Optimization and Engineering (2023) 24:681–707

https://doi.org/10.1007/s11081-021-09701-4

RESEARCH ARTICLE

Short‑term open‑pit production scheduling optimizing

multiple objectives accounting for shovel allocation
in stockpiles

Fabián Manríquez1 · Héctor González2 · Nelson Morales3

Received: 22 April 2021 / Revised: 8 November 2021 / Accepted: 8 November 2021 /

Published online: 4 January 2022
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature
2021

Abstract
Short-term open-pit mine production scheduling is a challenging task that must deal
with several objectives, like maximization of the productivity of equipment (plant
and mining), compliance of ore extraction, and others. Unfortunately, there are
trade-offs between these objectives that make the problem of finding well-balanced
short-term schedules complex to handle. To overcome this problem, we propose a
methodology based on mathematical programming and a hierarchical method to
generate short-term open-pit schedules considering multiple objectives. The math-
ematical program allocates shovels to different mining faces, including stockpiles.
It considers plant capacity constraints, ore blending, precedences between mining
faces, shovels throughput, and shovels’ traveling time between mining faces. We
also propose several compliance indicators, which we use to evaluate and compare
different short-term schedules. We apply the proposed optimization model to a real
iron open-pit mine to compare how the hierarchical method performs with regards
to a single objective approach and show increases of waste extraction compliance up
to 29% (when priority is production) and of 73% in production (when the priority
is waste extraction). Moreover, in general, we observe that the hierarchical method
produces more robust plans.

Keywords Mining · Short-term planning · Equipment allocation · Multiobjective ·

Optimization

The authors thank CONICYT grant PFCHA/DOCTORADO BECAS CHILE/2019 - 21190201 and
CONICYT/PIA Project AFB180004 grant “Advanced Mining Technology Center”, which funded
this research.

* Nelson Morales
[email protected]
Extended author information available on the last page of the article

13
Vol.:(0123456789)
682 F. Manríquez et al.

1 Introduction

Open-pit mining is a technique for the extraction of ore deposits near the Earth’s
surface that permits access to valuable sections of the deposit with relative ease,
therefore reaching high production capacities and lower cost compared to under-
ground mines. However, because the ore distribution is not uniform in the deposit,
the excavation process may provide access to parts with various ore content levels
(thus, economic value) or even require removing unwanted material (“waste”).
Therefore, the order in which different parts of the deposit are accessed has an
enormous impact on a mine operation’s economic value.
Because of its complexity, the process of planning an open-pit mine is organ-
ized into different levels: strategic (long-term), tactical (medium-term), and oper-
ational (short-term) levels (L’Heureux et al. 2013). These levels of planning are
oriented toward different purposes and designed to work with various levels of
information.
Strategic mine planning seeks to maximize the mine operation’s economic
value, usually associated with the Net Present Value (NPV). For this, it deter-
mines what portions of the ore body should be extracted, when this should
happen, the total lifespan of the mine, the production rate, and the investment
amount. A principal result of a long-term planning is a production schedule,
determining the tonnage of waste and ore to be mined from each bench-phase for
each year over the scheduling horizon.
Tactical scheduling determines the mining sequence for up to a typical period
of 5 years based on the production rate constraints. It takes as input the produc-
tion schedule determined in the long-term plan and defines what portions will be
extracted to meet the production targets with the available resources so that the
costs of this extraction are adequately projected and minimized.
Lastly, operational scheduling covers time horizons that may span from a week
and up to a year. Within this context, short-term planning seeks to ensure the fea-
sibility of the long and medium-term mine production schedules (Smith 1998) by
delivering the ore tonnes and grades to the processing plant (Chanda 1990).
To reach long-term goals, short-term planners look to make the best possible
use of all available resources: processing facilities, mining zones, stockpiles, and
equipment fleet. Unfortunately, there may exist some trade-offs between all these
aspects. For example, it may be possible to maximize the productivity of loading
equipment by assigning them to mining faces closer to their destination. How-
ever, as it turns out, these faces tend to be zones of waste because they are nearer
to the surface. Similarly, a mining operation may need blending material from
different areas of the mine to obtain certain product quality, thus relocating load-
ing equipment and decreasing its productivity. All this implies that considering
only a single criterion for optimization may harm other aspects which are relevant
too.
Another relevant aspect of short-term planning is related to stockpiles. Stock-
piles are essential because they allow keeping the balance between the ore
extracted from the mine and the processing capacity: First, stockpiles act as

13
Short‑term open‑pit production scheduling optimizing multiple… 683

buffers so that processes before them and processes after them can operate with-
out being constrained by each other. Second, stockpiles can be used to control the
blending of material to be processed, reducing its variability. Finally, stockpiles
can be used to sort material by grade or other properties (Jupp et al. 2014; Robin-
son 2004).
All the issues mentioned above make the problem of short-term mine planning
hard to be modeled. In fact, contrary to what happens in long-term and medium-
term planning, the short-term planning problem for open-pit mines has not been as
widely studied (Blom et al. 2018); therefore, there is a lack of mathematical models
and optimization tools to support short-term decisions.
Because of the above, we propose a Mixed-Integer Linear Programming (MILP)
problem to generate short-term open-pit schedules. The optimization model’s major
decision is the assignment, over several time-periods, of shovels to mining faces and
stockpiles but accounting for the time required to move from one mining face to
another. This integration allows the model to link and balance several of the objec-
tives mentioned before.
The model takes as input production targets (i.e., material to be sent to the mill),
mining targets, and waste extraction targets, and looks for a shovel assignment to
achieve these targets while also considering the following constraints: plant capac-
ity, ore blending requirements, precedences between mining faces, shovels through-
put, mine sequencing and movement of shovels between sectors of the mine.
We utilize the model in two configurations: the single objective approach, which
aims to optimize a relevant indicator, and the hierarchical method (Grodzevich and
Romanko 2006) that utilize priorities to sequentially addresses different objectives.
This work’s principal contribution is the MILP to generate a short-term schedule
for open-pit mines considering loading equipment allocation, stockpiles, and three
objectives for optimization: loading equipment productivity, plant productivity, and
grade compliance. However, the paper also contributes with several indicators to
assess and compare the different short-term schedules. Finally, the article presents
the model’s application to an iron open-pit mine case study showing the model’s
validity in different scenarios.
The remainder of this paper is organized as follows. Section 2 provides a review
of the related work associated with short-term open-pit scheduling. Section 3 states
the problem being solved by the optimization model. Section 4 presents the pro-
posed optimization model. Section 5 describes the real-scale open-pit mine case
study and outlines the scenarios that are analysed. Section 6 reports and discusses
the results of the case study. Finally, Section 7 concludes the study and outlines
future work.

2 Related work

This section briefly reviews some studies associated with production scheduling at
the long- and short-term horizons. We pay special attention to optimization models
considering stockpiles and multiple-objectives, which are the core elements of the
model proposed in this work.

13
684 F. Manríquez et al.

Most optimization models used to support open-pit planning rely on a discretiza-

tion of the deposit called the block model. A block model is a 3D array of blocks
with spatial coordinates and a vector of relevant attributes (tonnage, rock type, ore
grades, and others) estimated using samples of the deposit, geologic models, and
geostatistical methods. Therefore, optimization models abstract the scheduling pro-
cess, computing for each block, its extraction period, and its best destination for
processing, stocking, or dumping. The primary constraint in open-pit mining is that
extraction must happen so that the pit walls do not collapse. This is modelled using
precedence constraints: For each block, a specific set of blocks (called predeces-
sors) must be extracted to gain access to that block while keeping the walls of the pit
stable.
The open-pit mine production scheduling problem (OPMPSP) consists of sched-
uling the blocks’ extraction to maximize the Net Present Value (NPV) (Samavati
et al. 2018), subject to precedence and capacity constraints. The first formulation of
OPMPSP as an integer linear program is due to Johnson (Johnson 1968). Over the
years, several authors have extended this model. The Precedence Constrained Pro-
duction Scheduling Problem (PCPSP) (Espinoza et al. 2013) is currently regarded
as the reference problem for modeling long-term open-pit mining. This model con-
siders several time-periods and multiple potential destinations of the blocks and
slope constraints; however, it also considers an arbitrary number of side constraints,
including capacity and blending restrictions. Unfortunately, neither OPMPSP nor
PCPSP consider stocks in their modeling, except as potential block destinations
from where no material is retrieved during the planning period.
Indeed, stockpiles are hard to model using linear models because of the mixing
of material produced in these piles. As Moreno et al. (2015) states, the inclusion
of stockpiles in models for the open-pit production-scheduling problem has been
avoided, due to the difficulty of correctly modeling the mixing behavior of the mate-
rial inside a stockpile. Nevertheless, some authors have proposed various approaches
to address this issue.
A few nonlinear optimization models have been proposed to address the open-
pit mine production scheduling problem with stockpiles. Tabesh et al. (2015) states
that stockpiling should theoretically be modeled nonlinearly to optimize a compre-
hensive open-pit mine plan, and linearized the formulation by using a “sufficient
number" of stockpiles, each with a tight range of grades. Bley et al. (2009) proposed
a quadratic model for production scheduling that assumes ‘instant-mixing” inside
the stockpile, that is, that all grades of material inside the stockpile are averaged.
However, this type of model is computationally very difficult to solve, limiting its
use in real instances. Bley et al. (2012b) provides more details on computational
approaches for solving these models. Bley et al. (2012a) address the solution of the
open-pit mine production scheduling problem (OPMPSP) with a single stockpile
(OPMPSP+S). The addition of a stockpile adds a relatively small number of quad-
ratic constraints to the formulation of the OPMPSP and turns the problem from a
mixed-integer linear into a mixed-integer nonlinear program.
Hoerger et al. (1999) assume that material sent to a stockpile must have a grade
within a specific range and that the grade of material extracted from a stockpile is
the minimum value of that range. Akaike and Dagdelen (1999) consider that there

