M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
663–669
EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR
THE SURFACE FINISH ON A MAGNESIUM AZ91D ALLOY
USING AN ANN AND A GENETIC ALGORITHM
RAZVOJ OPTIMIZACIJE PROCESA @I^NE EROZIJE ZA KON^NO
POVR[INSKO OBDELAVO MAGNEZIJEVE ZLITINE AZ91D Z UPORABO
UMETNE NEVRONSKE MRE@E IN GENETSKEGA ALGORITMA
Madhesan Pradeepkumar
1*
, Thangaraj Jesudas
2
, Chandrasekaran Sasikumar
3
,
Mani Narasimharajan
4
1
Department of Mechanical Engineering, Sona College of Technology, Salem, Tamil Nadu, 636005, India
2
Department of Mechatronics, Mahendra Engineering College, Namakkal, India
3
Department of Mechanical Engineering, Bannari Amman Institute of Technology Sathyamangalam - 638 401, India
4
Department of Mechanical Engineering, Mahendra Institute of Technology, Namakkal, India
Prejem rokopisa – received: 2024-02-12; sprejem za objavo – accepted for publication: 2024-09-05
doi:10.17222/mit.2024.1107
In this research the optimizations of the wire EDM process parameters to achieve a minimal surface roughness on a magnesium
AZ91D alloy have been carried out. The experiments were conducted with three machining factors, i.e., the pulse-on time, the
pulse-off time, and the wire feed, using a Box-Behnken design of experiment. The effects of the Artificial Neural Network
(ANN) and the Response Surface Methodology (RSM) models were compared and studied, and it has been found that the ANN
approach predicts the perfect output response. The genetic algorithm (GA) was then utilized to determine the best machining pa-
rameters to provide a better surface finish using the projected ANN outcomes, which were then used to build a quadratic equa-
tion. Furthermore, the optimum machining parameters were identified for a better surface finish through the integration of the
ANN and GA approach. Based on the aforementioned findings, this study showed that the suggested methods are capable of
predicting the optimum machining parameters, which would be beneficial in the low-cost manufacturing sector.
Keywords: ANN, genetic algorithm, magnesium alloy, wire EDM, optimization
V ~lanku avtorji opisujejo raziskavo optimizacije procesnih parametrov `i~ne erozije (EDM; angl.: Electrical Discharge Ma-
chining), da bi dosegli najmanj{o mo`no hrapavost povr{ine na Mg zlitini tipa AZ91D. Avtorji so, na osnovi Box-Behnkenovega
eksperimentalnega dizajna, izvajali preizkuse pri treh izbranih faktorjih mehanske obdelave in sicer: ~asu v`iga elektri~nega
impulza, ~asu izklopa elektri~nega impulza in hitrosti dodajanja `ice. Med seboj so primerjali in {tudirali u~inke modelov
umetne nevronske mre`e (ANN; angl.: Artificial Neural Network) in metodologije povr{inskega odgovora (RSM; angl.: Re-
sponse Surface Methodology). Ugotovili so, da z ANN modeliranjem lahko napovejo mnogo bolj{e izhodne odgovore kot z
RSM modelom. Genetski algoritm (GA) so uporabili za dolo~itev najbolj{ih procesnih parametrov erozije z uporabo dobljenih
ANN podatkov, ki so jih nato uporabili za izdelavo kvadratne ene~be. Nadalje pa so avtorji ugotovili, da so optimalne procesne
parametre mehanske obdelave z `i~no erozijo za najbolj{o kakovost povr{ine dobili z integracijo ANN in GA pristopa. Na
osnovi v ~lanku navedenih ugotovitev so avtorji sposobni napovedati optimalne procesne parametre EDM, ki bi lahko bili
cenovno zelo ugodni tudi za druge uporabnike mehanske obdelave materialov z postopkom `i~ne erozije.
Klju~ne besede: umetne nevronske mre`e, genetski algoritm, zlitina na osnovi magnezija, `i~na erozija, optimizacija
1 INTRODUCTION
Magnesium alloys have exceptional strength-to-
weight ratios and are extensively utilized in industries; a
number of aircraft and automobile industries are focus-
ing on products made with magnesium alloys.
1
Wire
EDM is an un-conventional thermo-electrical machining
process where the work materials are machined by a con-
tinuous discrete spark among the work and the tool elec-
trode while they are immersed in a liquefied dielectric
medium. The continuous discrete spark erodes the work
piece in complex shapes in accordance with a computer-
ized, numerically controlled program.
