GOSPODARKA SUROWCAMI MINERALNYMI
Tom 24 2008 Zeszyt 4/2
JAVAD SATTARVAND*, CHRISTIAN NIEMANN-DELIUS**
Perspective of metaheuristic optimization methods in open pit
production planning
Introduction
Open pit mining refers to a method of extracting minerals from the earth by which the ore
body is accessed by opening a large stretch of ground to expose the ore to air. Mining begins
with a small pit in the surface, and then proceeds to a larger pit which encloses the small pit,
and the process continues until a final pit is reached. These pits are known as mining
sequences or push backs. The long-term planning of an open pit mine refers to the optimized
designing of these push backs in a way that maximize the objective function. The results of
long-term calculations are used as guides for the short-term production planning which may
be for a quarter, a month or a week.
Generally all stages of planning are based on a block model which is a set of imaginary
adjacent rock blocks of the entire ore deposit and surrounding material (waste). The model
may have several million blocks depending on the size of the ore body and the size of the
blocks. Normally the average ore content of each block is estimated from available explo-
ration data using geostatistical methods or simulation.
The variables involved in open pit mine production planning interact in a circular fashion,
that is, without knowledge of one variable the next variable in the circle cannot be deter-
mined (Fig. 1). The time it will take to mine all the pits in the sequence will define the mine
life, and the boundaries of the last pit in the sequence will determine the ultimate pit limits
(UPL). To distinguish ore from waste, a cutoff grade must be established that is a function of
commodity price, mining and milling costs. These are not known at the outset of the project.
* Ph.D. student at the Institute of mining engineering III, RWTH Aachen University.
** Univ. Professor at the Institute of mining engineering III, RWTH Aachen University.
144
As it has been shown in Figure 1, a cutoff grade must be defined to determine the extension of
the ore body, hence, the ultimate pit limits; then the ultimate pit limits are used to devise
a production schedule that in the next step will be used to calculate the revenues and costs
which in turn establish the most economic values and cutoff grade. As seen in this circular
relation, the value of a given variable cannot be determined if the value of the preceding
variable is not known. Fixing or assuming the values of one or more variables along this
circle will lead to inferior planning.
Fig. 1. Circular fashion in open pit production scheduling (Wang, Sevim 1995)
Rys. 1. Metoda ko³owa w planowaniu produkcji odkrywkowej (Wang, Sevim 1995)
In fact this is a multi-variable optimization problem that requires simultaneous solution.
Unfortunately, such a solution is not easy to obtain and after three decades of continuing
efforts, the problem of production planning in open pit mining as a whole is still an unsolved
problem.
The way that is followed in practice to overcome this problem consists of partially
solution of the problem for one or two parameters by fixing the values of the other
parameters. Usually, the UPL with maximum value is obtained without considering time
value of the blocks by using a graph theory based algorithm. Once the UPL is determined,
different mining sequences are fit within the limits, so that, a series of nested pits is generated
within the predetermined UPL by use of the parameterization technique and the series is then
searched for a mining sequence that would satisfy the production or cutoff grade targets and
highest possible economic value.
There are two fundamental problems with this approach. The first one is that the cost of
mining a specific block is a function of the production capacity. Therefore, one must assume
a production rate in order to be able to calculate block values. The second problem is in the
determination of the UPL. The UPL, which is the final shape of the pit, must be a natural
145
outcome of the optimum mine sequence which in this approach, however, mine sequences
are forced to fit in the UPL that is determined a priori. Clearly, the UPL and the mining
sequence should not be determined independent of each other if optimization is desired.
In this paper, firstly the problem of long term production planning of an open pit mine
from mathematical point of view will be defined. Then the applicability of different meta-
heuristic algorithms will be discussed.
