FACTA UNIVERSITATIS

Series: Mechanical Engineering Vol. 11, No 1, 2013, pp. 29 - 44

COMPARISON OF META-HEURISTIC ALGORITHMS FOR

SOLVING MACHINING OPTIMIZATION PROBLEMS 

UDC 519.863; 621.7.01

Miloš Madić, Danijel Marković, Miroslav Radovanović

Universityof Niš, Faculty of Mechanical Engineering, Serbia

Abstract. Since optimization of the machining parameters not only increases machining
efficiency and economics, but it also enhances the end product quality, this topic is still
the subject of many studies. The selection of the optimal machining parameters is often
performed in a two-stage approach, i.e. mathematical modeling of machining performance
and optimization using an optimization method. Among the traditional optimization
methods, in recent years, the modern meta-heuristic algorithms are being increasingly
applied to solving machining optimization problems. Their ability to deal with complex,
multi-dimensional and ill-behaved optimization problems has made them choice number
one by most researchers and practitioners. In the present study, an attempt is made to
compare the optimization results of different meta-heuristic algorithms applied to solving
machining optimization problems. Four meta-heuristic algorithms are taken into
consideration, namely, real coded genetic algorithm (RCGA), simulated annealing
(SA), improved harmony search algorithm (IHSA) and cuckoo search algorithm (CSA).
These meta-heuristic algorithms are applied to searching for optimal combinations of
different machining parameters for five case studies taken from the literature. The
optimization results obtained by applying RCGA, SA, IHSA and CSA for parametric
optimization of these machining processes are compared with those derived by the past
researchers.
Key Words: Machining, Optimization, Meta-heuristic Algorithms

1. INTRODUCTION
In today's manufacturing world, it is a vital task to define optimal machining parameters
for achieving machining cost and efficiency [1]. Modeling of machining processes aimed at
better understanding, optimization and process control is very important in manufacturing
practice. This is usually achieved by integrating empirical models based on the regression

Received January 10, 2013




Acknowledgments. The paper is a part of the research done within the project TR35034. The authors would
like to thank to the Ministry of Education and Science, Republic of Serbia.
Corresponding author: Miloš Madić
University of Niš, Faculty of Mechanical Engineering, Str. Aleksandra Medvedeva 14, 18000 Niš, Serbia
E-mail: [email protected]
30 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

analysis or artificial neural networks with an optimization method. The application of the
Taguchi method without formulating any kind of model is also an attractive alternative, par-
ticularly in the case of multi-objective optimization problems [2].
Identification of the optimal machining parameters is very important for reduction of
machining costs, product quality improvement and increased productivity and profit. There-
fore, the machining processes optimization is one of the most investigated issues. Thus,
many researchers have tried various conventional methods for solving machining optimiza-
tion problems including design of experiments, graphical methods, analytic methods and
mathematical programming methods such as linear and nonlinear programming, dynamic
programming, geometric programming, goal programming, integer programming, stochastic
programming, etc. [3]. Despite the fact that the machining optimization problems have been
extensively investigated, the complexity of machining economics problems has led to the re-
quirement for increasingly effective optimization algorithms [4].
The convergence speed of the meta-heuristic algorithms to the global (or nearly
global) optimal results is better than that of traditional techniques. Therefore, the meta-
heuristic algorithms have been increasingly used to further improve the solution of ma-
chining optimization problems with complex nature in many applications [1]. It is re-
ported that the meta-heuristic algorithms have been applied in machining because of their
ability to deal with highly complex, non-linear, and multi-dimensional machining optimi-
zation problems [5]. In the current trend of optimizing machining process parameters,
various evolutionary or meta-heuristic algorithms have been used [6].
The most recent research of the meta-heuristic algorithms for machining process pa-
rameters optimization have been demonstrated by Zain et al. [7], Rao et al. [8], Samanta
and Chakraborty [9], Madić et al. [10], Yildiz [4, 1, 11], Ciurana et al. [12], Pansare and
Kavade [13], Liu et al. [Liu]. A comprehensive review paper regarding machining pa-
rameters optimization by means of the meta-heuristic algorithms is presented by Yusup et
al. [6]. As has been reported in the literature, three types of meta-heuristic-based search
algorithms viz. genetic algorithm (GA), simulated annealing (SA) and particle swarm op-
timization (PSO) have been mostly applied in the domain of the machining parameters
optimization. However, in recent years there is an increasing trend in the application of
other meta-heuristic algorithms such as ant colony optimization (ACO), artificial bee col-
ony (ABC), improved harmony search algorithm (IHSA), and cuckoo search algorithm
(CSA) for solving machining optimization problems.
This paper aims at comparing the performance of the GA, SA, IHSA and CSA when
applied to different machining optimization problems. After a brief discussion of the basic
methodology, characteristics and principles of the meta-heuristic algorithms in general,
five machining optimization case studies are considered. The obtained optimization re-
sults of the GA, SA, IHSA and CSA are compared and discussed. The optimization re-
sults are also compared with those derived by the earlier researchers.

