Innovative optimization of iron ore blending in daily mine planning from a linear

programming model
Vidal Félix Navarro Torresª, Edilaila Moraesb
a
Vale Institute of Technology, Belo Horizonte, Minas Gerais, Brazil, [email protected]
b
Vale S.A, Belo Horizonte, Minas Gerais, Brazil, [email protected]

Abstract: This work aims to develop models and algorithms to mine planning to explore opportunities for better use
of mineable reserves, seeking to optimize economic and quality objectives, in an integrated way. The research
methodology aims to understand the mine planning processes and tools applied, and to identify requirements that guide
the development of mathematical models to support the decisions to reuse opportunity materials (i.e. materials before
considered as waste can be reevaluated due to changes in mining plans, processing technologies more evolved, increase
in commodity prices, etc.) and minimizing the variability of ROM grades in short-term planning. Analysis of
operational configurations that support greater flexibility in the handling and storage of opportunity materials were
also held in order to enable an objective comparison between the economic benefits of the approach and the costs and
investments involved.

Keywords: daily, mine planning, optimization, linear programming.

Optimización innovadora de la mezcla de mineral de hierro en la planificación

minera diaria desde un modelo de programación lineal
Resumen: Este trabajo tiene como objetivo desarrollar modelos y algoritmos para la planificación minera y explorar
oportunidades para un mejor uso de las reservas explotables, buscando optimizar los objetivos económicos y de
calidad, de manera integrada. La metodología de investigación tiene como objetivo comprender los procesos y
herramientas de planificación minera aplicados, e identificar los requisitos que guían el desarrollo de modelos
matemáticos para respaldar las decisiones de reutilizar materiales de oportunidad (es decir, los materiales que antes se
consideraban desechos pueden ser reevaluados debido a cambios en los planes de minería, tecnologías de
procesamiento más evolucionadas, aumento de los precios de los productos básicos, etc.) y minimización de la
variabilidad de los grados ROM en la planificación a corto plazo. También se realizó un análisis de las configuraciones
operativas que respaldan una mayor flexibilidad en el manejo y almacenamiento de los materiales de oportunidad para
permitir una comparación objetiva entre los beneficios económicos del enfoque y los costos e inversiones
involucrados.
Palabras clave: planificación minera, optimización, programación lineal.