1. Problem statement
Mixed-integer linear programming can be effectively used to model the production
scheduling problem of an open pit mine. It can be defined as follows (Caccetta et al. 1998):
Max C i X i (1)
Subject to:
A i X i £ bi i = 1,2,... , T (2)
EX 1 £ 0
EX 1 + EX 2 £ 0 (3)
EX 1 + EX 2 + ... + EX T £ 0
M T
å å xntm £1 "n (4)
m =1 t =1
m
x nt =0 or 1 "n, "m, "t
where
T – number of periods over which the mine is being scheduled, N – total number of
blocks in the ore body, M – number of different possible types that a block may be mined as
(in this case only two different types will be considered; m = 1 if the block is mined as ore and
m = 2 if the block is mined as waste), K = number of mining and milling constraints for
a given period, R = the number of overlaying restricting blocks, Xt = column vector of the
m m m
(M.N) variables x nt (x nt = 1 if block n is mined as type m in time period t and if x nt =0
m
otherwise), Ct = row vector of (M.N) objective function coefficients c nt corresponding to
m m
the variables x nt (c nt represents the profited or cost return from mining the corresponding
block), At = a K by (M.N) matrix of mining and milling constraint coefficients for time period
t, bt = a K element column vector of RHS coefficients for the mining and milling constraints
and E = an (N.R.M) by (M.N) matrix (0, 1, –1) coefficients for the sequencing constraints.
146
The mining and milling constraints (equation 2) are the capacity constraints. The sequencing
constraints (equation 3) ensure that a block can only be removed if all overlaying blocks have
been removed in the previous or current periods. The reserve constraints (equation 4) are ne-
cessary mathematically to ensure that a block is mined only once.
Though simply stated, an integer linear programming formulation of the scheduling
problem usually involves a large number of variables and constraints. For example, a small
deposit containing 2000 blocks inside the UPL and 6 planning periods would require the
solution of an integer programming problem with 24,000 variables, 12 mining and milling
constraints, 312,000 sequencing constraints and 2000 reserve constraints. Clearly this is
beyond the capacity of current commercial integer programming packages. Several appro-
aches have been proposed in literature to solve this model.
Dagdelen and Johnson (1986) used lagrangian parameterization in order to relax mining
and milling constraints into objective function. Consequently the problem could be handled
by repetition of any UPL algorithm such as Lerchs-Grossmann (1965) graph theory based
algorithm. Caccetta et al. (1998) used the fact that mining and milling constraints are
relatively few in number but complicate the underlying structure of the problem. They
utilized lagrange multipliers to omit these constraints and solve the model using subgradient
optimization method in next step. Later Ramazan et al. (2005) described the application of
fundamental tree algorithm to reconstruct the mining blocks and decrease the number of
variables in scheduling problem without reducing the resolution of the model or optimality of
the results. A fundamental tree is defined as any combination of blocks such that the blocks
can be profitably mined respecting slope constraints.
2. Metaheuristic algorithms
Many of the problems arising in engineering applications such as production planning of
an open pit are NP-hard, that means, they can not be solved to optimality within polynomially
bounded computation time. Hence, to practically solve large instances one often has to use
approximation methods which return near-optimal solutions in a relatively short time.
Algorithms of this type loosely called heuristics. They often use some problem-specific
knowledge to either build or improve solutions.
Recently the researches have focused in a new class of heuristic algorithms, called
metaheuristics. A metaheuristic is a set of algorithmic concepts that can be used to define
heuristic methods applicable to a wide set of difficult problems. These concepts are usually
inspired from the biology and the nature. The use of metaheuristics has significantly
increased the ability of finding very high quality solutions to hard combinatorial problems
(that are often easy to state but very difficult to solve) in a reasonable time. This is
particularly true for large and poorly understood problems. The family of the metaheuristics
includes, but not limited to, genetic algorithm, simulated annealing, tabu search, ant colony
optimization, and particle swarm optimization.
147
2.1. G e n e t i c a l g o r i t h m ( G A )
Genetic algorithm is a search procedure that mimics the operation of genetics and natural
selection. It begins the search with a population of random solutions and evolves this
population over a series of generations by applying probability techniques and reproduction
operators to each member of the population. Reproduction operation consists of two main
steps known as crossover and mutation. The crossover operator combines selected pairs of
the solutions to produce new and potentially better solutions whereas the mutation operator
provides the possible diversity in the population. Survival possibility of solutions for next
generations depends on their quality (fitness value).
Denby and Schofield (1994) demonstrated the fundamental operation of a genetic
algorithm in optimization of an open pit mine production planning. The main advantage of
their method was its ability to solve UPL and long-term planning problems simultaneously.