2
The evaluation of
the quality of a component, the surface finish, is a neces-
sary element for production industries. Surface rough-
ness is an important performance characteristic for the
EDM process, and it became a factor in the quality and
economy characteristics. The operator’s expertise and the
manufacturer’s guidelines play a big role in choosing the
right machining parameters for the wire EDM. As a re-
sult, for newer materials, the machining parameters in
wire EDM must be optimized using experimental tech-
niques. The main objective of the wire EDM manufac-
turers and users is to succeed in establishing better sta-
bility and higher productivity of the wire EDM process.
3
Dewangan et al.
4
studied the influence of pulse cur-
rent, pulse-on time, tool work time, and tool lift time in
relation to the surface integrity on the EDM of AISI P20
tool steel using the RSM and the TOPSIS methods. The
study reports that the optimal solution was obtained
Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669 663
UDK 546.46:621.9.048 ISSN 1580-2949
Original scientific article/Izvirni znanstveni ~lanek MTAEC9, 58(5)663(2024)
*Corresponding author's e-mail:
pradeepkumar@sonatech.ac.in (Madhesan Pradeepkumar)

based on five decision makers’ preferences for different
process parameters. Vinoth and Pradeep
5
have studied
the conventional EDM and the cryogenic EDM processes
on the machining of an aluminum metal matrix compos-
ite and they reported that the discharge current, pulse-on
time, and gap voltage have a significant effect on the
electrode wear and surface roughness. Kavimani et al.
6
have studied the impact of the machining parameters on
wire electrical discharge machining performance in mag-
nesium composites, finding that the MRR and surface
roughness are factors to consider. Gangadharudu et al.
7
studied a multi-response optimization with the use of the
principal component analysis based on the grey tech-
nique to find the optimum process parameters for a
metal-matrix composite on the EDM process; they dis-
covered a setting of the process parameters through evo-
lutionary optimization techniques. Murahari et al.
8
have
evaluated the effects of the EDM process parameters in
terms of the output response material removal rate, sur-
face roughness, tool wear rate, and recast layer thickness
on the machining of Ti-6Al-4V using an analysis of the
variance and the F-test. Tripathy et al.
9
studied the ma-
chining parameter optimization for multiple responses on
the numeral process variables on the EDM of the H-11
die steel using the TOPSIS and the grey relational analy-
sis and they investigated the effect of process variables
like powder concentration, peak current, pulse-on time,
duty cycle, and gap voltage based on the response of the
material removal rate, tool wear rate, electrode wear ra-
tio, and surface roughness. Rai et al.
10
investigated the
cutting forces and the shear plane temperature in end
milling with the use of an ANN. They used 15 neurons
in the feedforward neural network with a backpropa-
gation algorithm. In terms of the cutting tool, the geo-
metrical parameters, the cutting parameters, and the
work piece material properties, as well as the three com-
ponents of the cutting forces and the shear plane
temperature were used as the output layers. The study
demonstrated that the ANN model shows a close agree-
ment between the experimental and the numerical re-
sults. Vishal et al.
11
studied the optimization of the pro-
cess parameters in terms of kerf width in the EDM
process on a stainless-steel 304L. They used the Taguchi
L32 orthogonal design of experiment to conduct experi-
ments and analysis of the variance, and the study discov-
ered that the mathematical model hasa4%error in com-
parison to the experimental results. Yýldýzel et al.
12
have
investigated the modulus of rapture values for glass fiber
reinforced in a concrete block with the use of an ANN
integrated with an artificial bee colony and they con-
cluded that the model provides most appropriate perfor-
mance compared with multiple linear regression.
The literature survey of this research has helped in
furthering a number of researchers on the process param-
eters optimization by the use of the RSM, and ANN
tools for different materials; however the researches on
the EDM process of advanced materials like magnesium
AZ91D alloy were rarely conducted. In addition, the
conventional optimization techniques were preferred in
the research over the evolutionary techniques. In this re-
search work we focus on a prediction of best machining
parameters to attain minimal surface roughness on wire
EDM of a Mg AZ91D alloy through the integration of
two evolutionary optimization techniques of the RSM
technique, the ANN methodology, and the GA methodol-
ogy.