1. Introduction availability in the pits are crucial to the

The definition of the ultimate pit of an ore reserve is accomplishment of production targets. The objective
associated with different economic and technical of mine planning and scheduling is to perform an
parameters that directly affect the viability of the optimization of the selection of the blocks to be
company. The orebody is represented by a block mined, in order to attend the blending constrains
model, and each block of this model is associated imposed by the plant.
with an economic value that can represent a gain (if However, a common difficulty to all economic
the block is classified as ore) or a cost (if the block is optimization algorithms for final pits regards to the
classified as waste). The objective of mine planning fact that the ultimate pit limits cannot be determined
is to generate a mining sequence that optimizes until the economic values of the blocks are known
economic and quality objectives, seeking the best with greater accuracy [2]. Actually, the economic
possible use of the deposit and respecting the values of the blocks cannot be known precisely until
restrictions imposed by the production process. At the mining sequence is specified. Hence, the mining
the interface between the mine and the processing schedule cannot be specified until the final limits are
plant, aspects such as quality of ore and its well defined, according to Carmo [3, 8]. Such
29
recurrences suggest an iterative process that requires utilization of the existing ore in reserve, despite the
careful att ention of those responsible for mine increased implementation complexity [6].
planning. Still, according to Rubio [4], the process of Alternatively, mathematical integer programming
determining the mine production plan should be done techniques allow the development of optimization
after the definition of the economic step, scheduling models for optimal selection of the ore body blocks
and the cut-off grades, which leads to simplification that, when properly blended, meet the specifications
of the optimization problem. The algorithms assign a of physical and chemical quality of the process.
default value to blocks without incorporating Maximizing the reservation involves selecting the
blending concepts or extraction strategy; there is no larger amount of quality blocks, respecting the
integration of operational variability and reserves precedence relations and the pit slope. Mathematical
characteristics in the calculation of production models that address this problem typically deal with
schedules. instances of great complexity, requiring the
The problem of recovery of mineral deposits is development of specific technical solution [7].
closely linked to methods for calculating the cut-off
grade and therefore, the ultimate pit design [1]. The 2. Objectives
benefit function of each block allows determining the The objective of this work is the development and
ultimate pit economically feasible, which depends on evaluation of mathematical models that support the
several variables, in particular the cut-off grade, decision-making process of mine production
which depends on commodity prices and planning aiming to better use of mineable reserves
technologies involved, directly influencing the life of and seeking lower variability of the feed grades of the
the mine, the average grade of the reserve and the processing plant in the short term planning, from
company actual value. The global average grade of optimizing the blending from planning and daily
an exploitable reserve is an important parameter scheduling of mining.
because it expresses, with some variability, the To achieve this main objective will be determined
material that can be turned into product. optimal ways for the use of mineable reserves,
The cut-off grade is a response to the non- through the evaluation of the blending of ore and
homogeneous nature of the materials exposed in waste blocks for a given planning horizon; short-term
working faces. Higher and lower grades materials production programs will be improved, in order to
can be blended, resulting in better use of the reserve minimize the variability of quality parameters in the
and in a more homogeneous feed to the plant. power of processing plants; and different operational
However, while the cut-off grade definition for one scenarios and their impact on the viability and
variable leads to a correct routine, the application of performance of the proposed plans will be evaluated.
various cut-off grades, in order to control the quality
of more variables, may lead to under-utilization of 3. Methods
reserves. Due to the rising trend of reduction in 3.1. Mathematical modelling
quality of mineral reserves, is crucial to control the The short-term planning seeks to determine a subset
several variables [5]. of blocks to be exploited in order to meet mass and
The parameterization of reserves is a widely used quality targets during a given period, monthly,
method for obtaining mass parameters and average quarterly, semiannual or yearly. However, when
cut-off grades. The simple parameterization seeks attention turns to the daily operations, it is essential
determining a grade-tonnage curve for the reserve, that the feeding of processing plants have the
setting different cut-off grades to obtain the expected minimal variation around the average quality. This is
average grade. The parameters for various cuts, in important to avoid negative impacts on the operating
turn, establishes cut-off grades for the variables of parameters and the key factors of plant performance.
interest, but seeks to minimize the deviation obtained Thus, for daily schedule for monthly production, the
by performing different cutting grades. Finally, the exploitation order of the blocks must be such that the
multivariable parameterization does not consider average quality of the total mass fed to the plant have
pre-established cut-off grades and seeks to determine the smallest possible variation over time. In this way,
a curve for obtaining a desired average grade. it was developed a preliminary model and two
Multivariate parameterization methods have been different approaches: the lower daily variation
studied more carefully, given the potential for better