2. META-HEURISTIC ALGORITHMS
The algorithms used for solving optimization problems can be very diverse, from con-
ventional algorithms to modern meta-heuristic algorithms [14]. The optimization algo-
rithms developed so far can be broadly classified into deterministic and stochastic. The
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 31

main difference between deterministic and stochastic algorithms is that in stochastic

methods, the points that do not strictly improve the objective function can also be created
and take part in the search process [15]. Most conventional or classic algorithms are de-
terministic. Some deterministic optimization algorithms use gradient information; they are
called gradient-based algorithms (such as the Newton-Raphson algorithm). These algo-
rithms use the function values and their derivatives and find a greater use in solving
smooth unimodal problems. Gradient-free/derivative-free algorithms do not use any de-
rivative, but only the function values. Hooke-Jeeves pattern search and Nelder-Mead
downhill simplex are examples of gradient-free algorithms [14]. Within stochastic algo-
rithms, there are heuristic and meta-heuristic algorithms. Loosely speaking, heuristic
means "to find" or "to discover by trial and error". Quality solutions to a tough optimiza-
tion problem can be found in a reasonable amount of time, but there is no guarantee that
optimal solutions are reached [14]. While the heuristic algorithms resemble trial and error
mechanisms, and depend on computational capacity, the meta-heuristic algorithms tend to
learn as they run, and tend to be more intelligent and adaptive [16].
The term meta-heuristics was introduced by Glover [17] and represents a class of
promising algorithms for solving hard optimization problems. The meta-heuristic algorithms
are aiming at efficient and comprehensive exploration of the search space, using the
governing mechanisms which imitate certain strategies taken from nature, social behavior,
physical laws, etc., in order to find near optimal solutions. Some popular global optimization
algorithms include: GA, SA, PSO, ACO, IHSA, CSA, artificial bee colony (ABC), taboo
search (TS), artificial immune system (AIS), teaching-learning-based optimization (TLBO)
algorithm, gravitational search algorithm (GSA), shuffled frog leaping (SFL), scatter search
(SS), firefly algorithm (FA), etc. Besides these well known algorithms, the investigations on
the meta-heuristic algorithms are still being done and new algorithms are being developed
continually. The past 20 years have witnessed the development of numerous meta-heuristic
algorithms in various communities that are at the intersection of several fields, including
artificial intelligence, computational intelligence and soft computing [18].
Optimization based on using meta-heuristic algorithms starts with an initial set of in-
dependent variables and then evolves to obtain the global minimum/maximum of the ob-
jective (fitness) function. The objective function is a mathematical model (function) that
assigns a value to each solution in the search space. Starting from an initial solution built
by some heuristic, meta-heuristics improve it iteratively until a stopping criterion is met.
The stopping criterion can be elapsed time, number of iterations, etc. The operation of
meta-heuristic algorithms works in such a way to determine the final solution, only some
existing solutions are actually visited. The search is conducted under a process that is spe-
cific to each meta-heuristic algorithm, but it is a way that attempts to intelligently find
good solutions. However, there is no guarantee that the solution returned by a meta-heu-
ristic is the best [19]. A universal step by step optimization procedure for any type of
meta-heuristic algorithm can be described as follows [20]:
 Initializing the population in the search domain by seeding the population with
random values.
 Evaluating the fitness for each individual of the population.
 Generating a new population by reproducing selected individuals through
evolutionary operations, such as crossover, mutation, and so on.
 Looping to step 2 until stopping criteria are satisfied.
32 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

All the meta-heuristic algorithms use certain trade-off of local search and global ex-
ploration. A variety of solutions is often realized via randomization which provides a
good way of moving away from the local search to that on the global scale [14].
The two main concepts of any meta-heuristic algorithms are: intensification and diversi-
fication, or exploitation and exploration [21]. Diversification means to generate diverse so-
lutions so as to explore the search space on the global scale, while intensification means to
focus on the search in a local region by exploiting the information that a current good solu-
tion is found in this region. This is in combination with the selection of the best solutions [14].
The fine balance between these two components is very important to the overall
efficiency and performance of an algorithm. Too little exploration and too much ex-
ploitation could cause the system to be trapped in local optima, which makes it very
difficult or even impossible to find the global optimum. On the other hand, in the
case of too much exploration but too little exploitation, it may be difficult for the
system to converge and thus it slows down the overall search performance [14]. The
main difference between developed meta-heuristic algorithms is in the means by
which they try to achieve this balance.
Meta-heuristic algorithms can be divided into two categories: single-solution meta-
heuristic algorithms where a single solution (and search trajectory) is considered at a time
and population meta-heuristic algorithms where a multiplicity of solutions evolve concur-
rently. Within each category, it is also possible to distinguish between primarily construc-
tive meta-heuristic algorithms, where a solution is built from scratch (through the intro-
duction of new elements at each iteration) and improvement meta-heuristic algorithms
which iteratively modify a solution [22].
The main idea behind designing the meta-heuristic algorithms is to tackle complex
optimization problems where other optimization methods have failed to be effective.
These methods are now recognized as some of the most practical approaches for solving
many real world problems [18]. There are several advantages of using meta-heuristic al-
gorithms for optimization, namely:
 Broad applicability: they can be applied to any problems that can be formulated as
function optimization problems.
 Hybridization: they can be combined with more traditional optimization techniques.
 Ease of implementation: typically easier to understand and implement.
 Efficiency and flexibility: they can solve problems larger problems faster. Moreover,
they are simple to design and implement, and are very flexible [18].
 The use of meta-heuristics can be justified due to: (i) complexity of the internal
problem that prevents the application of exact techniques and (ii) a very large
quantity of possible solutions that prevent the use of exhaustive algorithms [23]. [23].
However, there are some disadvantages of the meta-heuristic algorithms that should be
here noted:
 In general, the optimization performance is highly dependent on fine parameter tuning.
 They do not have "sound" mathematical foundation, when compared to more
traditional techniques [24].
 They cannot prove optimality.
 They cannot probably reduce the search space.
 Repeatability of optimization results obtained with the same initial condition
settings is not guaranteed.
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 33