13
Short‑term open‑pit production scheduling optimizing multiple… 685

are infinite potential stockpiles, so every block has its stockpile (i.e., there is no
blending in the stockpile). Fu et al. (2019) models the stockpiles using a series of
grade bins, allowing the model to allocate material with a different grade. The mod-
els of Moreno et al. (2017), Rezakhah et al. (2020), Rezakhah and Newman (2020)
utilize blending constraints to require a constant grade in the stockpiles by forcing
that the average grade sent to each stockpile is known and fixed in advance. Moreno
et al. (2017) proposed several linear integer optimization problems to schedule
open-pit mines considering stockpiling. Rezakhah et al. (2020) apply those models
to an operational poly-metallic (gold and copper) mine, where stockpiles are used
to blend materials based on multiple block characteristics. Rezakhah and Newman
(2020) extend the model to consider the deterioration of material when exposed to
the environment.
Asad (2005) presents a long-term cut-off grade optimization algorithm for open-
pit mining operations with stockpiling in a deposit with two economic minerals.
This algorithm is an extension of the theory of cut-off grades in deposits of two eco-
nomic minerals presented in Lane et al. (1984).
Finally, some articles account for stockpiles in the context of geological uncer-
tainty in the metal content, i.e., they incorporate the fact that geostatistical methods
produce only estimations of critical values like grades Ramazan and Dimitrakopou-
los (2013), Koushavand et al. (2014), Silva et al. (2015), Lamghari and Dimitrako-
poulos (2016), Levinson and Dimitrakopoulos (2019).
The works mentioned above addressed block scheduling in a long-term setting;
therefore, they are not concerned with operational aspects like equipment assign-
ment, which is central to the problem we address. We now review some works ori-
ented to the short-term.
Fioroni et al. (2008) seeks to reduce mine costs using optimization and simula-
tion models to generate short-term production schedules that can be executed in the
mine, considering the actual use of mining equipment.
Rehman and Asad (2010) propose a MILP model to define the short-term
sequence mining of blocks of a limestone quarry to meet plant quantity and quality
requirements at the lowest possible cost. Contrary to our approach, their model does
not consider shovel allocation to quarry blocks.
Eivazy and Askari-Nasab (2012) develop a MILP model to generate short-term
schedules. The model minimizes the total cost, considering: mining cost, process-
ing cost, waste rehabilitation cost, rehandling cost and total haulage cost. The model
takes into account multiple destinations and models: stockpiles as buffers and blend-
ing location, horizontal directional mining, and decisions on-ramps, blending con-
straints at the processing plants, mining and processing capacities, and mining prec-
edences. The model assumes that stockpiles’ output grade is equal to their average
grade; however, it does not consider the allocation of loading equipment in the mine
or to stockpiles.
Torkamani and Askari-Nassab (2015) integrates the optimal scheduling of short-
term production with the simulation of the shovel-truck operation; based on this, it
is possible to choose the optimal number of shovels and trucks needed to fulfill the
mine schedule.

13
686 F. Manríquez et al.

Upadhyay and Askari-Nasab (2016, 2018, 2019) present an MILP problem to

allocate shovels to mine faces to maximize production, achieve the desired head
grade and tonnage at crushers, and minimize shovel movements.
Mousavi et al. (2016) propose a MILP problem that considers precedences,
machine capacity, grade requirements, processing demands, and stockpile man-
agement. The objective function is to minimize the total cost, including the cost
of rehandling, holding, misclassification, and drop-cut. Their model considers the
assignment of loading equipment in the mine but not to stockpiles.
Matamoros and Dimitrakopoulos (2016) propose a formulation based on sto-
chastic mixed-integer programming to address scheduling of open-pit mines in
the short-term. The model considers uncertainty in both ore body metal quantity
and quality. It also takes into account fleet parameters and equipment availabil-
ity. The model allocates shovels to mine sectors and the number of truck trips
per shovel. The objective function considers operating fleet cost and mining cost.
Stocks are modeled implicitly by considering a penalization cost due to overpro-
duction at each time-period.
Blom et al. (2017) apply the hierarchical method to generate multiple, diverse
short-term schedules while optimizing for a customizable, prioritized sequence of
objectives. They use a rolling horizon-based algorithm to resolve instances.
Upadhyay and Askari-Nasab (2019) uses a mixed-integer linear objectives pro-
gramming model (MILGP) to obtain the optimal allocation of shovels and trucks
to meet the objectives aligned to long-term scheduling: maximize production,
minimize deviations of the head grade, minimize deviations of the plant feed ton-
nage, and minimize operating costs.
Alexandre et al. (2019) deals with the truck dispatching problem. That is, the
efficient allocation of trucks to shovels in operation at open-pit mines. The work
present multi-objective strategies for solving the problem of dynamically allocat-
ing a heterogeneous fleet of trucks in an open-pit mining operation.
Both and Dimitrakopoulos (2020) presents a stochastic optimization model
that optimizes the short-term extraction sequence of an open-pit mine allocating
shovels and trucks to mine faces. The model considers geological and equipment
performance uncertainty.
Shah and Rehman (2020) describes a mixed integer linear programming model
that optimizes the short-term extraction sequence of an cement quarry allocating
shovels and trucks to mine faces. The objective function minimizes the truck and
shovel total cost, subject to quantity and quality constraints.
Benlaajili et al. (2020) presents a tuck-shovel dispatching model that allocates
shovels to mine faces and shovels to shovels in two main steps. The first step
proposes a modeling of the allocation of shovels problem as a vehicle routing
problem. In the second step, a mixed integer linear programming model is pro-
posed to determine the optimal number of trips required to transport the quantity
of ore from each loading point to each dumping site, this model is used to dis-
patch available trucks to the appropriate shovel.
Table 1 summarizes and compares the characteristics of the short-term open-
pit mine scheduling articles that we have reviewed so far.

13
 Table 1  Comparison of short-term open-pit mine scheduling articles
Feature / Fioroni Torkamani Eivazy and Mousavi Matamoros Blom Upadhyay Alexan- Both and Shah and Benlaajili
Article et al. and Askari- Askari-Nasab et al. and Dimi- et al. and Askari- dre et al. Dimitrako- Rehman et al.
(2008) Nasab (2011) (2012) (2016) trakopoulos (2017) Nasab (2019) (2019) poulos (2020) (2020) (2020)
(2016)

Shovel alloca- ✓ ✓ ✗ ✓ ✓ ✓ ✓ ✗ ✓ ✓ ✓
tion in mine
faces
Shovel ✗ ✗ ✗ ✗ ✗ ✓ ✗ ✗ ✗ ✗ ✗
allocation in
stockpiles
Explicit ✗ ✗ ✓ ✓ ✗ ✓ ✗ ✗ ✓ ✗ ✗
modeling of
stockpiles
Truck alloca- ✓ ✓ ✗ ✗ ✓ ✓ ✓ ✓ ✗ ✓ ✓
tion in mine
Short‑term open‑pit production scheduling optimizing multiple…

faces
Grade uncer- ✗ ✗ ✗ ✗ ✓ ✗ ✗ ✗ ✓ ✗ ✗
tainty
Equipment ✗ ✗ ✗ ✗ ✓ ✗ ✗ ✗ ✓ ✗ ✗
perfor-
mance
uncertainty
Multi-objec- ✓ ✗ ✗ ✗ ✓ ✓ ✓ ✓ ✓ ✗ ✓
tive optimi-
zation (not
only costs)
687

13
688 F. Manríquez et al.