30
around the monthly average and lower daily
variation. Subject to the following constraints:
The proposed model seeks to maximize the ROM −
𝑦𝑗𝑘 ≥ −𝑀(1 − 𝑧𝑗𝑘 ) ∀𝑗, 𝑘 (11)
mass supplied in a plan, by selecting ore blocks +
𝑦𝑗𝑘 ≥ −𝑀𝑧𝑗𝑘 ∀𝑗, 𝑘 (12)
presented in a set that represents the planning ∑𝑘 𝑦𝑖𝑘 = 1 ∀ 𝑖 ∈ 𝐵 (13)
horizon, according to the following objective ∑𝑖 𝑚𝑖 𝑦𝑖𝑘 ≤ 𝑢 ∀𝑘 (14)
function:
𝑦𝑖𝑘 ∈ {0,1} ∀𝑖 ∈ 𝐵 (15)
𝑚𝑎𝑥 ∑𝑖 𝑚𝑖 𝑦𝑖 (1)
𝑧𝑗𝑘 ∈ {0,1} ∀𝑗, 𝑘 (16)
Subject to the following constraints:
∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖 3.2. Application development
∑𝑖 𝑚𝑖 𝑦𝑖𝑘
≤ 𝑙𝑗 ∀ 𝑗 ∈ 𝑄 − (2)
∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖 Two different tools were developed.
≥ 𝑙𝑗 ∀ 𝑗 ∈ 𝑄 + (3)
∑𝑖 𝑚𝑖 𝑦𝑖 a) Maximize the overall mass of ore from the set
Where: of blocks available in the planning horizon.
k: 1, 2, ... K: periods (for daily schedule, K = 30); The weighted average of the quality of
mi: block mass i ∈ B; selected blocks must meet the quality
tij: grade of j ∈ (Q+ ∪ Q− ) present in the block i ∈ specifications of the power plant.
B; b) Minimize the daily variability of feeding the
lj : limit of j ∈ (Q+ ∪ Q− ) in the total mass; processing plants, by determining the optimal
yik: selection of block i ∈ B in the period k, yik ∈ {0, sequence of exploitation of certain blocks in
1}. the mining plan.
This objective function seeks to maximize the total
selected mass from the available blocks in the plan. For both approaches described above, the final
The lower daily variation around the monthly product comprises an integrated set of technologies
average approach uses the following objective to facilitate the use. The models were implemented
function: using Python programming language, in connection
𝑤𝑗 ∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖𝑘 with the solver Gurobi through the API available for
𝑚𝑎𝑥 ∑𝑗 𝐾
∑𝑘 |
∑𝑖 𝑚𝑖 𝑦𝑖𝑘
− 𝑙𝑗 | (4)
that language. The models were integrated into an
Where wj is the relative importance of the parameter Excel spreadsheet through SolverStudio supplement.
j ∈ (Q+ ∪ Q−) compared with the other, as wj ∈ [0, This is the spreadsheet user interface, and use is
1] and Pj wj = 1 where the parameters and input data is defined for the
Subject to the following constraints: problem.
∑𝑘 𝑦𝑖𝑘 = 1 ∀ 𝑖 ∈ 𝐵 (5)
∑𝑖 𝑚𝑖 𝑦𝑖𝑘 ≤ 𝑢 ∀𝑘 (6)
𝑦𝑖𝑘 ∈ {0,1} ∀𝑖 ∈ 𝐵, 𝑘 (7)
The objective function aims to minimize the sum of
the daily quality deviations related to the monthly
average, for each parameter. The constrains require
that each block only could be used once in the
considered plan, and guarantee that the daily feed
capacity of the processing plant have not be
exceeded.
The next formulation do not allow that large variation
occur from one day to another, this formulation seeks
to minimize the average daily variation, in order to
minimize impacts on the processing plant. The lower
daily variation approach use the following objective
function:
+
𝑦𝑗𝑘 = ∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖𝑘 − ∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖𝑘−1 ∀𝑗, 𝑘 (8)
−
𝑦𝑗𝑘 = ∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖𝑘−1 − ∑𝑖 𝑡𝑖𝑗 𝑚𝑖 𝑦𝑖𝑘 ∀𝑗, 𝑘 (9)
𝑤𝑗 Figure 1 Application integrated interface running to solve
+ −
𝑚𝑖𝑛 ∑𝑗 ∑𝑘(𝑦𝑗𝑘 + 𝑦𝑗𝑘 ) (10) the optimization model
𝐾

31
 3.3. Application on real cases

The application of this work has its focus in open pit

iron mines. In general, iron ore can be divided into
two main types: hematite (high iron grade ore, with
more than 63% grade) and itabirites (rocks formed by
alternating quartz and hematite beds).
The application of the proposed model provides a
higher theoretical baseline to the quality levels
reached in that reserve, considering the blending of
one or more blocks. For monthly plans, the selected
blocks are positioned so that there is need to consider
schedule and precedence constrains. In addition, the
manipulation of materials geographically and
temporally distant in the pit may require a feasibility
analysis of the construction and use of intermediate
stockpiles, which can be more consistently evaluated
Figure 2 Daily variation of contents, for Mine 1
through simulation models of discrete events.