3. APPLICATION OF META-HEURISTIC ALGORITHMS FOR SOLVING MACHINING

OPTIMIZATION PROBLEMS
In this section, the performance of the selected meta-heuristic algorithms for the
optimization of real engineering optimization problems will be compared. Reviewing the
literature one can see that meta-heuristic algorithms have been successfully applied to a
wide range of optimization problems. For illustrative purposes, this section will thus focus
on a particular class of problems, namely machining optimization problems, as they
exhibit an impressive record of successful implementations. This paper aims at comparing
the optimization results of the real-coded GA (RCGA), SA, IHSA and CSA for different
machining optimization problems. A brief description of these algorithms is provided in
the following subsections.

3.1. Real coded genetic algorithms

Genetic algorithms, developed by Holland [25], are artificial genetic systems based on
the process of natural selection. They are a particular class of evolutionary algorithms that
use techniques inspired by evolutionary biology such as inheritance, mutation, selection
and crossover. The evolution of population is performed through a specific number of
generations where the next generation gives a better solution than the previous one. In
RCGA, the solution is directly represented as a vector of real parameter decision vari-
ables; thus, the representation of the solutions is very close to the natural formulation of
many problems. The use of real-parameters makes it possible to use large domains for
variables. The RCGA have been used to solve engineering problems that are complex and
difficult to solve by conventional optimization methods. Implementation of the RCGA re-
quires the determination of six fundamental issues: chromosome representation, selection
function, the genetic operators, initialization, termination and evaluation function.

3.2. Simulated annealing

Initially presented by Kirkpatrick et al. [26], SA is a random search technique for
global optimization problems able to escape local optima. The concept of simulated an-
nealing is taken from nature and it mimics the metals recrystallization in the process of
annealing. Annealing is the slow cooling of metal that produces good low energy state
crystallization, whereas fast cooling produces poor crystallization. SA uses single point
search method. It is a memoryless search algorithm in the sense that no information is
saved from previous searches [23]. SA algorithm starts with an random initial design
vector (solution) Xi and high temperature T. A second design point is created at random in
the vicinity of the initial point and the difference in the function values ∆E at these two
points is calculated as [27]:
E  f  fi 1  fi  f ( X i 1 )  f ( X i ) (1)

If the new solution's objective function value is smaller, the new solution is automati-
cally accepted and becomes the current solution from which the search will continue.
Otherwise the point is accepted with a probability e(−∆E/kT) where k is the Boltzmann's con-
stant. This completes one iteration of the SA. Due to the probabilistic acceptance of a
nonimproving solution, SA can escape from local optima. At a certain temperature T, a
34 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

predetermined number of new points are tested. The algorithm is terminated when current
value of temperature is small enough or when changes in function values (f) are suffi-
ciently small.

3.3. Improved harmony search algorithm

Harmony search algorithm (HSA), developed by Geem [28] has been successfully ap-
plied to various benchmark and real world problems. It is a meta-heuristic optimization
algorithm conceptualized by using the musical process of searching for a perfect state of
harmony. Musical performances seek to find pleasing harmony (a perfect state) as deter-
mined by an aesthetic standard, just as the optimization process seeks to find a global so-
lution (a perfect state) as determined by an objective function. Optimizations procedure of
the HSA includes five steps [29]. The algorithm requires several parameters [26], includ-
ing harmony memory (HM), number of improvisations (NI), harmony memory consider-
ing rate (HMCR), pitch adjusting rate (PAR), bandwidth vector (bw). Mahdavi [29] sug-
gested an improvement of the traditional HSA with the key difference in the way of ad-
justing PAR and bw. Namely, to improve the performance of the HSA and eliminate the
drawbacks that originate from fixed values of PAR and bw, the improvement of the tradi-
tional HSA uses variables PAR and bw in the improvisation step. In this paper, the im-
proved harmony search algorithm (IHSA) is used.

3.4. Cuckoo search algorithm

Cuckoo search algorithm (CSA) is a novel population based stochastic global search
meta-heuristic algorithm developed by Yang and Deb [30]. CSA is inspired by natural
mechanisms and mimics, the breeding behavior of some cuckoo species that lay their eggs
in the nests of host birds. Each egg represents a solution, and a cuckoo egg represents a
new solution. The goal is to use new and potentially improved solutions (cuckoos) to
replace worse solutions in the nests. CSA can be briefly described using the following
three idealized rules [30]:
 Each cuckoo lays one egg at a time, and dumps it in a randomly chosen nest.
 The best nests with high quality of eggs (solutions) will carry over to the next generations.
 The number of available host nests is fixed, and a host can discover an alien egg
with a probability Pa [0, 1].