With regard to works in Table 1 that consider multiples objectives, Fioroni et al.
(2008) considers the following objectives: maximizing ore production, minimiz-
ing grade deviation to ore plant, and minimizing the loss of production due to load
equipment movements. Torkamani and Askari-Nasab (2011) minimizes the opera-
tional cost associated with the mine considering: cost of moving shovels to new
faces, total transportation cost of trucks moving to the waste dump or to the mill, and
cost of negative deviation from the production target at the mill. Eivazy and Askari-
Nasab (2012) also minimizes the total cost, considering: mining cost, processing
cost, waste rehabilitation cost, rehandling cost and total haulage cost. Mousavi et al.
(2016) minimizes the total cost, which includes rehandling and holding costs, mis-
classification and drop-cut costs. The misclassification cost is monitored to ensure
that material is assigned to the right destination. A drop-cut is a condition such that
a block is extracted while all the adjacent blocks have not yet been extracted. In
contrast to a drop-cut, a side-cut is performed when the excavator is located in the
same bench as the block. Finally, a drop-cut cost is considered in order to give prior-
ity to the side-cut extraction, unless a new working bench is required to be opened.
The optimization problem of Matamoros and Dimitrakopoulos (2016) minimizes
the overall extraction cost. This cost considers the following components: cost of
extracting material from the mine, hauling cost given the uncertainty in the trucks’
hauling time and mechanical availability, cost of shovel movements among sectors,
lack of production per shovel given uncertainty in its mechanical availability, penal-
ize the lack of mining blocks that match the required mining width and minimize
the geological risk with respect to the quality and quantity of ore production and
penalize deviation from production targets. Blom et al. (2017)’s formulation con-
siders multiple objectives such as maintaining alignment with a longer-term plan
and maintaining product grades within desired bounds. The optimization problem
described by Upadhyay and Askari-Nasab (2019) considers the following objectives:
maximization of production, meeting the desired feed to processing plants, meeting
the grade blending requirements of the processing plants, and minimizing shovel
movements. The objective function for the truck allocation described by Alexandre
et al. (2019) is to maximize production and minimize costs. The optimization model
proposed by Both and Dimitrakopoulos (2020) maximizes metal production and
profit of the mining complex as a whole, instead of minimizing operational costs.
It considers: revenue and costs in the mining complex, penalties for deviations from
production targets, reduce the risk of not achieving shovel production targets per
mining area, reduce risk of falling short of truck haulage capacity, shovel movement
cost, account for the cost of trucks in operation, and smoothing of the mining sched-
ule. For this part, the objective function of the optimization problem described in
Shah and Rehman (2020) considers the minimization of truck/shovel cost. Finally,
Benlaajili et al. (2020) addresses two problems. The first problem (the allocation
of shovel to mining faces) minimizes the total travel cost of shovels. The second
problem (allocation of trucks to shovels) minimizes the number of trips required to
transport ore during a working shift.
From Table 1 we observe that the only two works that consider uncertainty
are Matamoros and Dimitrakopoulos (2016) and Both and Dimitrakopoulos
(2020). These articles take into account geological uncertainty and the equipment

13
Short‑term open‑pit production scheduling optimizing multiple… 689

performance uncertainty. We understand that addressing the mentioned uncertain-

ties is critical to risk assessment and decision making in short-term open-pit mine
production scheduling; however, we consider stochastic modeling out of the scope
of this study and a potential improvement for future research.

2.1 Multi‑objective optimization

There are several different multi-objective evolutionary algorithms like the genetic
algorithm which fundamentally operates on a set of candidate solutions. The
weighted sum and hierarchical method are commonly used to optimize multiple
objectives.
In the weighted sum method, each different objective is assigned a positive coef-
ficient that represents its relative importance. Then, a unified objective function
is constructed as the addition of the objectives multiplied by their corresponding
weights.
In the hierarchical method, the decision-maker ranks the objective functions,
sorting the objectives in descending order of importance. In this method, as many
optimization problems as objective functions are solved sequentially, following the
rank previously defined. Each problem is then solved for its corresponding objec-
tive, but with an additional constraint requiring it to perform as well as the optimal
solutions of the criteria already considered.
Figure 1 provides a schematic view of this procedure for N objective functions
to be minimized in descending order of importance. The j-th objective function is
denoted as Aj , and Aj as the optimal value of this function in the j-th problem. The
set of general constraints of the problem are represented as x ∈ S , where x is the vec-
tor of variables and S the set of feasible points. The parameters 𝜆i ≥ 1, ∀i ∈ {1, ..., N}
represent the tolerance of the deviation of the value Ai . Notice that if 𝜆i = 1, it means
that the i + 1-th sub-problem must find a solution that replicates the i-th optimum
value obtained for the previous sub-problem, while 𝜆 > 1 implies that some deterio-
ration of the i-th objective value is allowed.
At the first iteration, the first problem is solved and so a solution x1 with value
A1 = A1 (x1 ) is found. Then, for the j-th iteration, solutions xk , k = 1, … , j − 1 have
been found, each with a value Ak = Ak (xk ), hence the constraints Ak (x) ≤ 𝜆k Ak are
added for k = 1, … , j − 1 and the problem with objective function Aj is solved. The

Fig. 1  Scheme of the hierarchical method considering N objectives

13
690 F. Manríquez et al.

process continues until the last problem, with j = N is solved and its solution xN and
its value AN (xN ) are reported.
The hierarchical method seems more suited for short-term scheduling and there-
fore is the one utilized in this article. In general, the articles that consider multi-
ple objectives described in Table 1 use the weighted sum method to perform multi-
objective optimization. Applying the weighted sum method is straightforward and
easier to understand than the hierarchical method. However, it requires selecting the
weight coefficients, which makes its application difficult (Grodzevich and Romanko
2006). An advantage to use the hierarchical method to perform a multi-objective
optimization is that we do not have to compute or select the value of the weight
coefficients. As a drawback, we need to solve n optimization problem to perform a
multi-objective optimization with n hierarchical objectives.

3 Problem statement

This section provides an overview of all the modeling aspects incorporated in the
mathematical model. The precise optimization problem, notation and parameters are
introduced in Sect. 4.
Each mining face is characterized by a tonnage, type of material (ore or waste),
and the grade of the metal of interest. We assume two kinds of faces: mine or stock-
piles. Mining faces only allow extraction, while stockpile faces also allow the accu-
mulation of material extracted at the mine. Mining faces are also related to each
other by precedences, meaning that some need to be depleted before others’ extrac-
tion begins.
The material at the mining faces may contain different components (for example,
different metals or pollutants). We will assume that this composition is known and
there are targets or ranges allowed at the plant, for each period and component.
In short-term planning, the traveling time of shovels can be a significant portion
of the total time; thus, we consider the travel time required to move a shovel between
two different faces as not available for production.
In our approach, stockpiles are potential destinations of material extracted at
the mine and locations to which shovels can be assigned. In terms of the grades,
we adopt the same model as (Hoerger et al. 1999) and assume that the stockpiles
are homogeneous and the ore reclaimed from each stockpile has a grade equiva-
lent to the average grade of the stockpile, which is constant over the scheduled hori-
zon. Moreover, we assume that the stockpile to which the face sends its material
is defined in advance, i.e., the model decides the amount to be sent but not what
stockpile to use.
We consider that a shovel is unavailable during meals, shift changes, mainte-
nance, and failures and assume that the total unavailable time per period is known in
advance for every shovel; thus, the model can use a shovel only during its available
time. Notice that the model does not utilize the internal distribution of unavailable
time explicitly. In particular, no maintenance or failures are incorporated directly
and the model can use only available time.

