingly longer distance with steeper gradi-
mining sites. ent, resulting in considerable energy fuel consumption.
Thus, mining industry practitioners are keen to imple-
ment advanced engineering and management decisions
3.4. Bottleneck stage improvement (i.e. more efficient mine excavator/truck schedule) for the
reduction of energy intensity. In this case study, in com-
The proposed system is able to identify the bottleneck
parison to the initial random schedule, we analyse that
stage and quantitatively evaluate the usage of mining
the obtained optimal MPS schedule could bring around
resource units at this stage. The identification of bottle-
20 hours less of equipment movement time, which results
neck stage is a useful decision-making tool on whether
in the saving of over 5000 litres diesel in fuel consumption
the mining industry practitioners need to lease/assign
(Energy Efficiency Opportunities, 2013). Our proposed
additional resource units to improve the production effi-
MPS model is promising to provide analytical techniques
ciency or cut the redundant resource units at some non-
for improving the energy efficiency in synchronisation of
bottleneck stages to reduce the equipment costs. In this
mining operations.
case study, the bottleneck stage is identified as the exca-
vating stage. If the number of available excavators is
increased from 5 to 6, the makespan greatly declines 3.6. Integration benefits
from 2953.42 hours to 2586.4 hours, implying the effi-
The integration of MDP–MBS–MPS model has bene-
ciency improvement by (2953.4−2586.4)/2953.4 × 100 =
fits for determining the appropriate sizes of mining jobs,
12.43%. As a result, it is concluded that the proposed
accelerating the process of all operations and increasing
methodology can provide quantitative advices on how
the mining throughput. Our case study shows that a more
to deploy the mining equipment in a better way for
efficient schedule is obtained, resulting in a significant
significantly improving the short-term mine production
decrease of 205.67 hours on the makespan or the produc-
efficiency.
tion efficiency improvement by 6.96%. This is reflected by
such an interactive mechanism that can mathematically
bridge the gap between the planning and scheduling deci-
3.5. Fuel consumption reduction
sions. In practice, the decision-making process in plan-
In the Modular Report (2010), it is reported that ‘the aver- ning and scheduling should be interactive. When a tenta-
age fuel consumption rate of a 240-ton mine truck is 75 tive schedule under the current planning information is
litres diesel per hour under normal idling engine speed unsatisfied, the mine scheduling engineers may adjust the
and nearly doubles to 135 litres diesel per hours under resource capacities or production targets. However, such
excessive engine speed’. As the pit’s bench level becomes feedbacks between planners and schedulers might be
deeper and deeper, the materials will have to be hauled manually interplayed for several runs. In comparison, the
66 S. Q. LIU AND E. KOZAN

original production targets may be still achieved with less Notes on contributors
resource units or the overall mining throughput is fur- Shi Qiang (Samuel) Liu is currently a pro-
ther improved, because the sizes of mining jobs defined fessor in the School of Transportation &
as decision variables are able to be optimised after the Logistics at Southwest Jiaotong University
implementation of the proposed integrated planning– in China. He is also the Min Jiang Scholar
scheduling model. (the title of a distinguished Adjunct Pro-
fessor) in the School of Economics & Man-
agement at Fuzhou University in China.
4. Conclusion He was working in Australia for over 12
years (2005–2017) as a senior research
This paper represents a pioneering work to develop scientist in CRC ORE (Australian Government’s Cooperative
a whole mining management system by integrating Research Centre for Optimising Resource Extraction) and a
the strategic-level (long-term) MDP, tactical-level (mid- research fellow in Decision Science Discipline at Queensland
University of Technology, Brisbane, Australia. He was awarded
term) MBS and operational-level (short-term) MPS mod- the PhD degree (in Operations Research) with the Dean’s Award
els. With the analysis of many real-life mining constraints for Academic Excellence from the School of Mathematical Sci-
in a consecutive way, a series of CPLEX-solvable mathe- ences at Queensland University of Technology in 2009. Due
matical formulation models of MDP, MBS and MPS are to his academic contributions, he was awarded the New Out-
developed and integrated. Such integration was validated standing Researcher Medal by the Australian Society for Oper-
ations Research. He worked as a design engineer and soft-
by a real-world case study to bring the following practical
ware engineer in Singapore for three years (2002–2005). He
benefits to ore mining industry: obtained the Master degree in Industrial & Systems Engineer-
ing from National University of Singapore (2000–2002). He also
(1) it can provide analytical techniques on how to fur- obtained one Master degree (in Engineering Thermophysics)
ther improve the overall mining efficiency in a and two Bachelor degrees (in Thermal Engineering and Eco-
more systematic way; nomics) from Harbin Institute of Technology (one of top 9
universities in China) in six years (1993–1999). He has over
(2) interaction between mine planning and schedul- 40 refereed papers, most of which were published in lead-
ing engineers is optimised through a mathemati- ing SCI-indexed journals including Transportation Science (the
cal approach rather than by a traditional manual foremost journal in Transportation), Decision Support Systems,
way; International Journal of Production Economics, Expert Systems
(3) more synchronised mining schedules can be with Applications, Computers & Operations Research, Comput-
ers & Industrial Engineering, Journal of Heuristics, Engineer-
obtained to further improve the production
ing Optimisation, Journal of the Operational Research Society,
throughput, because the sizes of aggregates are Flexible Services & Manufacturing Journal, Optimisation Let-
defined as decision variables in an integrated way; ters, Advances in Engineering Software, Asia-Pacific Journal of
and Operational Research, etc. He has solid background in oper-
(4) the operational-level cost can be reduced due to ations research, decision science, combinatorial optimisation,
less idle time or less set-up time of mining equip- train scheduling, machine scheduling, mine scheduling, hospi-
tal scheduling, robotics scheduling, network flow algorithms,
ment units. and metaheuristics.
Erhan Kozan is an adjunct professor of
Regarding the future work, due to urgent demands, operations research in the School of Math-
some new mining jobs may be inserted into a partially- ematical Science, Queensland University
completed scheduling horizon at a specific time point. of Technology and an honorary professor
Thus, a reactive and robust rolling-horizon scheduling in the Sustainable Minerals Institute, Uni-
versity of Queensland, Australia. He has
methodology will be developed and included in a more had over 40 years industrial, managerial,
comprehensive system. teaching, and research experience in the
areas of operations research. He has acted
as a principal investigator for over 20 competitive national and
Acknowledgements international research grants since 1996 in the area of health,
finance, mining, car and truck production, railways, seaports
The authors would like to acknowledge the partial support of
transportation, logistics, and supply chain. He is the author of
CRC ORE, established and supported by the Australian Gov-
a book, ten software and over 200 refereed papers, most of
ernment’s Cooperative Research Centres Programme.
which were published in leading SCI-indexed journals. He is
the editor/associate editor of five journals and works as a ref-
eree of over 40 international journals. He has supervised over
Disclosure statement 35 postgraduate research students. He is the former president of
No potential conflict of interest was reported by the authors. the Asia Pacific Industrial Engineering and Management Soci-
ety (APIEMS) and Australian Society for Operations Research
INTERNATIONAL JOURNAL OF SYSTEMS SCIENCE: OPERATIONS & LOGISTICS 67