3.5. Parameter settings of the RCGA, SA, IHSA and CSA

Due to the unique functionality of each type of the meta-heuristic algorithm, the com-
parison of meta-heuristic algorithms is in many ways more difficult than other algorithmic
comparisons [31, 32]. As it is known, each type of the meta-heuristic algorithm has a
number of parameters that must be set before the algorithm execution. These parameters
are vital components of an algorithm and can be changed to alter the performance of the
algorithm. Although some general guidelines about selection of these parameters exist in
relevant literature, it has been widely reported that the optimal setting is strongly related
to the design problem under consideration. In that sense, it is decided to use the most
common parameter settings for each of the meta-heuristic algorithms. The list of the pa-
rameter settings used for each meta-heuristic algorithm applied is given in Table 1.
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 35

As meta-heuristic algorithms have stochastic nature, each run will usually pro-
duce different results. Hence, to reduce randomness, each optimization problems has
been attempted 20 times with different seeds and the best solution for each algo-
rithm is recorded.

Table 1 Meta-heuristics and their parameter settings used

Population size = 20
Number of generations = 100
Selection function = Stochastic uniform
Reproduction: Elite count: 2; Crossover rate: 0.8
RCGA
Mutation: constraint dependent default
Crossover function: Scattered
Migration: Migration direction: forward; Migration interval: 20;
Migration fraction: 0.2.
Annealing function: Fast annealing
Reannealing interval: 100
Temperature update function: Exponential temperature update
SA
Initial temperature: 100
Acceptance probability function: Simulated annealing acceptance
Start point: set such that all variables take lower bound values
Harmony memory size: 10
Harmony memory consideration rate: 0.95
Minimum pitch adjusting rate: 0.1
IHSA Maximum pitch adjusting rate: 0.85
Minimum bandwidth: 0.001
Maximum bandwidth: 0.8
Number of improvisations: 50000
Number of nests: 20
CSA
Discovery rate : 0.25

3.6. Machining optimization problem formulation

In machining optimization problems the aim is usually to minimize/maximize an
objective function, often representing a machining performance, under some machining
parameter constraints. The problem may be expressed as follows:
Minimize(or maximize) f ( x)
subject to: g i ( X )  0, i  1...m
X lj  X j  X uj , j  1...n
(2)
where, X is the vector of machining parameters; f (x) is the objective function to be
l u
minimized (maximized); gi (X) is the i-th functional constrain; and Xj and Xj are lower
and upper bounds of j-th machining parameter Xj.
36 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

4. CASE STUDIES
The search for papers dealing with machining optimization problems is restricted to
those based on empirical models developed using the regression analysis. The reason for
this is that, since regression models are explicitly expressed, the obtained optimization
results can readily be checked and compared.

4.1. Case study I

Pansare and Kavade [13] present an experimental work to investigate the effects of the
cutting parameters (feed rate, cutting speed and depth of cut) on surface roughness in
turning oil hardened non-shrinkable steel. Based on the experimental results from the Ta-
guchi's L9 orthogonal array, the authors have developed the relationship between the cut-
ting parameters and the surface roughness using multiple linear regression. The regression
equation developed by Pansare and Kavade [13] is as follows:

Ra  8.11  0.0217  A  25.9  B  6.37  C  0.0563  AB  0.0153  AC  19.4  BC (3)

where A is the cutting speed (m/min), B is the feed rate (mm/rev), C is the depth of cut
(mm), and Ra is the surface roughness (m).
In an attempt to obtain optimum turning parameters for minimum surface
roughness value, the authors apply ACO. The turning optimization problem is for-
mulated as follows:

Minimize Ra  f ( A, B, C )
subject to: 150  A  250 (m/min); 0.1  B  0.2  mm/rev  ; 0.5  C  1.5 (mm) (4)

The optimum turning parameter values and surface roughness value obtained by Pan-
sare and Kavade [13] are given in Table 1. The optimization problem formulated in Eq. 4
is attempted using the RCGA, SA, IHSA and CSA with the parameter settings as dis-
cussed before. The best obtained optimization results are summarized in Table 2.

Table 2 Optimization results for case study I

Cutting speed Feed rate Depth of cut Surface roughness
Algorithm
(m/min) (mm/rev) (mm) (m)
ACO* 150 0.1032 1.4677 0.0114
RCGA 150 0.1 1.5 0.093
SA 150.011 0.1 1.5 0.092
IHSA 150 0.1 1.5 0.093
CSA 153.3242 0.1015 1.4391 0.0727
*
Results reported by Pansare and Kavade [13]

Regarding the optimization performance of the selected meta-heuristic algorithms in terms

of the solution quality obtained and the computational time, the following observations are
made: (i) IHSA is most consistent, (ii) RCGA yields scatter results, (iii) SA has the slowest
convergence, (iv) although better than ACO, CSA performs poorer that the other meta-heuristic
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 37

algorithms, (v) although statistically is adequate, in some cases, mathematical model may not
have adequate physical meaning and should be used with care. In such situations, an experi-
mental verification of the obtained optimization results should be carried out.