2 EXPERIMENTAL PART
In this paper the study was conducted on a wire EDM
on a modern material, i.e., a magnesium AZ91D alloy,
and the effective RSM Box-Behnken experimental
M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
664 Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669
Figure 1: Proposed investigational methodology

design
13
was chosen to conduct the experiments.
14
Box-Behnken designs are a robust and efficient design
for a quadratic model for optimization by avoiding un-
necessary combinations of experiments like corner
points and star points of the design. The purpose of the
identical experiments in Box-Bhenken design is to iden-
tify the proper experimentation and minimize bias from
uncontrolled variables. The results were analyzed as well
as compared with the ANN methodology,
15
and the ANN
resulting model was employed to produce a quadratic
equation, which was then used as the fitness function for
a GA to forecast the optimum machining parameters for
an excellent surface quality. The proposed investigational
methodology is shown in Figure 1.
2.1 Experimental Process
The wire cut EDM process was conducted on a mag-
nesium alloy with use of a CNC Wire EDM machine
ELECTRONICA-ELCOT EL-10 VGA. The machining
parameters of pulse-on time (T
on
), pulse-off time off
(T
off
), and wire feed (W
f
) were set at the three levels as
shown in Table 1. The cutting process was performed
with the use of a copper wire of 0.25 mm diameter. A to-
tal of 17 trials were conducted using the interaction of
the three machining parameters at the lower, medium,
and higher levels in accordance with the recommenda-
tions of the Box-Behnken design of experiment. The re-
sults of the output response surface-roughness measure-
ment using the Mitutoyo Surftest 211 equipment are
shown in Table 2.
Table 1: Machining parameters and their levels
Wire-EDM-
Machining
Parameters
Unit
Levels
Lower (–1) Medium (0) Higher (1)
T
on
μs 4 7 10
Toff
μs 4 7 10
Wf
mm/min 1 2 3
Table 2: Box-Behnken Design of Experiments
Run
T
on
(μs)
T
off
(μs)
W
f
(mm/min)
Experiment
R
a (μm)
1 –1 –1 0 1.25
2 1 –1 0 1.04
3– 1101 . 4
4110 0.74
5 –1 0 –1 1.19
6 1 0 –1 0.81
7 –1 0 1 1.13
8101 0.71
9 0 –1 –1 0.89
10 0 1 –1 1.15
11 0 –1 1 1.12
12 0 1 1 0.69
13 0 0 0 0.84
14 0 0 0 0.89
15 0 0 0 0.91
16 0 0 0 0.87
17 0 0 0 0.90
2.2 Response surface methodology (RSM)
An experimental modeling method known as the
RSM is used to establish the relationship among the re-
sponses and the control variables
16
. To determine the ef-
fects of wire feed (W
f
), pulse-on time (T
on
), and pulse-off
time (T
off
) on the surface finish (R
a
), the present research
employs RSM.
The response
Y =  N, f, ap Y =  (T
on
, T
off
,W
f
) (1)
R
a
=  (T
on
, T
off
,W
f
) (2)
The present study’s second-order RSM model is pro-
vided by:
YXX XX
ii
i
k
ij i j
ij
k
ii i
i
k
=+ + +
===
∑∑∑
 
0
1
2
1
(3)
n are the interacting coefficient terms.
17
2.3 Artificial Neural Network (ANN)
ANNs are based on biological nervous systems. They
consist of an enormous number of interconnected factors
called neurons. The interactions between the neurons and
the process information are stored in the brain by regula-
tion among the neurons; as a result, the neurons form a
network which mimics a biological nervous machine.
18
The input layer, hidden layer, and output layer are the
three levels that make up the ANN computational
method. Each layer contains a distinct type of neuron,
and all the layers are connected in succeeding layers.
19
In this wire EDM operation, the ANN tool was uti-
lized to identify the optimum machining parameters for
obtaining a minimal surface roughness. A 3-8-1 neural
network structure was developed by using the three ma-
chining parameters T
on
, T
off
, and W
f
as the input neurons
(input layer) and one hidden layer of eight neurons to
generate one output neuron as surface roughness (output
layer). The experimental data was used as the training
data for the ANN model. The experiment was conducted
with the training and testing sample datasets of 15 each.
A random training data set was utilized to train the neu-
ral network, and a backpropagation technique was em-
ployed to train the feedforward network. The backpropa-
gation algorithm is based on the gradient-descent
method, and it uses iterations of the weights to update
the goal and training mean square error values. The hid-
den-layer and output-layer activation functions were set
to logsig and tansig, respectively to achievie the surface
roughness output.