13
Short‑term open‑pit production scheduling optimizing multiple… 691

The available time of a shovel during a period can be split as follows:

Available Time = Traveling Time + Operation Time + StandBy Time . Traveling
occurs when the shovel moves from one mining face to another during a certain
planning period. Operational time happens when the shovel is loading material
and sending it to its destination. Stand-By time occurs only in the case when the
face is depleted, that is if the operational time is more than enough to load all the
material.
To compute the material loaded, we consider that shovels have a known opera-
tive throughput that quantifies the material that they can load per unit of time
(tonnes per hour). We assume that this parameter depends on the shovel and the
mining face. This is to allow the model to consider different shovels, but also dif-
ferent conditions that may effect the performance of a shovel, like limited space
for operating the equipment, delays due to long hauling distances, and others.
We assume that the time (in hours) that is required by a given shovel to move
between two mining faces is also known in advance. However, as a shovel may visit
more than one face per time-period, we utilize routes. A route is a list of sectors in
which a shovel travels sequentially within a time-slot. Thus, the traveling time is
computed as the addition of all movements between mining faces in the route.
We propose several indicators to compare and evaluate the performance of the
results beyond the optimization targets. These indicators aim to measure the com-
pliance of the plan, i.e., how close the material flows scheduled by the plan are
to their targets. The indicators are based on the material flows depicted in Fig 2:
O, the tonnage of ore sent directly from the mine to the processing facilities; R
the tonnage of rehandling, that is, material sent from stockpiles to the processing
facilities; S, the tonnage of material sent from the mine to stockpiles; and W, the
tonnage of material sent to waste dumps.
Two critical material flows set by long-term scheduling are: P = R + O , the
total material sent to the plant, and M = S + O + W , the total material extracted
from the mine. Therefore, we consider total targets P0 and M 0 , respectively, to be
moved during the planning horizon. We consider also W 0 , a waste removal target.
It is worth noting that, W 0 is not usually considered in long-term planning, where
all ore extraction occurs during the length of time being scheduled; however, in
short-term planning it is a critical target, because it enables reaching ore in peri-
ods beyond the planning horizon.
Based on the material flow targets described before, we introduce compliance
performance indicators, which are presented in Table 2: waste extraction, plant

Fig. 2  Conceptual diagram of

a material’s flow network in a
mining operation

13
692 F. Manríquez et al.

Table 2  Compliance Index Formula Description

Performance indicators
𝛤W W
⋅ 100 Waste extraction compliance.
W0
𝛤P O+R
⋅ 100 Plant utilization.
P0
𝛤O O
⋅ 100 Plant utilization due to ore
P0 directly sent from mine.
𝛤M O+S
⋅ 100 Ore extraction compliance
M0

utilization, plant utilization due to ore extracted from the mine, and ore extrac-
tion. These indicators will be later used to analyze the plans obtained using the
model.

3.1 Multiple criteria for measuring short‑term planning performance

We consider several potential criteria (and therefore objective functions) to be mini-

mized using the model. In this article, we focus on the first three of the list below,
although the model could be used also to analyze the others:

• 𝛥O is the deviation (in tonnage) between ore sent to the plant from the mine and
the plant capacity.
• 𝛥P is the deviation (in tonnage) between ore sent to the plant from the mine and
stockpiles and the plant capacity.
• 𝛥W is the deviation (in tonnage) between waste hauled and its target.
• 𝛥D is the deviation (in tonnage) between ore sent to the plant and the ore plant
capacity per period.
• 𝛥Gradej , the deviation (in tonnage) between the content of component j sent to
the plant and its target.
• Total Travel Time, the total traveling time of the shovel fleet (in hours).
• Total Travel Cost, the total traveling cost of the shovel fleet (in dollars).

4 Optimization model

This section presents in detail the mathematical model that implements the prob-
lem described in the previous section as a mixed integer linear program. The main
decision made by the model is the allocation of loading equipment to mining faces,
which is done considering the fraction of a time-slot that the shovel will be assigned
and the fraction of time that it will be working in a mining face.

13
Short‑term open‑pit production scheduling optimizing multiple… 693

4.1 Sets and Indexes

• p ∈ P , the set and index for shovels.

• f , f � ∈ F , the set and indexes for mining faces.
• Fore , Fwaste ⊂ F , sets of mining faces that contain ore and waste, respectively.
• Fsp ⊂ F set of mining faces that correspond to stockpiles. (Notice that
Fsp ⊂ Fore, because all stockpiles contain ore).
• t ∈ T = {1, 2, … , T}, the set and index for time periods. T is the time horizon.
• j ∈ J , the set and index for material components to be controlled (like grades,
pollutants or others).
• r ∈ R , the set of routes. A route r is a ordered sequence of mining faces
r = (f1 , f2 , … , fk ), for some k, which corresponds to the length of the route,
denoted as |r|.
• Rf , the set of routes that go through mining face f.
• Sf ⊂ F, f ∈ Fsp is the set of faces that send material to stockpile f.
• Hr , the set of routes whose last face is equal to the first face of route r.
• (f , f � ) ∈ Q, the set of precedences between the mining faces ( f ′ has to be mined
before f).

4.2 Parameters

The parameters of the optimization model are detailed below.

• Tonf ≥ 0, the total material to be mined in mining face f (in tonnes).

• Gradej,f ∈ [0, 1], the fraction of the content j in the mining face f.
• Lengtht > 0, the length of period t (in hours).
• Avl Timep,t , hours that shovel p is available during period t. This time is com-
puted considering the expected maintenance time Maint Timep,t , failure time
Fail Timep,t , and the delays Delay Timep,t as follows:
AvlTimep,t = Lengtht − (Maint Timep,t + Fail Timep,t + Delay Timep,t ).

• Prodp,f , the tonnage of material that can be mined by the shovel p in the mining
face f. If 𝜃p,f is the estimated throughput of shovel p at face f (in tonnes per hour)
then
Prodp,f = Avl Timep,t ⋅ 𝜃p,f .

• Capt > 0, the processing capacity of the plant in period t (in tonnes).
• Min Prodt > 0, the minimum desired tonnage to be sent to the ore processing
plant in period t (in tonnes).
• Max Gradej ≥ 0, j ∈ J , the maximum grade of the content j to be sent to the ore
processing plant in period t.
• Min Gradej ≥ 0, j ∈ J , the minimum grade of content j to be sent to the ore pro-
cessing plant in period t.
• Travel Timep,r > 0, p ∈ P, r ∈ R is traveling time of shovel p along the sectors
of route r (in hours).

13
694 F. Manríquez et al.

• Max Movesp , p ∈ P , the maximum number of movements between sector of

shovel p over the scheduling horizon.
• Target Gradej ≥ 0, j ∈ J , the grade target of the content j by the ore processing
plant (in percentage).
• N, a very large number.

4.3 Variables

To ease the reading of the model, all binary variables have a bar on top of it, while
continuous variables do not have the bar.
The decision variables are used to determine the location and duration of shovel
assignments, to control the movement of the shovels, to determine what mining
faces are currently active, and to account for tonnages:

• xp,f ,t ∈ [0, 1], fraction of period t ∈ T that shovel p ∈ P is assigned to face

f ∈ F.
• x̄ p,f ,t ∈ {0, 1}, equal to 1 if shovel p ∈ P is allocated to mining face f ∈ F in
period t ∈ T , 0 otherwise.
• yp,f ,t ∈ [0, 1], fraction of the period t ∈ T where shovel p ∈ P is operational at
mining face f ∈ F , i.e., the shovel is loading and sending material to its destina-
tion.
• zp,f ,t ∈ [0, 1], the time percentage of the period t ∈ T where shovel p ∈ P is oper-
ative at mining face f ∈ F sending material to its predefined stockpile.
• w̄ f ,t ∈ {0, 1}, equals to 1 if mining face f ∈ F finishes its exploitation in period
t ∈ T or before, 0 otherwise.
• v̄ p,r,t ∈ {0, 1}, equals to 1 if shovel p ∈ P goes through route r ∈ R in period
t ∈ T , 0 if not.
• 𝛽f ,t ≥ 0, tonnage of the stockpile f ∈ Fsp at the end of the period t ∈ {0} ∪ T .
(For simplicity, in the model, 𝛽f ,t=0 is fixed to the initial tonnage in stockpile f.)

Thus, the optimization model decides if the material contained in mining face of ore
f that is not a stockpile is either sent directly to the plant or sent to the predefined
stockpile.