4.2. Case study II

Shivakoti et al. [33] have carried out turning experiments on mild steel workpiece by
considering the spindle speed and cutting speed at different feed rate values. The ultimate
goal is to find the optimum parameters values for turning operations for maximizing the
material removal rate (MRR). Based on the experimental results, the authors have developed
explicit mathematical model based on the regression analysis. The regression equation
developed for calculation of MRR is as follows:

MRR  1.42  1.83  A  0.9  B  10  C  103  AB  112  AC  0.000014 BC (5)

where A is the feed rate (mm/rev), B is the spindle speed (rpm), and C is the cutting speed
(m/min).
To identify optimal cutting parameters, Eq. 4 is selected as objective function and
solved using GA by formulating the optimization problem as:

Maximize MRR  f ( A, B, C )
subject to: 0.62  A  0.98 (mm/rev); 40  B  1000 (rpm); 3.5  C  95.5 (m/min) (6)

The obtained optimization results by Shivakoti et al. [33] and the optimization results
of the RGCA, SA, IHSA and CSA are compared in Table 3.

Table 3 Optimization results for case study II

Feed rate Spindle speed Cutting speed
Algorithm MRR (mm3/s)
(mm/rev) (rpm) (m/min)
GA* 0.582 891.520 25.580 51247.549
RCGA 0.98 999.926 3.936 99591
SA 0.98 1000 3.5 99691
IHSA 0.98 1000 3.5 99691
CSA 0.9717 949.3741 5.9017 93581
*
Results reported by Shivakoti et al. [33]

Regarding the optimization performance for the second case study, the following
observations are made: (i) again, IHSA produces most consistent solutions, (ii) scattered
nature of the RCGA solutions is evident, (iii) Although SA converges too slowly in
comparison to other algorithms, it yields the best solution.

4.2. Case study III

Mohruni [34] has conducted a machining experiment to measure the surface rough-
ness value in the end milling operation. Based on the data for the real machining experi-
ments, Zain et al. [7] have developed regression models for each cutting tool to predict
surface roughness. The obtained equations are obtained as:
38 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

Runcoated  0.451  0.00267  A  5.671  B  0.0046  C (7a)

RTiA1N  0.292  0.000855  A  5.383  B  0.00553  C (7b)
RSNTR  0.237  0.00175  A  8.693  B  0.00159  C (7c)

where: where A is the cutting speed (m/min), B is the feed rate (mm/tooth), and C is the
radial rake angle (°).
Zain et al. [7] propose Eq. 7c as the fitness function in the GA optimization module.
The minimization of the fitness function value of Eq. 7c is subjected to the boundaries
(limitations) of cutting condition values. The optimization problem is formulated as:

Minimize RSNTR  f ( A, B, C )
subject to: 124.53  A  167.03 (m/min); 0.025  B  0.083  mm/tooth  ; 6.2  C  14.8 ( ) (8)

To determine best optimal milling parameter values, Zain et al. [7] have tried several
combinations of GA parameter rates. The obtained optimization results are given in Table
4. To investigate the efficiency of the PSO algorithm for machining problem solving, the
optimization problem formulated in Eq. 8 is attempted by Pare et al. [35]. Although the
authors report superior results than those obtained by Zain et al. [7], the optimization
results are not valid since the constraint for the feed rate is not satisfied.

Table 4 Optimization results for case study III

Cutting speed Feed rate Radial rake angle Surface Roughness
Algorithm
(m/min) (mm/tooth) (°) (µm)
0.138*
GA* 167.029 0.025 14.769
0.13854154***
**
PSO 124.53 0.0025 6.2 0.0186
- 167.03 0.025 14.8 0.13849225
*
Results reported by Zain et al. [7]
**
Results reported by Pare et al. [35]
***
Results obtained after substituting the obtained parameter values in Eq. 6c

Regarding this case study, it should be noted that application of any classical mathe-
matical or meta-heuristic optimization technique is redundant. Since the mathematical
model is expressed as a linear combination of independent parameters with no interaction
terms, and since there are no other additional non-linear constraints, the optimization so-
lution of the above machining optimization problems is obvious. For this case study, from
Eq. 7c it is seen that surface roughness increases with an increase in feed rate, and de-
creases with an increase in cutting speed and radial rake angle. Therefore, considering
machining constraints, the optimal machining parameters values can be easily determined,
i.e. they represent the boundary points in the experimental space covered. The best opti-
mal machining parameter values and corresponding surface roughness value are given in
Table 4.
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 39

4.2. Case study IV

Saravanakumar et al. [36] have conducted turning experiments on Inconel 718 in or-
der to investigate the influence of machining process parameters such as the cutting
speed, feed rate, and depth of cut on the output parameters such as MRR and surface
roughness. Using the experimental data, the authors have developed regression equations
for material removal rate (MRR) and surface roughness Ra in the following form:

MRR  19158  298  A  112136  B  91493  C  1749  AB  1417  AC  (9)

 537343  BC  7880  ABC

Ra  23.6  0.331  A  110  B  88  C  1.66  AB  1.29  AC  463  BC  6.93  ABC (10)

where A is the cutting speed (m/min), B is the feed rate (mm/rev), and C is the depth of
cut (mm).
To identify optimal cutting parameters, Eq. 9 is to be maximized whereas Eq. 10 is to
be minimized considering the following machining parameter constraints:

60  A  80 (m/min); 0.15  B  0.25  mm/rev  ; 0.1  C  0.25 (mm) (11)

The optimization problem is then selected as an objective function and solved using
GA. The optimization results obtained by Saravanakumar et al. [36] as well using RCGA,
SA, IHSA and CSA are given in Table 5.