By choosing proper values for genetic parameters, that was capable to get good results in an
acceptable time. On the other hand, the method was suffering from the fact that the results
were not reproducible because of stochastic nature of the algorithm. The procedure has been
summarized in Figure 2. The major steps of GA approach are:
2.1.1. Chromosome representation of the pits
The shape of any feasible (respecting slope angles) pit in a given block model could be
represented by means of a simple array of integer numbers. Each element of this array shows
the depth of pit geometry in a desired column of block model. Therefore for a block model
sizing Inum × Jnum × Knum , the length of any pit chromosome will be Inum × Jnum where the pit
depth at grid location of (i,j) will be stored in the array at position (Inum × i) + j. Similarly a gi-
ven long-term schedule of an open pit mine could be represented by a longer array considering
the depth of mine at each column of block model for every production period separately.
2.1.2. Initial population
The process starts with the generation of a population of random feasible schedules. The
size of the population is one of the controllable parameters in the system that is set to 50 on
the basis of experience. It is envisaged that the intelligent selection of feasible schedules may
improve the efficiency of the technique.
2.1.3. Assessment of pit fitness
The calculation of fitness is fundamental to the success of a genetic algorithm. A fitness
value is calculated for each of the schedules in the population that is the net present value
(NPV) of the schedule. Authors have reported that the introduction of highly complex fitness
functions has little or no effect on the overall efficiency of the system.
148
Fig. 2. Process of open pit production planning by means of genetic algorithm (Denby, Schofield 1996)
Rys. 2. Proces planowania produkcji odkrywkowej za pomoc¹ algorytmu genetycznego
(Denby, Schofield 1996)
2.1.4. Reproduction of pit population
During the reproduction phase a new generation is created and individual schedules either
survive to the next generation or are removed altogether. In this process schedules with high
fitness values have a greater chance of surviving than those with lower fitness values. The
reproduction phase is critical in which it must ensure that sufficient genetic diversity is
maintained from generation to generation as well as convergence to an optimum result is
sufficiently rapid by allowing good schedules to reproduce faster than bad schedules.
During crossover approximately 70% of schedules are randomly combined in pairs on
a probabilistic basis. This will result in the crossed pairs having modified schedule cha-
racteristics some of which will have higher fitness values, improving their chances of
survival into future generations, whereas some will have lower fitness values, reducing their
chances of survival.
Mutation is also performed on a probabilistic basis on approximately 0.1% of the cells in
the schedule randomly to maintain genetic diversity and prevents the system from con-
verging to a false optimum.
2.1.5. Normalizing of the pits
Both crossover and mutation can cause the resulting pits to violate scheduling constraints
and after each such operation the produced pits must be normalized. This entails modification
of the schedule as little as possible, to ensure that the extraction constraints, such as the
149
number of cells in each scheduling period or the sequencing constraints, are not broken. It
can sometimes result in significant modifications to a schedule.
2.1.6. Termination condition of the algorithm
The number of generations required to reach the optimum schedule varies depending on
the complexity and the scale of the problem. However, the way in which the genetic
algorithm is formulated has a significant effect on the efficiency of the system.
2.2. S i m u l a t e d a n n e a l i n g ( S A )
Annealing is the process in which a metallic or glass solid is heated up sufficiently to
allow the atoms and molecules to rearrange themselves into a stress-free state (but not so
much as to cause melting) and then cooled gradually so as to lock the new configuration into
place. In simulated annealing method, the value of the function being optimized is analogous
to the energy of the solid. It begins with a random solution and then modifies that solution
slightly to create a new potential solution. If the new solution satisfies the constraints and has
a better objective function value than the existing solution, it is accepted without question.
However, if the new solution has a worse value, then the decision on its acceptance will be
made on the basis of current temperature of the system. This allows the process to get out of
potentially sub-optimal solutions. The temperature of the system is lowered gradually until
the optimum solution has been found.
Critical factors to the success of a simulated annealing process include the initial tem-
perature and the cooling rate. If the starting temperature is too low, the process will converge
too quickly and may produce a sub-optimal solution. On the other hand too high initial
temperature might cause spending a long time on poor solutions. Similarly very quick cooling
of the system potentially gets locked around a local-optimum solution and produces a sub-
-optimal result. Clearly too slow cooling rate increases the computation time unnecessarily.
Kumral and Dowd (2005) investigated solving of the open pit mine production sche-
duling problem by using of SA metaheuristic. The main advantage of this routine is that it
utilizes a multi-objective function comprised three minimization components, on the other
hand, the separate determination of UPL and production schedule would be counted as
a disadvantageous for this method.
