M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
367–373
EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING
OF A MAGNESIUM ALLOY USING AN ARTIFICIAL NEURAL
NETWORK AND A GENETIC ALGORITHM
OVREDNOTENJE INTEGRITETE POVR[INE PO MEHANSKI
OBDELAVI MAGNEZIJEVE ZLITINE Z UPORABO UMETNE
NEVRONSKE MRE@E IN GENETSKEGA ALGORITMA
Madhesan Pradeepkumar
1
, Rajamanickam Venkatesan
2
, Varatharajan Kaviarasan
1
1
Sona College of Technology, Department of Mechanical Engineering, Salem, Tamil Nadu, 636005, India
2
Kumaraguru College of Technology, Department of Mechatronics Engineering, Coimbatore, Tamil Nadu, 641049, India
pradeepkumar@sonatech.ac.in
Prejem rokopisa – received: 2017-11-23; sprejem za objavo – accepted for publication: 2018-01-09
doi:10.17222/mit.2017.198
Magnesium alloys are advanced, light materials used widely in industries and milling is one of the material removal processes
that are extensively used. In this present study, the experimental work has been carried out based on a Box-Behnken design by
mainly considering three factors, i.e., cutting speed, feed, and depth of cut. The first part of in this study, the effects of Response
Surface methodology (RSM) and Artificial Neural Network (ANN) models were evaluated and compared. The RSM and ANN
models provide the average error of 2.40 % and 1.52 %, respectively, it recommends that ANN is a more efficient methodology
for the prediction of the optimal output response than RSM. The predicted model has been coupled with evolutionary
optimization technique genetic algorithm (GA) to determine the optimum cutting parameters to attain the minimal surface
roughness with respect to the wide ranges of machining parameters and GA suggest the surface roughness of 1.0926 μm with
the optimized machining parameters. The evaluation of these observations proves that the proposed methods are capable of
determining the optimum machining parameters for modern materials.
Keywords: artificial neural network, genetic algorithm, magnesium alloy, milling, optimization
Zlitine na osnovi magnezija (Mg) so napreden lahek material, ki se pogosto uporablja v razli~nih vejah industrije. Njegova
mehanska obdelava z rezkanjem je ena od najpogostej{ih metod odstranjevanja materiala za dokon~no oblikovanje razli~nih
industrijskih izdelkov. V predstavljeni {tudiji avtorji opisujejo eksperimentalno delo izvedeno na osnovi Box-Behnkenovega
dizajna z upo{tevanjem treh faktorjev: rezalne hitrosti, pomika in globine reza. V prvem delu te {tudije so avtorji ovrednotili in
primerjali u~inke dveh uporabljenih modelov: metodologije reakcije povr{ine (RSM; angl.: Response Surface Methodology) in
umetne nevronske mre`e (ANN, angl: Artificial Neural Network). RSM in ANN modela sta predvidela povpre~no 2,40 % oz.
1,52 % napako. To pomeni, da je ANN u~inkovitej{a methodologija za napoved optimalne (reakcije) storilnosti. Model za
prognozo so avtorji zdru`ili z evolucijsko optimizacijsko tehniko genetskih algoritmov (GA) in na ta na~in dolo~ili optimalne
rezalne parametre za dosego minimalne povr{inske hrapavosti v obmo~ju dokaj {iroko izbranih parametrov mehanske obdelave.
GA napoveduje povr{insko hrapavost 1.0926 μm pri optimiziranih parametrih mehanske obdelave. Ovrednotenje teh opazovanj
je potrdilo da, je s predlaganimi metodami mo`no dolo~iti optimalne parametre mehanske obdelave modernih materialov.
Klju~ne besede: umetna nevronska mre`a, genetski algoritem, zlitina na osnovi Mg, rezkanje, optimizacija
1 INTRODUCTION
Magnesium alloys are new composite materials and
nowadays used widely in industries due to their excellent
strength-to-weight ratio.
1
According to that many aircraft
and automobile industries focus on making products in
magnesium alloys. Surface roughness has become a sig-
nificant factor in terms of quality and economy charac-
teristics.