Table 5 Optimization results for case study IV

Cutting speed Feed rate Depth of cut Surface MRR
Algorithm
(m/min) (mm/rev) (mm) Roughness (µm) (mm3/min)
79.99 0.25 0.1 - 2122.23
GA*
79.9 0.15 0.1 0.69 -
79.998 0.25 0.1 - 2123. 89
RCGA
79.999 0.15 0.1 0.6896 -
80 0.25 0.1 - 2124.275
SA
79.943 0.15 0.1 0.689956
80 0.25 0.1 - 2124.275
IHSA
80 0.15 0.1 0.689 -
79.9374 0.2493 0.1397 - 2054.9
CSA
78.2857 0.1492 0.1086 0.7871 -
*
Results reported by Saravanakumar et al. [36]

Regarding the optimization performance for this case study, the following observa-
tions are made: (i) SA yields the best solution, (ii) For maximization of MRR, the best
and the most consistent results are obtained by SA and IHSA, (iii) CSA performs poorly
with very diversified solutions, (iv) for surface roughness minimization best optimization
solution as achieved with IHSA.
40 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

4.2. Case study V

Sharma et al. [37] have conducted turning experiments on aluminum 6061 alloy and
metal matrix composites of aluminum. An attempt has been made to establish mathemati-
cal relationships between the cutting parameters and surface roughness based on the re-
gression analysis. Mathematical relationships between the responses and the machining
parameters for Al/SiC (5%) and Al/SiC(10%) composite material are obtained as follows:

Ra Al SiC  5%   18.7  0.00122  A  443  B  10.4  C  0.000001  A2 

(12a)
2541  B 2  4.71  C 2  0.0015  AB  0.00229  AC  40.7  BC

Ra Al SiC 10%   3.61  0.00252  A  14.2  B  13.2  C  0.000002  A2 

(12b)
640  B 2  9.67  C 2  0.0407  AB  0.00212  AC  21.2  BC

where A is the cutting speed (m/min), B is the feed rate (mm/rev), and C is the depth of
cut (mm).
In an attempt to obtain optimum turning parameters for minimum surface roughness
value, Sharma et al. [37] apply PSO considering the following constraints for the
machining parameters:

228  A  740 (mm/min); 0.05  B  0.1 mm/rev  ; 0.4  C  1 (mm) (13)

The optimization results obtained by Sharma et al. [37] as well using RCGA, SA,
IHSA and CSA are given in Table 6.

Table 6 Optimization results for case study V

Cutting speed Feed rate Depth of cut Ra (µm) Ra (µm)
Algorithm
(mm/min) (mm/rev) (mm) Al/SiC (5%) Al/SiC (10%)
1.2883*
233 0.05 0.4 
0.874501**
PSO*
1.558*
228 0.05 0.4 
1.84469***
726.542 0.05 0.4 0.23937 
RCGA
228 0.05 0.4  1.84469
724.214 0.05 0.4 0.2411 
SA
228 0.05 0.4  1.84469
740 0.05 0.4 0.22936 
IHSA
228 0.05 0.4  1.84469
736.2804 0.0534 0.4026 0.9216 
CSA
370.7913 0.0539 0.4880  2.9578
*
Results reported by Sharma et al. [37]
**
Results obtained after substituting the obtained parameter values in Eq. 11a
***
Results obtained after substituting the obtained parameter values in Eq. 11b
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 41

Regarding the optimization performance for this case study, the following observa-
tions are made: (i) IHSA yields the best optimization results followed by RCGA, (ii) For
the first objective functions, SA and CSA perform poorly with very diversified solutions,
(iii) When optimizing second objective function, all algorithms except CSA have proved
efficient and found best solution.

5. CONCLUSION
In this paper, an attempt is made to compare the optimization solutions obtained by
the meta-heuristic algorithms applied to solving machining optimization problems. Four
meta-heuristic algorithms are taken into consideration, namely, RCGA, SA, IHSA and
CSA and applied to solving five machining case studies. On the basis of the obtained
results the following conclusions can be made:
In the case of machining optimization problems IHSA has proved to be the most effi-
cient meta-heuristic algorithm in terms of computational time and solution quality. The
IHSA parameter settings as proposed in this study enable IHSA efficient and fast explo-
ration of the search space without getting trapped in local minima.
 Although SA has the slowest convergence, it can be used efficiently for solving
machining optimization problems. In some cases, it is found to be superior to the
other meta-heuristic algorithms.
 CSA performs much poorer that the other meta-heuristic algorithms indicating a
need for fine tuning of the parameters for solving machining optimization prob-
lems.
 Not intending to be suspicious about the search space exploration capability and
efficiency of other meta-heuristic algorithm, the optimization results indicate that
more effort is needed for efficient implementation of RCGA and CSA to the
solving machining optimization problems.
 An approach presented in [32] based on Taguchi's experimental design has
proved efficient for fine tuning of IHSA parameters as used in this study. There-
fore, in the authors' opinion, it represents an appropriate way for fine tuning of
the RCGA and CSA algorithm parameters so as to improve its performance for
solving machining optimization problems.
 The analysis of the obtained optimization results indicate that, regardless of the
meta-heuristic algorithm applied, there is a need to check the boundary points in
the hyperspace of the machining parameters in order to check the existence of
optimal solution.
 An analysis of the mathematical model and optimization problem formulation is
to be done prior to the application of a meta-heuristic algorithm, since in some
situations it may happen that the optimal conditions can be determined in a much
easier way.
 Finally, it is necessary to emphasize the need for validation of the optimization
solutions obtained using meta-heuristic algorithms.
42 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