2.2.1. Objective function
The goal of optimization is to minimize weighted sum of three cost components in-
cluding:
— Cost1: cost of deviation from required tonnage,
— Cost2: penalty and opportunity cost for each content variable,
— Cost3: cost of the content variability for each content variable.
150
Cost1 represents the penalty costs (such as stock piling cost) related to the mine pro-
duction out of the required tonnage boundary. Cost2 is considered if the average metal
content of a period exceeds from the desired limits. It could be either a kind of penalty cost
that is induced by industrial, operational, qualitative or environmental reasons or the oppor-
tunity costs such as loosing of quality requirements if a very high-quality ore is extracted in
the beginning periods. Cost3 refers to the costs corresponding to the decreasing of efficiency
in consequence of non-uniform production. These components transformed into a single
objective problem using a weighting scheme. The selection of weights (priorities) depends
on the ore body, sales contract, ore market structure and plant characteristics.
2.2.2. Initial solution
It has been experienced that the computational time to converge towards a good
solution could be very long using a random initial solution. Kumral and Dowd (2005) used
lagrangian parameterization method suggested by Dagdelen (1985) to obtain a proper
initial solution.
2.2.3. Perturbation mechanism
During each iteration of the algorithm the current solution is perturbed by transferring
a random number of blocks to the next or the previous scheduling periods randomly. Any
transition that causes the access constraints or the ore-waste ratio to be exceeded is
rejected.
2.2.4. Cooling schedule
The possibility of accepting a perturbed solution could be expressed as:
T å i
-0. 005 200 w ´DCost i (5)
P =e
where
T – the temperature of system,
wi – the priority weights of the ith elements of the objective function (Costi).
Kumral and Dowd (2005) proposed the value of 1.67 as initial temperature. Every time
a move is accepted the system cools according to the function T ¬ T/(1 + bT) and for each
rejected move the system is heated by the function T ¬ T/(1 + aT). Where a is
accepted as a coefficient between 0.80–0.95 and the value of b is selected in a
way that the ratio of b/a to be kept equal to the ratio of rejected solutions number to
that of accepted solutions. This ratio is also used as the stopping (termination) criterion.
151
2.3. A n t c o l o n y o p t i m i z a t i o n
The field of “ant algorithms” studies derived from the observation of real ants’ behavior,
and uses these models as a source of inspiration for the design of novel algorithms for the so-
lution of optimization problems. One of the most successful examples of ant algorithms is
known as ant colony optimization (ACO) which is inspired by the foraging behavior of ant co-
lonies. In nature a foraging ant deposits a chemical known as pheromone on the ground which
increases the probability that other ants will follow the same path. ACO has been successfully
implemented on several combinatorial optimization problems such as traveling salesman,
sequential ordering, general assignment, multiple knapsack and network routing problems.
A study has conducted in the institute of surface mining and drilling technology, RWTH
Aachen University, aiming the application of ACO for optimization of long-term open pit
mine production planning. The process has the ability to optimize UPL and long-term
planning problems simultaneously according to a multi-objective target by utilizing a po-
pulation of mine schedule solutions.
In the same way as chromosome construction in GA approach, any given feasible
schedule of an open pit mine could be represented as an array of integer numbers (schedule
array). Accordingly n number of integers (corresponding to the mine level at n planning
periods) is considered for each vertical column of the block model (Fig. 3). The algorithm
also consists of saving n number of variables (ti) for each block of the model
corresponding to the n desired planning periods. ti reveals that how often the block has been
belonged to the ith planning period so far. The higher value of ti the higher possibility of
falling this block in the ith planning period again.
Fig. 3. Array representation of the production schedule of an open pit mine for ACO
Rys. 3. Prezentacja tablicowa harmonogramu produkcji kopalni odkrywkowej dla ACO
2.3.1. Pheromone initialization
Experiments showed that the convergence time is very long when the algorithm starts
from a uniform or random initial pheromone values; therefore a scheduling solution is
created by means of Lerchs-Grossmann’s algorithm (1965) in combination with the Wang,
152
Sevim’s heuristic method (1995). Then a relatively higher value of pheromone (ti) is
assigned to the lowest blocks of in the ith planning period in each column.