20
In MATLAB, learning functions
were assigned to traingdx and learngd, respectively, for
the ANN training function.
2.4 Genetic Algorithm (GA)
The use of evolutionary biology principles is utilized
to address problems using the evolutionary optimization
technique known as Genetic Algorithm (GA), and it in-
M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669 665

cludes genetic inheritance, natural selection, mutation,
and crossover.
21
Starting from a set of chromosomes or
capability solutions that are randomly selected from the
character of bit strings, a GA optimal solution is gener-
ated. As a result, a population is formed by a whole set
of these chromosomes, and the chromosomes change
across many iterations or generations; as a result, new
generations are created using the crossover and mutation
approach. The chromosomes are further evaluated using
certain fitness condition standards and the most appropri-
ate chromosomes are stored, while the others are dis-
carded. This system repeats till one chromosome ascends
to high-quality fitness, whereupon it is assumed to be the
most appropriate solution to the problem.
22
The GA has
been successfully used in an extensive variety of prob-
lems due to its simplicity, ease of procedure, and global
perception. The ANN resultant quadratic model Equa-
tion (4) was utilized for the objective function, the popu-
lation size was set at 200, and the crossover probability
was at 0.8, the mutation rate was distributed uniformly at
0.1, the stall generation was set at 100, and the stopping
criteria were best fit. The GA was conducted to mini-
mize the objective function, and the ranges of the input
parameters are as shown in Table 3.
Table 3: Parameter range for GA optimization
Parameter Range for GA optimization
Ton
1μs T
on
 10 μs
Toff
1μs T
on
 10 μs
Wf
1 mm/min W
f
 5 mm/min
Objective function as Minimize: R
a
(T
on
, T
off
, W
f
)
3 RESULTS AND DISCUSSION
3.1 Parameter Effects on Surface Roughness
The analysis of the values of the machining parame-
ters in the wire EDM process, developed by the use of
the RSM method, is presented in Figure 2. It demon-
strates that as the value of T
on
increases the surface
roughness decreases linearly, predicting that a small
change in T
on
would severely affect the surface rough-
ness. In terms of T
off
, a lower rate of T
off
produces a
higher surface roughness; moreover, when the T
off
in-
creases the roughness value decreases until the middle
level. Thereafter, it slightly increases until the higher
level. In terms of W
f
, the figure depicts that increasing
W
f
provides a greater surface roughness till level 2 (2
mm/min); thereafter, the roughness value decreases until
the higher level of W
f
.
The analysis of the variance (ANOVA) was derived
on the basis of the RSM Box-Behnken experimental de-
sign and is presented in Table 4. The primary objective
of the ANOVA is to pinpoint cutting parameters that in-
fluence thc characteristics of quality. According to
ANOVA Table 4, the model contributed 99.50 percent of
the total variance, whereas the errors percentage contri-
bution was only 0.65 percent. This indicated that the ex-
periment was carried out with the proper machining pa-
rameters. Therefore, it can be observed that in the
first-order interactions T
on
has a larger percentage
(52.05%) contribution to the surface roughness while the
remaining parameters T
off
and W
f
have a significantly
smaller contribution of 1.911% and 2.83%, respectively.
Table 4: Analysis of variance
Source DF SS MS F-Value
Contribu-
tion
T
on
1 0.348613 0.026042 41.86 52.05%
T
off
1 0.012800 0.000165 0.26 1.911%
W
f
1 0.019012 0.039296 63.16 2.83%
Ton
*Ton
1 0.056963 0.052346 84.14 8.50%
Toff
*Toff
1 0.053188 0.054720 87.95 7.94%
W
f
*W
f
1 0.053188 0.004725 7.60 0.70%
T
on
*T
off
1 0.050625 0.050625 81.37 7.55%
Ton
*Wf 1 0.000400 0.000400 0.64 0.05%
Toff
*Wf 1 0.119025 0.119025 191.31 17.77%
Error 7 0.004355 0.000622 0.65%
Total 16 0.669706
Furthermore, the analysis suggests that a small varia-
tion in the T
on
factor would affect the surface-finish char-
acteristics on the magnesium material. In view of the
second order, T
on
has a higher influence of 8.50 % com-
pared to the other factors; moreover, Table 4 presents
that the T
on
and T
off
has an effect on the surface rough-
ness. The regression value for this operation was
99.35 %, which says that the Box-Behnken design of the
experiment was the best technique to design for evaluat-
ing the parameters in this wire EDM process. The
ANOVA suggests that T
on
has a larger contribution in the
wire cut EDM operation in comparison to the other pa-
rameters.