REFERENCES
1. Yıldız, A.R., 2009, A novel particle swarm optimization approach for product design and manufacturing,
The Internation Journal of Advanced Manufacturing Technology 40(5). pp. 617–628.
2. Marinković, V., Madić, M., 2011, Optimization of surface roughness in turning alloy steel by using
Taguchi method, Scientific research and Essay, 6 (16), pp. 3474-3484.
3. Madić, M., Radovanović, M., 2010, Savremene metode optimizacije obradnih procesa, IMK-14 Istraživanje
i razvoj, 37(4), pp. 19-24.
4. Yıldız, A.R., 2009, An effective hybrid immune-hill climbing optimization approach for solving design
and manufacturing optimization problems in industry, Journal of Materials Processing echnology 209(6),
p. 2773–2780.
5. Mukherjee, I., Ray, P.K., 2006, A review of optimization techniques in metal cutting processes,
Computers & Industrial Engineering 50(4), pp. 15–34.
6. Yusup, N., Zain, A.M., Hashim, S.Z.M., 2012, Evolutionary techniques in optimizing machining
parameters: Review and recent applications (2007-2011), Expert System with Applications, 39(10), pp.
9909-9927.
7. Zain, A.M., Haron, H., Sharif, S., 2010, Application of GA to optimize cutting conditions for
minimizing surface roughness in end milling machining process, Expert Systems with Applications
37(6), pp. 4650–4659.
8. Rao, R.V., Pawar, P.J., Davim, J.P., 2010, Optimisation of process parameters of mechanical type advanced
machining process using a simulated annealing algorithm, International Journal of Materials and
Product Technology, 37 (1-2), pp. 83-101.
9. Samanta, S., Chakraborty, S., 2011, Parametric optimization of some non-traditional machining processes
using artificial bee colony algorithm, Engineering Applications of Artificial Intelligence 24(6), pp. 946–957.
10. Madić, М, Marković, D, Radovanović, M, 2012, Optimization of surface roughness when turning
polyamide using ANN-IHSA approach, International Journal of Engineering & Technology, 1(4), pp.
432-443.
11. Yildiz, A.R., 2013, Cuckoo search algorithm for the selection of optimal machining parameters in
milling operations, The International Journal of Advanced Manufacturing Technology, 64(1-4), pp. 55-61.
12. Ciurana, J., Arias, G., Özel, T., 2009, Neural network modelling and particle swarm optimization (PSO)
of process parameters in pulsed laser micromachining of hardened AISI H13 steel, Materials and
Manufacturing Processes, 24(3), pp. 358–368.
13. Pansare, V.B., Kavade, M.V., 2012, Optimization of cutting parameters in multipass turning operation
using ant colony algorithm, International Journal of Engineering Science & Advanced Technology, 2(4),
pp. 955 – 960.
14. Yang, X.S., 2011, Review of Metaheuristics and Generalized Evolutionary Walk Algorithm, International
Journal of Bio-Inspired Computation, 3(2), pp. 77-84.
15. Al-Harkan, I.M, Trafalis, T.B., 2002, A hybrid scatter genetic tabu approach for continuous global
optimization, 11-31: In Combinatorial and Global Optimization, P.M. Pardalos, A. Migdalas, R. Burkard,
2002, World Scientific Publishing Co.
16. Manda, K,. Satapathy, S.C., Poornasatyanarayana, B., 2012, Population based meta-heuristic techniques for
solving optimization problems: A selective survey, International Journal of Emerging Technology and
Advanced Engineering, 2(11), pp. 206-211.
17. Glover, F., 1986, Future paths for integer programming and links to artificial intelligence, Computers &
Operations Research, 13(5), pp. 533–549.
18. Gholizadeh, S., Barati, H., 2012, Comprative study of three metaheuristics for optimum design of
trusses, International Journal of Optimization in Civil Engineering, 3, pp. 423-441.
19. Gomes de Freitas, F.G., Brito Maia, C.L., Lima de Campos, G.A., Teixeira de Souza, J., 2010, Optimization
in Software Testing Using Metaheuristics, Revista de Sistemas de Informação da FSMA, 5, pp. 3-13.
20. Khajehzadeh, M., Raihan Taha, M., El-Shafie, A., Eslami, M., 2011, A Survey on Meta-Heuristic
Global Optimization Algorithms, Research Journal of Applied Sciences, Engineering and Technology
3(6), pp. 569-578.
21. Blum, C., Roli, A., 2003, Metaheuristics in combinatorial optimisation: Overview and conceptural
comparision, ACM Computing Surveys, 35(3) 268-308.
 Comparison of Meta-Heuristic Algorithms for Solving Machining Optimization Problems 43