4.4 Objective functions

We consider different possible objective functions, which are listed below. These
objective functions correspond to deviations from targets like production, desired
grades, or waste removal.
The deviations that we consider were introduced in Sect. 3.1.The explicit math-
ematical expressions of these quantities are presented in Eqs. (1, 2, 3, 4, 5, 6, 7, 8).
(Notice that 𝛥D is defined in terms of 𝛥dt.)
∑
𝛥O = M 0 − Prodp,f ⋅ yp,f ,t
ore sp
(1)
p∈P,f ∈F ⧵F ,t∈T

13
Short‑term open‑pit production scheduling optimizing multiple… 695

∑
𝛥P = P0 − Prodp,f ⋅ yp,f ,t
(2)
p∈P,f ∈Fore ,t∈T

∑
𝛥W = W 0 − Prodp,f ⋅ yp,f ,t
(3)
p∈P,f ∈Fwaste ,t∈T

∑
𝛥dt = Capt − Prodp,f ⋅ yp,f ,t ∀t ∈ T
(4)
p∈P,f ∈Fore

0 ≤ 𝛥dt ≤ 𝛥D ∀t ∈ T (5)
∑
Prodp,f ⋅ Gradej,f ⋅ yp,f ,t + 𝛥Grade−j,t − 𝛥Grade+j,t
p∈P,f ∈Fore (6)
= Target Gradej ∀j ∈ J, t ∈ T

𝛥Grade+j,t ≤ 𝛥Gradej ∀j ∈ J, t ∈ T (7)

𝛥Grade−j,t ≤ 𝛥Gradej ∀j ∈ J, t ∈ T (8)

4.5 Constraints

Now, we present the mathematical expressions corresponding to the constraints of

the optimization model. Second, we describe them in groups associated to different
concepts. Due to the complexity of the model, we group the constraints and provide
separate descriptions.

4.5.1 Production constraints

Constraints (9)–(12) impose the ore plant capacity and models the distribution of
time of each shovel.
Constraint (9) imposes that the total material extracted in each mining face along
the planning horizon must be less or equal than the total material contained in that
mining face. Constraint (10) sets the minimum and maximum ore tonnages sent to
the ore processing plant. Constraint (11) limits the minimum and maximum con-
tents of component j sent to the ore processing plant. Constraint (12) models the
shovel time by imposing that the effective shovel time plus the shovel movement
time between sectors are less than or equal to the maximum shovel time utilization.
∑
Prodp,f ⋅ yp,f ,t ≤ Tonf ∀f ∈ F
(9)
p∈P,t∈T

13
696 F. Manríquez et al.

∑
Min Prodt ≤ Prodp,f ⋅ yp,f ,t ≤ Capt ∀t ∈ T
(10)
p∈P,f ∈F

∑
Min Gradej ⋅ Prodp,f ⋅ yp,f ,t
p∈P,f ∈Fore
∑
≤ Gradej,f ⋅ Prodp,f ⋅ yp,f ,t
(11)
p∈P,f ∈Fore
∑
≤ MaxGradej ⋅ Prodp,f ⋅ yp,f ,t ∀t ∈ T, j ∈ J
p∈P,f ∈Fore

∑ ∑
Lengtht ⋅ xp,f ,t + Travel Timep,r ⋅ v̄ p,r,t ≤ Avl Timep,t ∀p ∈ P, t ∈ T
f ∈F r∈R
(12)

4.5.2 Shovel allocation constraints

The dynamics of shovel assignment to mining faces is determined by constraints

(13)–(17): Constraint (13) sets the precedences between faces, not allowing a shovel to
work in a face before the previous face is completely extracted. Constraint (14) imposes
that a shovel can allocate time for working in face f during period t only if it has been
assigned to that f during t. Constraint (15) ensures that when a mining face is depleted,
no other shovel can be assigned to it (this prevents shovels from using ”old” min-
ing faces as shortcuts to move between different places in the mine). Constraint (16)
imposes that extraction of mining face f is not fulfilled in period t until all the material
in the face is completely extracted. Constraint (17) ensures that if extraction of mining
face f is concluded in a period t, then it remains in that state until the end of the sched-
uling horizon.
xp,f ,t ≤w̄ f � t ∀(f , f � ) ∈ Q (13)

xp,f ,t ≤̄xp,f ,t ∀p ∈ P, f ∈ F, t ∈ T (14)

x̄ p,f ,t ≤1 − w̄ f ,t−1 ∀p ∈ P, f ∈ F, t ∈ T ⧵ {1} (15)

∑ Prodp,f
w̄ f ,t ≤
Tonf
⋅ yp,f ,t ∀f ∈ F, t ∈ T (16)
p∈P,t∈T

w̄ f ,t ≥w̄ f ,t−1 ∀f ∈ F, t ∈ T ⧵ {1} (17)

13
Short‑term open‑pit production scheduling optimizing multiple… 697

4.5.3 Shovel movement

Constraints (18)–(21) model the movement of each shovel between mine faces of
different sectors.
Constraint (18) sets the maximum number of movements between mining faces
along the scheduled horizon for each shovel. Constraint (19) ensures that shovel
p uses at most one route during period t. Constraint (20) makes sure that, during
period t, shovel is allocated only to mining faces that are visited during that period.
Constraint (21) ensures that the movement between different faces is consistent, i.e.,
that the last mining face of the route r and the first mining face of the r′ route must
be equal.
∑
(|r| − 1) ⋅ v̄ p,r,t ≤ Max Movesp ∀p ∈ P (18)
r∈R,t∈T

∑
v̄ p,r,t ≤ 1 ∀p ∈ P, t ∈ T (19)
r∈R

∑
x̄ p,f ,t ≤ v̄ p,r,t ∀p ∈ P, f ∈ F, t ∈ T
(20)
r∈Rf

∑
v̄ p,r,t ≤ v̄ p,r� ,t−1 ∀p ∈ P, r ∈ R, t ∈ T ⧵ {1}
(21)
r� ∈Hr

4.5.4 Stockpile constraints

Constraints (22)–(25) models the stockpiles.

Constraint (22) defines the time allocated to send material to stockpiles. Con-
straint (23) imposes that stockpiles do no sent material to stockpiles. Constraint (24)
sets the initial ore tonnage of the mining faces that are of type stockpile. Constraint
(25) defines the inventory of ore tonnage in a stockpile.
yp,f ,t + zp,f ,t ≤ xp,f ,t ∀p ∈ P, ∈ F, t ∈ T (22)

zp,f ,t = 0 ∀p ∈ P, ∈ Fsp , t ∈ T (23)

𝛽f ,0 = Tonf ∀f ∈ Fsp (24)

∑ ∑
Prodp,f � ⋅ zp,f � ,t + 𝛽f ,t−1 = Prodp,f ⋅ yp,f ,t + 𝛽f ,t
p∈P,f � ∈Sf p∈P (25)
∀f ∈ Fsp , t ∈ T

13
698 F. Manríquez et al.

Table 3  Length of periods and Period Period length [days] Ore plant
ore plant capacity capacity
[kt]

1 1 0
2 4 42
3 3 0
4 4 42
5 3 0
6 4 42
7 3 0
8 4 39
9 3 0
10 2 15
Total 31 180

Table 4  Shovel fleet parameters Shovel Effective through- Maximum utiliza- Speed [km/h]
put [t/h] tion [%]

1 1350 43 2
2 1300 35 2
3 1200 47 15
4 1150 63 7
5 1150 45 7
6 1200 40 15

5 Case study

In this section, we describe the real-scale open-pit mine case study and outline the
different experiments that were performed.

5.1 Mine operation description

The case study comprises mining faces distributed in six sectors, so the notation
for mining face names is N_LEVEL, where N is the sector and LEVEL is the z coor-
dinate. The mine planning horizon is one month, split into ten periods, each last-
ing between one and four days. The main parameters related to the mine operation
are summarized in Tables 3 to 7: Table 3 summarizes the processing capacities and
period lengths for each period ( Prodt = 0, ∀t , i.e., there is no minimum production
requirement). Table 4 shows the parameters related to shovels. Table 5 presents the
tonnage of ore and waste contained in each mining face. In this case study, min-
ing face 6_1380 is the only stockpile and has 0 tons of inventory at the beginning

13
Short‑term open‑pit production scheduling optimizing multiple… 699

Table 5  Ore and waste tonnage Mining face Sector Waste [kt] Ore [kt]
of mining faces (in [kt])
1_1380 1 335 4
2_1330 2 40 2
2_1320 2 299 41
3_1280 3 21 2
3_1270 3 102 20
3_1260 3 131 10
3_1250 3 22 2
3_1240 3 618 94
4_1230 4 85 4
5_1240 5 59 0
5_1230 5 288 1
6_1380 6 0 0
Total tonnage 1,999 180

Table 6  Material flow targets Indicator Value [kt] Description

for the planning horizon
P0 180 Total Ore sent to the plant.
W0 1999 Total waste movement.
M0 180 Total extracted material from the mine.