2.3.2. Solution construction
Algorithm begins by determination and elimination of the blocks that are outside of the
modeling. Then k number of artificial ants travel through the schedule array to construct
a population of scheduling solutions. This is done by selecting the integer numbers corres-
ponding to the pit depths in every planning period. Accordingly the depth of UPL is selected
first for each column by choosing an integer number between 0 and maximum possible pit
depth on that location. The possibility of selecting the block i as UPL depth in a column of
block model by kth ant is calculated by the following formula:
[ t1 ] a [h i ]b (6)
Pik =
å [ t1 ]a [h i ]b
where
t1 – the pheromone value related to the last period (i.e. UPL),
h – the heuristic information of the block (the economic value of the block in this
case) and a and b are the priority constants.
In each step, after determination of UPL by kth ant, the other nested pits are also found
out similarly employing different t values related to each period. Eventually k
numbers of mine schedules are obtained whose objective function value could be
computed.
2.3.3. Pheromone update
After all the ants have constructed their schedules, the pheromone trails are updated. This
is done by first lowering the pheromone value in all blocks by a constant factor (pheromone
evaporation), and then adding pheromone in the blocks that the ants have considered them
inside a specific period.
An improvement (known as elitist strategy) is possible on the above mentioned simple ant
system, by providing a strong extra reinforcement (additional pheromone) to the blocks
belonging to the best schedule found since the start of the algorithm; this schedule is denoted
as Sbs (best-so-far schedule). Another improvement can be achieved by letting only the best
ranked ants to deposit pheromone.
2.4. O t h e r m e t a h e u r i s t i c s
153
The world of metaheuristics is rich and multifaceted and, besides the following approa-
ches, a number of other successful metaheuristics are available in the literature. Tabu search
and particle swarm optimization are the best known and most widely applied among them,
which their application in the field open pit mine optimization could be studied.
Tabu search uses a neighborhood search procedure to iteratively move from a solution to
another in the neighborhood, until some stopping criterion has been satisfied. To explore
regions of the search space that would be left unexplored by the simple local search
procedure, tabu search modifies the neighborhood structure of each solution as the search
progresses by using a tabu list. The list contains the solutions that have been visited in the
recent past. Algorithm prevents certain moves towards the tabu solutions list.
Particle swarm optimization (PSO) is another stochastic, population-based evolutionary
algorithm which has inspired from social-psychological principles of swarms. In order to
solve a problem by means of PSO a population of individuals (particles) is defined randomly.
Then a communication structure (social network) is defined through assigning neighbors for
each individual to interact with. An iterative process is employed to improve these candidate
solutions by iteratively evaluation of their fitness and remembering the location where they
had their best success. Each particle makes this information available to their neighbors and
they are also able to see where their neighbors have had success. Movements through the
search space are guided by these successes and the population converges to better solutions in
each iteration.
Conclusion
The paper discusses the application of metaheuristic algorithms in optimization of the
long-term production planning of open pit mines. It concludes that only a few numbers of
researches have been addressed so far in this field, in spite of high potential of these
algorithms in solving large and complex combinatorial optimization problems. At first,
employing of genetic algorithm and simulated annealing methods which were already
investigated and reported in the literature, have been briefly surveyed. Then the application
of another metaheuristic (Ant Colony Optimization) has been proposed. This approach has
the ability to solve a multi-objective mine planning and UPL determination problems
simultaneously. Finally scanning of some other metaheuristics revealed the potential benefits
of the research in this field.
REFERENCES
[1] C a c c e t t a L., K e l s e y P., G i a n n i n i L., 1998 – Open pit mine production scheduling. Third regional
APCOM symposium, Kalgoorlie, 65–72.
154
[2] D a g d e l e n K., J o h n s o n T.B., 1986 – Optimum open pit production scheduling by lagrangian pa-
rameterization. 19th APCOM Symposium, 127–141.
[3] D e n b y B., S c h o f i e l d D., 1994 – Open-Pit Design and Scheduling by Use of Genetic Algorithms. Trans.
IMM (Section A: Mining Industry), Vol. 103, A21–A26.
[4] D e n b y B., S c h o f i e l d D., 1995a – The Use of Genetic Algorithms in Underground Mine Scheduling.
Proceeding of APCOM XXV, 389–394.
[5] D e n b y B., S c h o f i e l d D., 1995b – Genetic algorithm for pit scheduling – Extension into 3-dimentions.
Proceeding of mine planning and equipment selection (MPES) conference, 177–185.
[6] D o r i g o M., S t t u e t z l e T., 2004 – Ant Colony Optimization. A Bradford Book (The MIT Press).