For better visibility and understanding of the effects
of machining parameters on the output response surface
roughness, the 3D surface plots in Figure 3 were cre-
ated. Figure 3a indicates that a lower rate of T
on
and a
higher rate of T
off
produce a higher surface roughness of
1.3 μm on this wire EDM process; furthermore, higher
M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
666 Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669
Figure 2: Main effect plots for surface roughness

rates of both T
on
and T
off
would yield a better surface fin-
ish. Figure 3b displays the contribution of T
on
and W
f
in
the wire EDM process. It shows that both the factors at
lower levels give a higher surface roughness; conse-
quently, higher levels of both the parameters yield a
better surface finish. In terms of the interaction of T
off
and W
f
, the 3D surface plot was drawn, see Figure 3c.
The figure shows that when the EDM process is con-
ducted with a lower level of T
off
in correspondence to a
higher level of W
f
, a higher surface roughness on the ma-
terial is produced. Similarly, a lower level of W
f
with a
higher level of T
off
also produces a higher roughness on
the surface of the magnesium alloy. However, higher lev-
els of both the parameters gave a better surface finish of
0.69 μm in this EDM process on the magnesium alloy.
The RSM evaluation reveals the relationship between
input parameters and the output response in terms of a
quadratic model and the 3D surface plots indicate that
T
on
and W
f
have the most impact on the output response
when considering the surface roughness. The RSM pre-
dicted values are closer to the experimental values at a
higher level of T
on
at 10 μs, a middle level of T
off
at 7 μs,
and a higher level of W
f
at 3 mm/min provide a better
surface finish of approximately 0.69 μm and displayed in
Table 5.
R
a
= 1.18947 – (0.1488*T
on
) + (0.0118*T
off
)+
+ (0.5110*W
f
) + (0.0123*T
on
*T
on
)+
+ (0.01267*T
off
*T
off
) – (0.0335*W
f
*W
f
)–
– (0.0125*T
on
*T
off
) – (0.00333*T
on
*W
f
)–
– (0.0575*T
off
*W
f
) (4)
3.2 Optimization
Once the proposed neural network model has under-
gone 1000 training iterations, the predicted surface
roughness results were generated. An analysis of the net-
work reveals a 0.99 correlation coefficient, meaning that
there is a strong relationship between the experimental
and projected values for this ANN model. Figure 4 dis-
plays the results of the ANN model’s evaluation of the
network’s performance, which is based on the correlation
coefficient between the output and target values for the
test data. A well-trained network can accurately predict
surface roughness values, as shown by the ANN model’s
average relative error between the experimental and pre-
dicted values of 1.31 percent.
An error % line chart of the predicted surface rough-
ness values using RSM and ANN approaches is shown in
Figure 5. A comparison between the chart and actual
values shows that the ANN model generates a smaller er-
ror than the RSM model. The ANN model’s predicted
values are closer to the experimental values; conse-
quently, it provides a better surface finish value of
0.71 μm with the optimal combination of the machining
parameters in the higher level of T
on
at 10 μs, the middle
level of T
off
at 7 μs, and the higher level of W
f
at
3 mm/min. In this optimization study, the ANN model
M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669 667
Figure 4: Artificial neural network - regression model
Figure 3: 3D Surface plot for surface roughness

provides accurate values of the output response in com-
parison to the experimental values; thereby proving its
ability to train the model more efficiently.
The GA optimization was conducted based on the ge-
netic parameters stated above with integration of results
obtained from ANN; moreover, the best fit was attained
for the given machining parameters ranges. Figure 6 de-
picts that the better the surface roughness has achieved a
value of 0.4820 μm with the optimized values of the cut-
ting parameters are at the higher levels with T
on
at 10 μs,
T
off
at 10 μs, and W
f
at 3 mm/min. The outcomes of the
confirmation test, which was carried out in accordance
with the suggestions of the ANN and GA procedures, are
shown in Table 5. It is noted that both the ANN and GA
model provide a lower error percentage; moreover, both
approaches work better as optimization tools for machin-
ing parameter optimization for modern materials.