(ASOR). He is an expert in disciplinary research across decision Kozan, E., & Liu, S. Q. (2012). A demand-responsive decision
science and scheduling theory. His current research focuses on support system for coal transportation. Decision Support
the area of healthcare process optimisation, train scheduling, Systems, 54(1), 665–680.
and mine optimisation Kozan, E., & Liu, S. Q. (2016a). A new open-pit multi-stage
mine production timetabling model for drilling, blasting
and excavating operations. Mining Technology, 125(1), 47–
53.
ORCID
Kozan, E., & Liu, S. Q. (2016b). An operational-level multi-stage
Shi Qiang Liu http://orcid.org/0000-0002-2370-5519 mine production timetabling model for optimally synchro-
Erhan Kozan http://orcid.org/0000-0002-3208-702X nising drilling, blasting and excavating operations. Interna-
tional Journal of Mining, Reclamation and Environment, 1–
18. doi:10.1080/17480930.2016.1160818
Kozan, E., Liu, S. Q., & Wolff, R. (2013). A short-term produc-
References tion scheduling methodology for open-pit mines. In Inter-
national Symposium on the 36th Applications of Computers
Bley, A., Boland, N., Fricke, C., & Froyland, G. (2010). A
and Operations Research in the Mineral Industry (the 36th
strengthened formulation and cutting planes for the open
APCOM), Brazil (pp. 465–473). Porto Alegre, Brazil: Fun-
pit mine production scheduling problem. Computers &
dacao Luiz Englert.
Operations Research, 37, 1641–1647.
Kumral, M., & Dowd, P. A. (2005). A simulated annealing
Boland, N., Dumitrescu, I., Froyland, G., & Gleixner, A. M.
approach to mine production scheduling. Journal of the
(2009). LP-based disaggregation approaches to solving the
Operational Research Society, 56, 922–930.
open pit mining production scheduling problem with block
Lamghari, A., Dimitrakopoulos, R., & Ferland, J. A. (2015).
processing selectivity. Computers & Operations Research,
A hybrid method based on linear programming and vari-
36, 1064–1089.
able neighborhood descent for scheduling production in
Caccetta, L., & Giannini, L. M. (1988). An application of dis-
open-pit mines. Journal of Global Optimization, 63(3), 555–
crete mathematics in the design of an open pit mine. Dis-
582.
crete Applied Mathematics, 21, 1–19.
Lerchs, H., & Grossmann, I. F. (1965). Optimum design of
Caccetta, L., & Hill, S. P. (2003). An application of branch and
open-pit mines. Transactions on CIM, LXVIII, 17–24.
cut to open pit mine scheduling. Journal of Global Opti-
Liu, S. Q., & Kozan, E. (2009). Scheduling a flow shop with com-
mization, 27, 349–365.
bined buffer conditions. International Journal of Production
Chicoisne, R., Espinoza, D., Goycoolea, M., Moreno, E., &
Economics, 117(2), 371–380.
Rubio, E. (2012). A new algorithm for the open-pit
Liu, S. Q., & Kozan, E. (2011). Scheduling trains with priori-
mine production scheduling problem. Operations Research,
ties: a no-wait blocking parallel-machine job-shop schedul-
60(3), 517–528.
ing model. Transportation Science, 45(2), 175–198.
Cullenbine, C., Wood, R. K., & Newman, A. (2011). A sliding
Liu, S. Q., & Kozan, E. (2012a). An interactive planning
time window heuristic for open pit mine block sequencing.
and scheduling framework for optimising pits-to-crushers
Optimization Letters, 5(3), 365–377.
operations. Industrial Engineering and Management Sys-
Energy Efficiency Opportunities: Analyses of diesel use for
tems, 11(1), 94–102.
mine haul and transport operations. (2013). Department
Liu, S. Q., & Kozan, E. (2012b). A hybrid shifting bottleneck
of resources, energy and tourism, Australian government,
procedure algorithm for the parallel-machine job-shop
public report.
scheduling problem. Journal of the Operational Research
Espinoza, D., Goycoolea, M., Moreno, E., & Newman,
Society, 63, 168–182.
A. (2013). MineLib: a library of open pit mining
Liu, S. Q., & Kozan, E. (2016a). New graph-based algorithms
problems. Annals of Operations Research, 206(1), 93–
to efficiently solve large scale open pit mining optimisation
114.
problems. Expert Systems with Applications, 43, 59–65.
Ferland, J. A., Amaya, J., & Djuimo, M. S. (2007). Application
Liu, S. Q., & Kozan, E. (2016b). Parallel-identical-machine job-
of a particle swarm algorithm to the capacitated open pit
shop scheduling with different stage-dependent buffering
mining problem. In S. C. Mukhopadhyay & G. S. Gupta
requirements. Computers & Operations Research, 74, 31–
(Eds.), Autonomous Robots and Agents. Studies in Computa-
41.
tional Intelligence (Vol. 76, pp. 127–133). Berlin, Heidelberg:
Liu, S. Q., & Kozan, E. (2017). A hybrid metaheuristic algorithm
Springer.
to optimise a real-word robotic cell. Computers & Opera-
Hochbaum, D. S., & Chen, A. (2000). Performance analysis and
tions Research, 84, 188–194.
best implementations of old and new algorithms for the
Modular Report. (2010). Reduce haul truck fuel consumption
open-pit mining problem. Operations Research, 48(6), 894–
(Working Paper No.). Brazil: Modular Mining Systems.
914.
Mousavi, A., Kozan, E., & Liu, S. Q. (2016a). Open-pit block
IBM. (2012). IBM ILOG CPLEX optimization studio 12.4:
sequencing optimization : a mathematical model and solu-
optimization model development toolkit for mathematical
tion technique. Engineering Optimization, 48(11), 1932–
and constraint programming. Retrieved from https://www-
1950. doi:10.1080/0305215X.2016.1142080
01.ibm.com/software/websphere/products/optimization/
Mousavi, A., Kozan, E., & Liu, S. Q. (2016b). Comparative
cplex-studio-community-edition/.
analysis of three metaheuristics for short-term open pit
Kozan, E., & Liu, S. Q. (2011). Operations research for mining:
block sequencing. Journal of Heuristics, 22(3), 301–329.
a classification and literature review. ASOR Bulletin, 30(1),
doi:10.1007/s10732-016-9311-z
2–23.
68 S. Q. LIU AND E. KOZAN