4. Results and discussion

For an evaluation of the solution provided by the

updated model, two fundamental aspects of the
solutions provided by the model are illustrated. First,
we highlight the daily variation in the quality
parameters around the monthly average, which is the
objective function of the model. Figure 1 to Figure 4
show graphs of this variation to the databases of the
two mines provided: Mine 1 and Mine 2.

Figure 3 Daily variation of contents, for Mine 2

Figure 1 Daily variation of contents, for Mine 1

32
 Figure 6 Total monthly mass in the sequence determined
Figure 4 Daily variation of contents, for Mine 2 by the model for the Mine 2

Finally, it was indicated the monthly mass value of 4. Conclusions

ore achieved, compared with the possible upper limit Through testing it was found that the implementation
of the processing plant. Figure 5 and Figure 6 of the model, its extent and the corrections made
illustrates the total weight for each of the quality satisfy the demand, and the results of optimization
parameters for both Mine 1 and 2. were fully satisfactory in the context of the daily
mine planning.
The implementation of the model, using the Gurobi
solver, allows executing the instance chosen by the
user and optimizing the planning of daily sequencing
exploitation of mining blocks. Through an Excel
spreadsheet integrated into Solver Studio is provided
to the user an interface for insertion of quality targets
parameters, daily mine capacity and other problem
input data. Through tests, it was found that the
implementation of the model, its extension and
corrections made are consistent with the demand and,
therefore, it is concluded that the project was
successfully completed.
Acknowledgments
Special thanks to the work of researchers of Vale
Institute of Technology (ITV) and Vale S.A.
Company.
Figure 5 Total monthly mass in the sequence determined
by the model for the Mine 1 References

[1] A. S. Oliveira, L. Esawa, J. Gonçalves e T.

Bianchi, “Projeto Parametrização Multivariável,”
2012.
[2] A. Lamghari, R. Dimitrakopoulos e J. A. Ferland,
“A variable neighbourhood descent algorithm for the
open-pit mine production scheduling problem with

33
metal uncertainty,” Journal of the Operational http://brasilminingsite.com.br/includes/modulos/mdl
Research Society, vol. 65, pp. 1305-1314, 2014. _artigo/exibir_artigo.php?id=24, 2008.
[3] F. A. R. do Carmo, A. Curi e W. T. de Sousa, [7] J. Amaya, D. Espinoza, M. Goycoolea, E.
“Otimização econômica de explotações a céu Moreno, T. Prevost e E. Rubio, “A scalable approach
aberto,” Revista Escola de Minas, vol. 59, n. 3, 2006. to optimal block scheduling,” em Application of
[4] E. Rubio, “Planificación Minera: Desafíos y Computers and Operations Research in the Mineral
Nuevas Tendencias en la Industria Minera,” em Industry (APCOM), Vancouver, 2009.
Primer Taller de Planificación Minera, Santiago, [8] F. A. R. do Carmo, “Metodologias para o
Chile, 2009. Planejamento de Cavas Finais de Minas a Céu Aberto
[5] A. C. Girodo e J. L. Beraldo, “Elementos Básicos Otimizadas,” Ouro Preto, 2001.
de um Projeto de Mineração,” em Curso Projetos de
Mineração, Belo Horizonte, Instituto Brasileiro de Artículo recibido: 16.10.2019
Mineração, 1981.
Aceptado en: 04.11.2019
[6] A. C. Campos, J. Valente e A. C. Girodo,
“Parametrização Multivariável de Recursos e
Reservas,” Brasil Mining Site, n.

34