Table 7  Distance between Sectors 1 2 3 4 5 6

sectors (in km)
1 – 1.7 2.4 2.8 3.2 2.4
2 1.7 – 0.7 1 1.5 1.6
3 2.4 0.7 – 0.4 0.5 1.3
4 2.8 1 0.4 – 0.1 1.8
5 3.2 1.5 0.5 0.1 – 2.1
6 2.4 1.6 1.3 1.8 2.1 –

of the scheduling horizon. Table 6 presents the targets for production and material
flow defined by the long-term plan. Finally, Table 7 reports the traveling distance
between sectors (travel time for mining faces in the same sector is assumed to be
negligible).

5.2 Experimental design

The model is used under different configurations to analyze how the performance
indicators presented in Sect. 3.1 change under different circumstances. There are
a total of 28 configurations, which are obtained by considering the alternatives

13
700 F. Manríquez et al.

described below. The remaining constraints and parameters are common to all
instances.

• Optimization criteria In this case the following deviations are considered as

objectives for minimization (a) plant utilization (𝛥P ), (b) waste extraction (𝛥W ),
and (c) plant utilization due to ore directly sent from mine (𝛥O) (see Sect. 4).
Each of these criteria is evaluated using a single-step optimization and several
two steps hierarchical configurations. Table 8 summarizes the criteria considered
for the single optimization and hierarchical optimization in each case. For exam-
ple, configuration 𝛥P(𝛥W) corresponds to the case where 𝛥P is minimized sub-
ject to 𝛥W , i.e., first the waste deviation 𝛥W is minimized and then plant devia-
tion 𝛥P is minimized, subject to the waste deviation, hierarchically. Finally, in
terms of the tolerance parameters used in the hierarchical method, we set 𝜆i = 1,
that is, we do not permit any deterioration in the values when compared with a
single-objective approach. The main reason for this is that the focus of this work
is comparing how different configurations (objective functions and hierarchies)
impact the results in terms of their performance, which is difficult to be done
objectively if the 𝜆 parameters are different for each configuration.
• Presence or absence of the stockpile These two options are labeled as “Yes” and
“No”, respectively, in the corresponding results.
• Fixed or mobile shovel fleet For this, two possible configurations are considered:
a static fleet, meaning that shovels must remain in the same sector where they
start, and fully mobile fleet, in which case shovels could change sector at most
once during the planning horizon.

5.3 Computational resources

All the schedules presented in this study were obtained on a 2.60 GHz IntelⓇ XeonⓇ
CPU, with 256 GB RAM, running Windows 8 Ⓡ . The optimization model was solved
using Gurobi Optimizer version 8.1 (Gurobi Optimization 2019). We impose a mini-
mum MIP gap of 5.0% when solving the optimization problems. We believe that this

Table 8  Different optimization of short-term objectives considered in case study

Opt. criteria Notation Description of target to be Minimized

Single 𝛥W Waste deviation from target

𝛥P Total ore sent to the plant deviation from production target
𝛥O Deviation of ore sent directly from the mine to the plant
Hierarchical 𝛥P(𝛥W) Plant deviation s.t. minimum waste deviation
𝛥W(𝛥O) Waste deviation s.t. minimum ore deviation
𝛥W(𝛥P) Waste deviation s.t. minimum plant deviation
𝛥O(𝛥W) Total ore deviation s.t. minimum waste deviation

13
Short‑term open‑pit production scheduling optimizing multiple… 701

Table 9  Computational results: Running Time, MIP Gap and Objective Value for all optimization crite-
ria
Result Running Time [min] MIP gap [%] Objective Value
Fleet Static Mobile Static Mobile Static Mobile

Stockpile No Yes No Yes No Yes No Yes No Yes No Yes

𝛥W 0.5 0.9 30.8 0.4 0.0 0.0 0.0 0.0 85 85 35 0
𝛥W(𝛥O) 1.3 0.7 7.0 23.3 0.0 0.0 3.1 0.0 767 767 111 767
𝛥W(𝛥P) 1.5 22.5 10.2 102.9 0.0 0.0 0.0 4.1 767 767 112 767
𝛥O 0.3 0.6 6.3 15.5 0.0 0.0 0.0 0.0 49 49 18 11
𝛥O(𝛥W) 0.4 0.7 116.3 31.7 0.0 0.0 0.2 3.5 85 85 49 26
𝛥P 0.3 0.5 4.9 34.4 0.0 0.0 0.0 0.0 49 49 18 10
𝛥P(𝛥W) 0.5 16.0 45.9 518.2 0.0 0.0 0.0 4.7 85 85 49 26

value is an adequate trade-off between the resolution time and the objective func-
tion value. Indeed, Table 9 shows that most of the solutions obtained are optimal
and that the schedules with a MIP gap above 3.0% are schedules associated with a
mobile fleet of shovels, which are more complex for the solver to obtain the sched-
ule. We tried to solve all the schedules with a gap of 0, but it was not possible in
some instances due to the high computational time.

6 Results and discussion

This section presents the results and discussion based on applying the optimization
model proposed to the open-pit mine case study.
The resolution time, MIP gap, and objective function of the schedules are pre-
sented in Table 9. Overall, we observe that excepting the case 𝛥P(𝛥W), computation
times are reasonable (less than two hours). We also observe an impact of ×2, ×9
or even unbounded on 𝛥W (deviation of removed waste material) if its optimiza-
tion is secondary to production deviations 𝛥O or 𝛥P . In contrast, if the deviation in
waste is first optimized and then 𝛥P or 𝛥O are optimized, the impact in production
is significant. For example, the deviation in 𝛥P and 𝛥O augments from 49 to 85 (i.e.
, a 73.5% increase) when the fleet is static. , and relatively more when the fleet is
mobile, with an increase of 172% in 𝛥P and 𝛥O if there is no stockpile, and about
160% if there is a stockpile (again for both criteria). This implies that prioritizing
waste deviation can have a huge impact in the value of the plans.
We compare the schedules obtained in terms of the compliance performance indi-
cators presented in Table 2: waste movement (𝛤W ), plant utilization (𝛤P ), mine to
plant (𝛤O), and mining extraction (𝛤M).
Table 10 presents the results in the case where the stocks are deactivated in the
model. We observe that for the static fleet, when the priority is production (𝛥P and
𝛥O), the hierarchical method cannot improve the single-criteria results, i.e., the per-
formance of all strategies is equivalent, even in terms of waste extraction. However,

13
702 F. Manríquez et al.

Table 10  Performance of the Fleet Static Mobile

different strategies in terms
of compliance (%) , when no Indicator 𝛤W 𝛤P 𝛤O 𝛤M 𝛤W 𝛤P 𝛤O 𝛤M
stocks are available
𝛥W 96 20 20 20 98 72 72 72
𝛥W(𝛥O) 62 73 73 73 94 90 90 90
𝛥W(𝛥P) 62 73 73 73 94 90 90 90
𝛥O 62 73 73 73 65 90 90 90
𝛥O(𝛥W) 96 53 53 53 98 73 73 73
𝛥P 62 73 73 73 65 90 90 90
𝛥P(𝛥W) 96 53 53 53 98 73 73 73

if the fleet is allowed to move between sectors, then 𝛥W(𝛥O) and 𝛥W(𝛥P) gener-
ate plans which are equivalent for production and mining performance, but with
a significant increase in waste extraction, improving the compliance from 65% to
94%. This improvement is important because the compliance of 65% means that the
remaining 35% of waste would be left to be extracted in future periods, thus nega-
tively impacting production for incoming months.
If the mine planner has the extraction of waste as a priority, then the hierarchical
method outperforms 𝛥O. Indeed, the method improves compliance indicators asso-
ciated to production (i.e., 𝛤P , 𝛤O) considerably. If the fleet is static, these indicators
fall from 73% to 53%, instead of 20%. When the fleet is allowed to move between
sectors, it can increase the production’s compliance from 72% to 73%.
It is worth noting that having waste as the highest priority is not unrealistic.
Whenever there is a shortage in equipment (for example, due to mechanical fail-
ures), mine operations prefer assigning equipment to ore mining faces; therefore,
delaying waste removal. As this effect accumulates over time, planners eventually
need to set waste extraction as a priority to not risk shortages of ore in the future.
In such situations, the results suggests that using the hierarchical approach would
help planners to get up to date in waste removal with a much better performance in
terms of production, when compared to a single approach that only aims to maxi-
mize waste extraction.