Table 5: Evaluation of predicted model
Machining Parameters Surface Roughness R
a
(μm)
Error
% T
on
T
off
W
f
RSM
Pre-
dicted
ANN
Pre-
dicted
GA
Pre-
dicted
Confir-
mation
Test
10 7 3 0.691 0.73 5.47
10 7 3 0.711 0.74 3.91
10 10 3 0.4820 0.50 3.60
4 CONCLUSIONS
The application of evolutionary techniques to predict
optimized cutting parameters in the wire EDM process
on a modern material, i.e., the magnesium alloy AZ91D,
is the focus of this research. Through the Box-Behnken
experimental methodology, the RSM quadratic model
has been developed to predict and examine analytical
models for attaining better cutting parameters on achiev-
ing a minimal surface roughness. Using the evolutionary
method of the ANN and GA approaches, the following
conclusions were drawn:
• The RSM main effect plot suggests a minimum sur-
face roughness of 0.69 μm with the desired parameter
levels; also, the Box-Behnken experimental design
yields a predicted regression (R
2
) value of 96.36, in-
dicating the Box-Behnken design to be the most effi-
cient way of designing an experiment. The ANOVA
shows that the T
on
has a bigger contribution of 52.32
percent in this experiment; additionally, it shows that
even the smallest adjustment in this parameter affects
the component’s surface finish.
• When compared to t heexperimental values, the ANN
model projected values are more accurate, resulting
in a better surface finish value of 0.71 μm with the
optimal combination of machining parameters of T
on
at 10 μs, T
off
at7μsandW
f
at 3 mm/min. The RSM
and ANN methodologies had an average error of
1.38 % and 1.31 %, respectively, when compared to
the experimental results. The experiment shows that
the ANN model predicts the optimal output response
better than the RSM model.
• The evolutionary GA suggests that the minimal sur-
face roughness of 0.48 μm can be achieved with the
optimized cutting parameters of T
on
at 10 μs T
off
at
10 μs, and W
f
at 3 mm/min. In this study, the inte-
grated GA model provides a better output response in
terms of optimization of the machining parameters
with minimal surface roughness in comparison to the
RSM and the ANN models.
5. REFERENCES
1
G. Li, L. Zhou, S. Luo, F. Dong, N. Guo, Microstructure and me-
chanical properties of bobbin tool friction stir welded ZK60 magne-
sium alloy. Materials Science and Engineering: A, 776 (2020),
doi:10.3390/ma15010023
2
E. M. Rubio, M. Villeta, D. Carou, A. Saa, Comparative analysis of
sustainable cooling systems in intermittent turning of magnesium
pieces, International Journal of Precision Engineering Manufac-
turing, 15 (2014) 5, 929–940, doi:10.1007/s12541-014-0419-5
3
J Ashvin, J. I. Makadia, Nanavati, Optimisation of machining param-
eters for turning operations based on RSM, Measurement, 46 (2013),
1521–1529, doi:10.1016/j.measurement.2012.11.026
M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
668 Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669
Figure 6: Best fitness of surface roughness on GA
Figure 5: Error percentage chart of experimental vs. predicted values

4
S. Dewangan, S. Gangopadhyay, C. K. Biswas, Study of surface in-
tegrity and dimensional accuracy in EDM using Fuzzy TOPSIS and
sensitivity analysis, Measurement, 63 (2015), 364–376, doi:10.1016/
j.measurement.2014.11.025
5
S. Vinoth Kumar, M. Pradeep Kumar, Machining process parameter
and surface integrity in conventional EDM and cryogenic EDM of
Al–SiCp MMC, Journal of Manufacturing Processes, 20 (2015),
70–78, doi:10.1016/j.jmapro.2015.07.007
6
V. Kavimani, K. S. Prakash, T. Thankachan, Influence of machining
parameters on wire electrical discharge machining performance of
reduced graphene oxide/magnesium composite and its surface integ-
rity characteristics. Compos. Part B Eng, 167(2019) 621–630,
doi:10.1016/j.compositesb.2019.03.031
7
Gangadharudu Talla, Deepak Kumar Sahoo, S. Gangopadhyay, C.K.