22. Gendreau, M., Potvin, J.Y., 2005, Metaheuristics in Combinatorial Optimization, Annals of Operations
Research, 140(1), pp. 189-213.
23. Talbi, E., 2009, Metaheuristics: From Design to Implementation, Wiley Publishing.
24. Viana, A., Sousa, J.P., Matos, M.M, 2005, Constraint oriented neighborhoods – a new search strategy in
metaheuristics, In: Metaheuristics: Progress as real problem solvers, Edited by: T. Ibakari, K. Nonobe,
M. Yagiura, Springer.
25. Holland, J., 1975, Adaptation in Natural and Artificial Systems, University of Michigan Press, Ann Anbor.
26. Kirkpatrick, S., Gelatt, C., Vecchi, M., 1983, Optimization by Simulated Annealing, Science, 220(4598),
pp. 671-680.
27. Rao, S.S., 1996, Engineering Optimization: Theory and Practice, John Wiley & Sons.
28. Geem, Z.W., Kim, J.H., Loganathan, G.V., 2001, A new heuristic optimization algorithm: harmony
search, Simulation, 76(2), pp. 60–68.
29. Mahdavi, M., Fesanghary, M., Damangir, E., 2007, An improved harmony search algorithm for solving
optimization problems, Applied mathematics and computation, 188(2), pp. 1567–1579.
30. Yang, X.S., Deb, S., 2010, Engineering optimisation by cuckoo search, International Journal of
Mathematical Modelling and Numerical Optimisation 1(4), pp. 330 – 343.
31. Silberholz, J., Golden, B., 2010, Comparison of metaheuristics. In: Gendreau, M., Potvin, J.-V. (eds)
Handbook of Metaheuristics, 2nd edn, Springer, Heidelberg.
32. Marković, D., Madić, M., Petrović, G., 2012, Assessing the performance of improved harmony search
algorithm (IHSA) for the optimization of unconstrained functions using Taguchi experimental design,
Scientific research and essays, 7(12), pp. 1312-1318.
33. Shivakoti, I., Diyaley, S., Kibria, G., Pradhan, B.B., 2012, Analysis of Material Removal Rate using
Genetic Algorithm Approach, International Journal of Scientific & Engineering Research,3(5), pp. 1-6.
34. Mohruni, A.S., 2008, Performance evaluation of uncoated and coated carbide tools when end milling of
titanium alloy using response surface methodology. Thesis for Doctor of Philosophy, Universiti
Teknologi Malaysia, Skudai, Johor, Malaysia.
35. Pare, V., Agnihotri G., Krishna, C.M., 2011, Optimization of Cutting Conditions in End Milling Process
with the Approach of Particle Swarm Optimization, International Journal of Mechanical and Industrial
Engineering, 1(2), pp. 21-25.
36. Saravanakumar, K., Shaik Dawood, A.K., Pratheesh Kumar, M.R., Optimization of CNC Turning
Process Parameters on INCONEL 718 Using Genetic Algorithm, Engineering Science and Technology:
An International Journal, 2(4), pp. 532-537.
37. Sharma, A.V.N.L., Sandeep Kumar, P., Gopichand, A., Mohan Rao, R., 2012, Optimal machining
conditions for turning of Al/SiC MMC using PSO and Regression analysis, International Journal of
Engineering Research and Applications, 2(6), pp. 497-500.
44 M. MADIĆ, D. MARKOVIĆ, M. RADOVANOVIĆ

POREĐENJE META-HEURISTIČKIH ALGORITAMA

ZA REŠAVANJE PROBLEMA OPTIMIZACIJE
PARAMETRA OBRADE
Optimizacija parametara obrade ne utiče samo da efikasnost i ekonomičnost obrade već i na
finalni kvalitet proizvoda, pa samim tim ova tema je još uvek predmet izučavanja mnogih studija.
Izbor optimalnih parametara obrade često se obavlja u dve faze, odnosno matematičkim modeliranjem
performansi obrade i optimizacijom pomoću optimizacionih metoda. Njihova mogućnost da rešavaju
složene i višediemzionalne optimizacione probleme učinila je da postanu veoma popularan izbor od
strane većine istraživača. U ovom radu, autori su uporedili rezultate optimizacije raličitih meta-
heurističkih algoritama koji su primenjeni na rešavanje optimizacionih problema obrade. Razmatrana
su četiri meta-heuristička algoritma i to: realno kodirani genetski algoritam, simulirano kaljenje,
poboljšani algoritam harmonijskog pretraživanja i algoritam kukavice. Pomoću ovih meta-
heurističkih algoritmama su tražene optimalne kombinacije različitih parametra obrade za pet studija
slučaja uzetih iz literature. Rezultati optimizacije, dobijeni pomoću prethodno navedenih pet meta-
heuristička algoritma za parametrsku optimizaciju procesa obrade, su upoređeni sa rezultatima
poslednjih istraživanja.
Ključne reči: mašinska obrada, optimizacija, meta-heuristički algoritmi