Table 11  Performance of the Fleet Static Mobile

different strategies in terms of
compliance, when stocks can Indicator 𝛤W 𝛤P 𝛤O 𝛤M 𝛤W 𝛤P 𝛤O 𝛤M
be used
𝛥W 96 0 0 20 100 0 0 72
𝛥W(𝛥O) 62 73 73 80 62 94 94 95
𝛥W(𝛥P) 62 73 73 80 62 95 50 95
𝛥O 62 73 73 73 62 94 94 95
𝛥O(𝛥W) 96 53 53 60 100 86 86 100
𝛥P 62 73 44 73 62 95 68 95
𝛥P(𝛥W) 96 53 53 53 100 86 86 100

13
Short‑term open‑pit production scheduling optimizing multiple… 703

Table 11 presents the results when stocks are enabled in the model. First, we
observe that, as expected, stocks’ availability positively impacts compliance overall.
Second, the results are consistent with the ones for the case of Table 10:

• When waste extraction is a priority, the hierarchical method performs much bet-
ter than 𝛥W (which sends no mineral to the plant).
• When production has the highest priority, 𝛥P reaches values of 𝛤O and 𝛤M , which
suggests that it tends to rely on stocks more to achieve the same 𝛤P = 73% than
other strategies. Such a plan is risky because it may deplete the stockpile and also
more expensive because of the extra cost of rehandling.

Overall, 𝛥W(𝛥O) and 𝛥W(𝛥P) are the strategies with the best performance in terms
of production: for the static fleet, it produces the best results, and for the mobile fleet
case, it reaches a value of 𝛤P = 94%, which is 1% lower than the maximum possible,
but with the highest possible compliance of material being sent directly from the
mine to the plant.

6.1 Discussion

Overall, the model seems to abstract the mine operation in a proper manner and fol-
low some expected behaviour. For example, in general terms, compliance increases
with fleet mobility and with the possibility to use stockpiles. Also, it tends to favor
compliance indicators for the corresponding optimization targets.
Aplying the hierarchical method to short-term planning generates plans that are
more balanced and robust when compared to single objective optimization. This
is because a single objective approach cannot take into account other criteria and,
therefore, may produce plans that either have a poor impact during the planning
horizon, or may hide problems for future periods of time.
The case study shows that the results can change substantially depending on the
criteria used and their rank for the hierarchical method. We consider this, in fact, a
feature of the mathematical model, which is flexible enough to incorporate all the
criteria, therefore providing the planner with different choices, but also providing
valuable information about potential issues. This is the case, for example, of the plan
for 𝛥P and stockpiles (Table 11), which promises a high production compliance, but
at the expense of delaying extraction from the mine.
The results are also interesting for the specific mine, because even in the best
cases, there seems to be a trade-off between production and waste extraction. Indeed,
the best strategies for production may reach over 90% compliance, but at the expense
of having a 62% confidence for waste extraction, or conversely, when a 𝛤W = 100%
is reached, the compliance for production are 𝛤P = 𝛤O = 86%).
Overall, the hierarchical method (especially 𝛥W(𝛥O) and 𝛥W(𝛥P)) generates
plans that are more robust, because they do not leave production or waste extraction
for future periods and, therefore, promise a better adherence to long-term plans and
goals.

13
704 F. Manríquez et al.

7 Conclusions and future work

In this study, we propose a MILP optimization model to address the problem of

shovel assignment and movement in an open-pit short-term mine context, where
production and waste extraction targets have been set by long-term plans and
stockpiles are available as material buffers. The model can be used with different
optimization targets to implement single objectives or a hierarchical method that
allows optimization of multiple criteria.
To evaluate the performance of the model with different criteria, we propose
compliance performance indicators, which assess the quality of the plan gener-
ated in terms of how close planned tonnages are with regards to their targets.
Thus, mine planners can use these indicators to evaluate and compare multiple
short-term schedules.
We apply the proposed optimization model to a real-scale open-pit mine case
study, over a time horizon of a month, in which mining faces are distributed over
six mining sectors. The mine operation has one ore processing plant and uses six
shovels and has one stockpile for operation. In this setting, we utilize the model to
generate schedules under different scenarios; namely, single-optimization or hier-
archical optimization of different short-term objectives, presence or absence of a
stockpile, a mobile or a fixed shovel fleet. The objective is to study the impact of
the different scenarios on the schedule indicators.
The results of the case study show that: (a) the hierarchical method can gener-
ate short-term mine production schedules optimizing the considered objectives,
(b) when applying the hierarchical optimization method, both the objectives and
the order of optimization of these have a great impact on the values of the differ-
ent schedule indicators, (c) in general, schedules with a stockpile obtain higher
schedule indicators compared to the ones with no stockpile, and (d) schedules
with a mobile shovel fleet obtain higher schedules’ indicators than the ones with
a fixed shovel fleet.
More importantly, in general, we observe that the plans generated using the
hierarchical method are more robust, because they minimize potential delays on
other criteria (not considered in the single objective approach). It is also interest-
ing to see that by using all proposed strategies for optimization (single and hierar-
chical), the different plans can be analyzed in terms of their relative strengths and
weaknesses. That is, the planner can choose from different plans, making a more
informed decision.
As future work, we want to incorporate more aspects of the mining operation
in the optimization model such as: scheduled shovel maintenance, allocation of
drilling rigs to mining faces, multiple ore processing plants, multiple stockpiles,
and grade blending in the stockpiles. We also intend to simulate the short-term
mine production schedule generated by the optimization model. We plan to apply
discrete-event simulation to assess the probability of compliance of the schedule.
Finally, we plan to incorporate both geological uncertainty and equipment perfor-
mance uncertainty in the optimization model.

13
Short‑term open‑pit production scheduling optimizing multiple… 705

Acknowledgements This research was carried out during Fabián Manríquez PhD, which was funded by
CONICYT grant PFCHA/DOCTORADO BECAS CHILE/2019 - 21190201. Nelson Morales and Héctor
Gonzalez were partially funded by CONICYT/PIA Project AFB180004 grant “Advanced Mining Tech-
nology Center. The authors would also like to thank Dr. Carlos Arroyo, Luis Navea, Daniel Mariño and
Dr. Xavier Emery for contributing with ideas and revisions of the work, and Vilson Reis and Wesley
Almeida for providing the real-data of the case study.

References
Akaike A, Dagdelen K (1999) A strategic production scheduling method for an open pit mine. Proceed-
ings of the 28th application of computers and operation research in the mineral industry pp 729–738
Alexandre RF, Campelo F, Vasconcelos JA (2019) Multi-objective evolutionary algorithms for the truck
dispatch problem in open-pit mining operations. Learn Nonlinear Models 17(2):53–66. https://​doi.​
org/​10.​21528/​lnlm-​vol17-​no2-​art5
Asad MW (2005) Cutoff grade optimization algorithm with stockpiling option for open pit mining opera-
tions of two economic minerals. Int J Surf Min Reclam Environ 19(3):176–187. https://​doi.​org/​10.​
1080/​13895​26050​02586​61
Benlaajili S, Moutaouakkil F, Chebak A, Medromi H, Deshayes L, Mourad S (2020) Optimization of
truck-shovel allocation problem in open-pit mines. communications in computer and information
science 1207 CCIS:243–255, https://​doi.​org/​10.​1007/​978-3-​030-​45183-7_​19
Bley A, Boland N, Froyland G, Zuckerberg M (2009) Solving mixed integer nonlinear programming
problems for mine production planning with a single stockpile. In: Technical Report 2009/21, Insti-
tute of Mathematics, TU Berlin
Bley A, Boland N, Froyland G, Zuckerberg M (2012a) Solving mixed integer nonlinear programming
problems for mine production planning with stockpiling. Optimization Online http://www.optimiza-
tion-online org/DB_HTML/2012/11/3674 html ()
Bley A, Gleixner AM, Koch T, Vigerske S (2012b) Comparing MIQCP solvers to a specialised algo-
rithm for mine production scheduling. In: Modeling, simulation and optimization of complex
processes, Springer Berlin Heidelberg, Berlin, Heidelberg, pp 25–39, https://​doi.​org/​10.​1007/​
978-3-​642-​25707-0_3
Blom M, Pearce AR, Stuckey PJ (2017) Short-term scheduling of an open-pit mine with multiple objec-
tives. Eng Optim 49(5):777–795. https://​doi.​org/​10.​1080/​03052​15X.​2016.​12180​02
Blom M, Pearce AR, Stuckey PJ (2018) Short-term planning for open pit mines: a review. Int J Min Rec-
lam Environ 0930:1–22. https://​doi.​org/​10.​1080/​17480​930.​2018.​14482​48
Both C, Dimitrakopoulos R (2020) Joint stochastic short-term production scheduling and fleet manage-
ment optimization for mining complexes. Optim Eng 21(4):1717–1743. https://​doi.​org/​10.​1007/​
s11081-​020-​09495-x
Chanda EC (1990) An application of integer programming and simulation to production planning for
a stratiform ore body. Min Sci Technol 11(2):165–172. https://​doi.​org/​10.​1016/​0167-​9031(90)​
90318-M
Eivazy H, Askari-Nasab H (2012) A mixed integer linear programming model for short-term open pit
mine production scheduling. Min Technol 121(2):97–108. https://​doi.​org/​10.​1179/​17432​86312Y.​
00000​00006
Espinoza D, Goycoolea M, Moreno E, Newman A (2013) Minelib: a library of open pit mining problems.
Ann Oper Res 206(1):93–114
Fioroni M, Franzese L, Bianchi T, Ezawa L, Pinto L, de Miranda G (2008) Concurrent simulation and
optimization models for mining planning. In: 2008 Winter simulation conference, pp 759–767,
https://​doi.​org/​10.​1109/​WSC.​2008.​47361​38
Fu Z, Asad MWA, Topal E (2019) A new model for open-pit production and waste-dump scheduling.
Eng Optim 51(4):718–732. https://​doi.​org/​10.​1080/​03052​15X.​2018.​14765​01
Grodzevich O, Romanko O (2006) Normalization and other topics in multi-objective optimization.
Fields-MITACS industrial problems workshop. Canada, Toronto, pp 89–101
Gurobi Optimization L (2019) Gurobi optimizer reference manual. http://​www.​gurobi.​com