Myburgh, C., & Deb, K. (2010). Evolutionary algorithms in Topal, E., & Ramazan, S. (2012). Strategic mine planning model
large-scale open pit mine scheduling. In the Proceedings of using network flow model and real case application. Inter-
the 12th annual conference on Genetic and evolutionary national Journal of Mining, Reclamation and Environment,
computation (pp. 1155–1162), Portland, Oregon, USA. 26, 29–37.
Newman, A. M., Rubio, E., & Weintraub, R. C. A. (2010). A Underwood, R., & Tolwinski, B. (1998). A mathematical pro-
review of operations research in mine planning. Interfaces, gramming viewpoint for solving the ultimate pit prob-
40(3), 222–245. lem. European Journal of Operational Research, 107(1),
Pan, C. H. (1997). A study of integer programming formula- 96–107.
tions for scheduling problems. International Journal of Sys- Wetherelt, A., & Van Der Wielen, K. P. (2011). Introduction to
tems Science, 28(1), 33–41. open-pit mining. In P. Darling (Ed.), SME mining engineer-
Ramazan, S. (2007). The new fundamental tree algorithm for ing handbook 3rd version (pp. 857–876). Englewood, CO:
production scheduling of open pit mines. European Journal Society for Mining, Metallurgy, and Exploration.
of Operational Research, 177, 1153–1166.

View publication stats

You might also like