[7] G l o v e r F., L a g u n a M., 1997 – Tabu search. Kluwert academic publishers.
[8] G l o v e r F., K o c h e n b e r g e r G.A., 2003 – Handbook of metaheuristics. Kluwert academic publishers.
[9] K u m r a l M., D o w d P.A., 2005 – A simulated annealing approach to mine production scheduling. Journal
of operational reseach society vol. 56, 922–930.
[10] L e r c h s H., G r o s s m a n n I.F., 1965 – Optimum design of open-pit mines. Transactions of CIM, 17–24.
[11] O s a n l o o M., G h o l a m n e j a d J., K a r i m i B., 2008 – Long-term open pit mine production planning:
A review of models and algorithms – Int. J. Mine reclamation and environment vol. 22, 3–35.
[12] R a m a z a n S., D a g d e l e n K., J o h n s o n T.B., 2005 – Fundamental tree algorithm in optimising pro-
duction scheduling for open pit mine design. Trans IMM (Section A: Mining Industry) vol. 114, A45–A54.
[13] S e v i m H, L e i D.D., 1998 – Problem of production planning in open pit mines. INFOR J. vol. 36, 1–12.
[14] T h o m a s G.S., 1996 – Optimization and scheduling of open pits via genetic algorithm and simulated
annealing. First International Symposium on Mine Simulation via the Internet.
[15] W a n g Q., S e v i m H., 1995 – Alternative to parameterization in finding a series of maximum-metal pits for
production planning. Mining Engineering, 178–182.
PERSPEKTYWY METOD OPTYMALIZACJI METAHEURYSTYCZNEJ W PLANOWANIU PRODUKCJI ODKRYWKOWEJ
S³owa kluczowe
Algorytm genetyczny, populacja pocz¹tkowa, metody optymalizacji metaheurystycznej, produkcja odkryw-
kowa [red.]
Streszczenie
Z matematycznego punktu widzenia, problem planowania produkcji odkrywkowej jest bardzo du¿ym i z³o-
¿onym modelem. WyraŸna definicja masy rud zwykle wymaga ujêcia w modelu kilku milionów bloków. Niepewne
iloœci parametrów ekonomicznych i operacyjnych, takie jak przychody i koszty eksploatacji, szacowane klasy
i geomechaniczne w³aœciwoœci ska³ sprawiaj¹, ¿e model optymalizacji staje siê prawie niemo¿liwy do pe³nego
rozwi¹zania za pomoc¹ tradycyjnych metod programowania matematycznego. W niniejszym opracowaniu omó-
wiono mo¿liwoœæ zastosowania metod optymalizacji metaheurystycznej, takich jak Algorytm Genetyczny, Od-
prê¿anie Symulowane, optymalizacja Kolonii Mrówek, optymalizacja Roju Cz¹stek i poszukiwanie Tabu w opty-
malizacji planowania kopalni odkrywkowej. Metody, które s¹ powszechnie stosowane w sztucznej inteligencji
i innych dziedzinach in¿ynierii maj¹ potencja³ równoczesnego rozwi¹zywania bardzo du¿ych i z³o¿onych modeli,
pomimo du¿ego czasu obliczeñ, czemu mo¿na zaradziæ poprzez odpowiedni rozwój sprzêtu komputerowego
w bliskiej przysz³oœci.
PERSPECTIVE OF METAHEURISTIC OPTIMIZATION METHODS IN OPEN PIT PRODUCTION PLANNING
155
Key words
Genetic algorithm, initial population, metaheuristic optimization methods, open pit production [red.]
Abstract
From mathematical point of view, the problem of open pit production planning is a very large and complex
model to solve. Clear definition of ore bodies, normally needs that several millions of blocks to be included in the
block model. Uncertain amounts of economic and operational parameters like revenues and costs of mining,
estimated grades and geomechanical properties of rocks makes the optimization model almost impossible to be
solved totally using traditional mathematic programming methods. In this paper applicability of the metaheuristic
optimization methods such as Genetic Algorithm, Simulated Annealing, Ant Colony optimization, Particle Swarm
optimization and Tabu search in optimization of open pit mine planning will be discussed. These methods that are
widely used in artificial intelligence and other engineering fields have potential of simultaneous solution of very
large and complex models, despite their higher calculation time which could be rectify with computer hardware
developments in near future.