13
706 F. Manríquez et al.

Hoerger S, Seymour F, Hoffman L (1999) Mine planning at Newmont’s Nevada operations. Min Eng
51(10):26–30
Johnson TB (1968) Optimum open pit mine production scheduling. California Univ. Berkeley Operations
Research Center, Tech. report
Jupp K, Howard TJ, Everett JE (2014) Role of pre-crusher stockpiling for grade control in iron ore
mining. Transactions of the Institutions of Mining and Metallurgy, Section B: Appl Earth Sci
122(4):242–255. https://​doi.​org/​10.​1179/​17432​75814Y.​00000​00045
Koushavand B, Askari-Nasab H, Deutsch CV (2014) A linear programming model for long-term mine
planning in the presence of grade uncertainty and a stockpile. Int J Min Sci Technol 24(4):451–459.
https://​doi.​org/​10.​1016/j.​ijmst.​2014.​05.​006
Lamghari A, Dimitrakopoulos R (2016) Network-flow based algorithms for scheduling production in
multi-processor open-pit mines accounting for metal uncertainty. Eur J Oper Res 250(1):273–290.
https://​doi.​org/​10.​1016/j.​ejor.​2015.​08.​051
Lane KF, Hamilton DJ, Parker JJB (1984) Cutoff grades for two minerals. In: Application of computers
and mathematics in the minerals industries. International symposium. 18, pp 485–491
Levinson Z, Dimitrakopoulos R (2019) Simultaneous stochastic optimisation of an open-pit gold mining
complex with waste management. Int J Min Reclam Environ 00(00):1–15. https://​doi.​org/​10.​1080/​
17480​930.​2019.​16214​41
L’Heureux G, Gamache M, Soumis F (2013) Mixed integer programming model for short term plan-
ning in open-pit mines. Min Technol 122(2):101–109. https://​doi.​org/​10.​1179/​17432​86313Y.​00000​
00037
Matamoros ME, Dimitrakopoulos R (2016) Stochastic short-term mine production schedule accounting
for fleet allocation, operational considerations and blending restrictions. Eur J Oper Res 255(3):911–
921. https://​doi.​org/​10.​1016/j.​ejor.​2016.​05.​050
Moreno E, Ferreira F, Goycoolea M, Espinoza D, Newman A, Rezakhah M (2015) Linear programming
approximations for modeling instant-mixing stockpiles. In: Application of computers and operations
research in the mineral industry - proceedings of the 37th international symposium, APCOM 2015,
2009, pp 582–587
Moreno E, Rezakhah M, Newman A, Ferreira F (2017) Linear models for stockpiling in open-pit mine
production scheduling problems. Eur J Oper Res 260(1):212–221. https://​doi.​org/​10.​1016/j.​ejor.​
2016.​12.​014
Mousavi A, Kozan E, Liu SQ (2016) Open-pit block sequencing optimization: A mathematical model and
solution technique. Eng Optim 48(11):1932–1950. https://​doi.​org/​10.​1080/​03052​15X.​2016.​11420​80
Ramazan S, Dimitrakopoulos R (2013) Production scheduling with uncertain supply: a new solution to
the open pit mining problem. Opt Eng https://​doi.​org/​10.​1007/​s11081-​012-​9186-2
Rehman SU, Asad MWA (2010) A mixed-integer linear programming (MILP) model for short-range pro-
duction scheduling of cement quarry operations. Asia-Pacific J Oper Res 27(3):315–333. https://​doi.​
org/​10.​1142/​S0217​59591​00027​27
Rezakhah M, Newman A (2020) Open pit mine planning with degradation due to stockpiling. Comp
Operations Res. https://​doi.​org/​10.​1016/j.​cor.​2018.​11.​009
Rezakhah M, Moreno E, Newman A (2020) Practical performance of an open pit mine scheduling model
considering blending and stockpiling. Comput Oper Res 115:104638. https://​doi.​org/​10.​1016/j.​cor.​
2019.​02.​001
Robinson GK (2004) How much would a blending stockpile reduce variation? In: Chemometrics and
intelligent laboratory systems, Elsevier, https://​doi.​org/​10.​1016/j.​chemo​lab.​2004.​03.​010
Samavati M, Essam D, Nehring M, Sarker R (2018) A new methodology for the open-pit mine produc-
tion scheduling problem. Omega (United Kingdom) 81:169–182. https://​doi.​org/​10.​1016/j.​omega.​
2017.​10.​008
Shah KS, Rehman SU (2020) Modeling and optimization of truck-shovel allocation to mining faces in
cement quarry. J Mining and Environ 11(1):21–30. https://​doi.​org/​10.​22044/​jme.​2019.​8329.​1712
Silva MF, Dimitrakopoulos R, Lamghari A (2015) Solving a large SIP model for production scheduling at
a gold mine with multiple processing streams and uncertain geology. transactions of the institutions
of mining and metallurgy, section A: Mining Technol 124(1):24–33. https://​doi.​org/​10.​1179/​17432​
86314Y.​00000​00075
Smith ML (1998) Optimizing short-term production schedules in surface mining: Integrating mine mod-
eling software with ampl/cplex. Int J Surf Min Reclam Environ 12(4):149–155. https://​doi.​org/​10.​
1080/​09208​11890​89440​38

13
Short‑term open‑pit production scheduling optimizing multiple… 707

Tabesh M, Askari-Nasab H, Peroni R (2015) A comprehensive approach to strategic open pit mine plan-
ning with stockpile consideration. Application of computers and operations research in the mineral
industry - proceedings of the 37th international symposium, APCOM 2015 (2020):326–332
Torkamani E, Askari-Nasab H (2011) A mathematical programming model for optimal truck-shovel
allocation. MOL Research report three 2010/2011, MOL, Mining Optimization Laboratory
106(August):99–105
Torkamani E, Askari-Nassab H (2015) A linkage of truck-and-shovel operations to short-term mine plans
using discrete-event simulation. Int J Mining Miner Eng 6:97
Upadhyay S, Askari-Nasab H (2016) Truck-shovel allocation optimisation: a goal programming
approach. Min Technol 9009:1–11. https://​doi.​org/​10.​1179/​17432​86315Y.​00000​00024
Upadhyay S, Askari-Nasab H (2018) Simulation and optimization approach for uncertainty-based short-
term planning in open pit mines. Int J Min Sci Technol 28(2):153–166
Upadhyay S, Askari-Nasab H (2019) Dynamic shovel allocation approach to short-term production plan-
ning in open-pit mines. Int J Min Reclam Environ 33(1):1–20. https://​doi.​org/​10.​1080/​17480​930.​
2017.​13155​24

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

Authors and Affiliations

Fabián Manríquez1 · Héctor González2 · Nelson Morales3

1
Departamento de Ingeniería Metalúrgica y Minas, Universidad Católica del Norte, Antofagasta,
Chile
2
DELPHOS Mine Planning Laboratory, AMTC & DIMIN, Universidad de Chile, Santiago, Chile
3
Département de génies civil, geologique et des mines, Polytechnique Montréal, Montreal, QC,
Canada

13