2
In modern industries, CNC machines are
widely used to produce complicated shapes on advanced
materials with high quality and accuracy. An appropriate
choice of machining parameters for the CNC milling
greatly depends on the operator’s expertise and manu-
facturer guidelines.
3–4
Hence, machining parameters for
newer materials need to be optimized by experimental
methods.
K. Shi et al.
5
have studied the process parameter
effects on the milling of magnesium alloys using Taguchi
methodology and grey rational analysis. They reported
that the feed rate is the most significant factor affecting
the surface integrity and they found the optimum com-
bination of process parameters through the grey rational
technique. They suggest the combination of both
methodologies were an effective method to obtain the
desired surface quality for the milling process. N. Haq et
al.
6
studied the effects of drilling parameters Al-SiC
(10p) MMC on LM25 alloy using grey relational
analysis. Cutting speed, feed rate, and point angle were
taken as the cutting parameters. Surface roughness,
cutting force and torque were determined by experi-
ments. Taguchi’s method of orthogonal array L
9
was
followed to the experimental process. They concluded
that point angle is the most significant machining
Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373 367
UDK 67.017:669.721.5:519.254 ISSN 1580-2949
Original scientific article/Izvirni znanstveni ~lanek MTAEC9, 52(3)367(2018)

parameter for affecting the surface roughness, followed
by cutting speed and feed. A. M. Zain et al.
7
conducted
milling experiments with two-level factorial design and
they studied the surface roughness effects on the milling
process with respect to the high cutting speed, low feed
rate and rake angle of machining parameters using ANN.
J. Kopac et al.
8
predicted the optimum surface roughness
and material removal rate in the milling of an Al alloy
with respect to the flank milling parameters and they ob-
tained an optimal combination in the integration of
Taguchi and Grey relational analysis. J. A. Ghani et al.
9
applied the Taguchi methodology and ANOVA to opti-
mize the cutting parameters, i.e., cutting speed, feed rate
and depth of cut in end milling of hardened steel AISI
H13 with TiN-coated P10 carbide insert tool, and they
recommend the optimal combination of parameters as
higher cutting speed, lower feed rate and lower depth of
cut would produce a good surface finish with low
resultant cutting force. J. S. Pang et al.
10
evaluated the
depth of cut, cutting speed and feed rate machining
parameters on the milling of the halloysite nanotubes
(HNTs) with aluminum reinforced epoxy hybrid
composite material under dry condition to predict the
optimal machining parameters to attain the best surface
finish using the Taguchi method. E. Kuram and B.
Ozcelik
11
evaluated multi-objective optimization
methodology using Taguchi based grey relational
analysis for the micro-milling of Al 7075 material with a
ball-nose end mill. They investigated the effects of
spindle speed, feed per tooth and depth of cut on the tool
wear, force and surface roughness and finally, the
responses were optimized using grey relational analysis.
From the literature survey, it is seen that the most of
the researchers have carried out the process parameter
optimization using Taguchi, RSM and ANN tools on
different materials; however, there is very little research
on advanced materials like the magnesium AZ91D alloy.
This paper deals with the optimization of the end milling
process parameters through evolutionary optimization
techniques such as RSM, ANN and GA methodologies
applied to an advanced material of magnesium alloy
AZ91D to obtain a better surface finish.
2 EXPERIMENTAL PART
In this experimental work, the modern material mag-
nesium AZ91D alloy was taken for evaluating the
optimum machining parameters in the end milling pro-
cess to achieve the minimal surface finish. The milling
process was done as per the recommendation of RSM
Box-Behnken design of experiments and this work were
analyzed, compared with RSM and ANN. The antici-
pated model coupled with the evolutionary optimization
technique GA to attain the better surface finish with the
diversity of machining parameters and the proposed
experimental methodology shown in Figure 1.