Biswas, Modeling and multi-objective optimization of powder mixed
electric discharge machining process of aluminum/alumina metal
matrix composite, Engineering Science and Technology, an Interna-
tional Journal, 18(2015), 369–373, doi:10.1016/j.jestch.2015.01.007
8
Murahari Kolli, Adepu Kumar, Effect of dielectric fluid with
surfactant and graphite powder on Electrical Discharge Machining of
titanium alloy using Taguchi method, Engineering Science and Tech-
nology, an International Journal, 18 (2015), 524–535, doi:10.1016/
j.jestch.2015.03.009
9
S. Tripathy, D. K. Tripathy, Multi-attribute optimization of machin-
ing process parameters in powder mixed electro-discharge machining
using TOPSIS and grey relational analysis, Engineering Science and
Technology, an International Journal, 19 (2016), 62–70, doi:10.1016/
j.jestch.2015.07.010
10
J. K. Rai, L. Villedieu, P. Xirouchakis, Mill-cut: a neural network
system for the prediction of thermo-mechanical loads induced in
end-milling operations, International Journal of Advanced Manufac-
turing Technology, 37 (2008), 3–4, 256– 264, doi:10.1007/s00170-
007-0973-4
11
Vishal Parashar, A. Rehman, J. L. Bhagoria, Y. M. Puri, Kerf width
analysis for wire cut electrical discharge machining for SS 304L us-
ing design of experiment, Indian Journal of Science and Technology,
3( 2010) 4, 369–373
12
S.A. Yýldýzel, A.U. Öztürk, A study on the estimation of prefabri-
cated glass fiber reinforced concrete panel strength values with an
Artificial Neural Network Model, CMC-Computers Materials &
Continua, 52 (2016) 1, 41–52
13
H. Akkus, Optimising the effect of cutting parameters on the average
surface roughness in a turning process with the taguchi method, Ma-
ter. Tehnol,.52 (2018) 6, 781–785, doi:10.17222/mit.2018.110
14
H. Raymond, D. C. Myers, C. M. Montgomery, Anderson-Cook. Re-
sponse Surface Methodology: Process and Product Optimization
Using Designed Experiments, John Wiley & Sons, 2016
15
K. M. Hamdia, X. Zhuang, T. Rabczuk, An efficient optimization ap-
proach for designing machine learning models based on genetic algo-
rithm, Neural Computing and Applications, 33(6) (2021), 1923–335,
doi:10.1007/s00521-020-05035-x
16
D. C. Montgomery, Design and analysis of experiments, John Wiley
and sons, New York, 1991.
17
S. Jayachandran, M. Raman, T. Ramasamy, Experimental investiga-
tion for the optimization of the WEDM process parameters to obtain
the minimum surface roughness of the AL 7075 aluminium alloy em-
ployed with a zinc-coated wire using RSM and GA, Mater. Tehnol,
53 (2019) 3, 349–356, doi:10.17222/mit.2018.166
18
M. Pradeep Kumar, R. Venkatesan, V. Kaviarasan, Evaluation of sur-
face integrity in milling of magnesium alloy using artificial neural
network and genetic algorithm, Mater. Tehnol, 52 (2018) 3, 367–373,
doi:10.17222/mit.2017.198
19
N. E. Karkalos, N. I. Galanis, A. P. Markopoulos, Surface roughness
prediction for the milling of Ti-6Al-4V ELI alloy with the use of sta-
tistical and soft computing techniques, Measurement, 90 (2016)
25–35. doi:10.1016/j.measurement.2016.04.039
20
A. Srinivasan, K.G. saravanan, R. Viswanathan and S.K.karthikayan,
Optimization and effect of dielectric fluid with Zr and Ni on electri-
cal discharge machining of die steel material, Chemical Industry &
Chemical Engineering Quarterly, 29(4) (2023), 311–318,
10.2298/CICEQ221215005A
21
P. Suresh, R. Venkatesan, T. Sekar, N. Elango, Optimization of Inter-
vening Variables in MicroEDM of SS 316L using a Genetic Algo-
rithm and Response-Surface Methodology, Strojni{ki vestnik – Jour-
nal of Mechanical Engineering, 60 (10) (2014), 656–664,
doi:10.5545/sv-jme.2014.1665
22
K. Deb, P. C. Roy, R. Hussein, Surrogate Modeling Approaches for
Multiobjective Optimization: Methods, Taxonomy, and Results,
Math. Comput. Appl, 26 (2021) 5, doi:10.3390/mca26010005
M. PRADEEPKUMAR et al.: EVOLUTIONARY OPTIMIZATION OF WIRE EDM PROCESS FOR THE SURFACE FINISH ...
Materiali in tehnologije / Materials and technology 58 (2024) 5, 663–669 669