2.1 End milling process
The work material with a size of 70 mm × 70 mm ×
10 mm of length, height and width, respectively, were
taken and high-speed steel tool has been utilized for
machining process under dry conditions. The cutting
speed (N) feed rate (f) and depth of cut (ap)w e r e
considered as cutting parameters for this experiment
work. The 15 experiments were carried out based on the
recommendation of Box-Behnken design of experiment
by the interaction of three machining parameters with
three levels. The output response surface roughness is
M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
368 Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373
Figure 1: Proposed experimental methodology

measured with use of Mitutoyo Surftest 211 instrument
corresponding to the experimental run and reported in
Table 1.
Table 1: Box-Behnken design of experiments with experimental
values and predicted values
Expe-
riment
run
N
(min
–1
)
f
(mm/
rev)
ap
(mm)
Expe-
riment
actual
RSM
pre-
dicted
ANN
pre-
dicted
1 750 75 0.4 1.28 1.3225 1.271
2 1250 75 0.4 1.2 1.1875 1.1712
3 750 125 0.4 2.1 2.1125 2.0766
4 1250 125 0.4 1.38 1.3375 1.3805
5 750 100 0.3 1.62 1.5925 1.6221
6 1250 100 0.3 1.2 1.2275 1.199
7 750 100 0.5 1.7 1.6725 1.7032
8 1250 100 0.5 1.1 1.1275 1.1112
9 1000 75 0.3 1.47 1.455 1.4711
10 1000 125 0.3 1.85 1.865 1.8508
11 1000 75 0.5 1.4 1.385 1.4015
12 1000 125 0.5 1.9 1.915 1.9004
13 1000 100 0.4 1.72 1.59 1.6011
14 1000 100 0.4 1.48 1.59 1.6011
15 1000 100 0.4 1.57 1.59 1.6011
2.2 Response surface methodology (RSM)
The RSM is an experimental modeling technique
through which the relationship between control variables
and responses are found.
12,13
The objective of using RSM
in this research is to investigate the effect of cutting
speed, feed rate and depth of cut on the surface finish
(R
a
).
The response Y =  (N,f,ap) (1)
The surface finish R
a
=  (N,f,ap) (2)
The second-order RSM model for the present re-
search is given by:
YXX XX
i
i
k
ii
ij
k
ij i i
i
K
i
=+ + +
===
∑∑∑  
0
11
2
(3)
Where a
0
is the free term of the regression equation;
X
1
, X
2
, ... X
n
are variable terms, 1
, 2
, ..., k
are the linear
coefficient terms;  11
,  22
, ...,  kk
are the quadratic
coefficient terms; and  12
,  13
, ...  k–1
are the interacting
coefficient terms.
14,15
2.3 Artificial neural network (ANN)
ANNs are inspired with the aid of the biological
nervous system. It consists of a huge number of
noticeably linked factors called neurons. To achieve the
specific output target the neural network was trained by
adjusting the values of the weights between neurons. In
an artificial intelligence the network was trained based
on the several input and target pairs, furthermore the
network is adjusted based on an assessment of the output
and the target till the network output suits the goal.
16,17
The three machining parameters speed, feed, and depth
of cut were used as an input layer with one hidden layer
of eight neurons to attain one output layer as surface
roughness. The neural network was trained by a random
training data set and a feed-forward back-propagation
algorithm was used for the train the network. The back
propagation algorithm
18,19
is works based on the method
of gradient descent and it updates the mean square error
values with respect to the target and training values by
the iteration of weights. In this ANN model, the hidden
layer and output layer activation functions were fixed as
logsig and tansig, respectively, for the output response of
the surface roughness. In terms of the ANN training
function, learning functions were placed as traingdx and
learngd, respectively. The developed neural network
model was trained based on the parameters, training was
carried out for 1000 iterations (epochs) and the predicted
results for surface roughness are presented in Table 1.
2.4 Genetic algorithm (GA)
GA is an evolutionary optimization procedure to find
the solutions for problems involving the application of
principles of evolutionary biology.
20,21
GAs use biologi-
cally inspired methods like genetic inheritance, natural
selection, mutation, and crossover. The solution of an
optimization with GA begins with a set of ability
solutions or chromosomes inside the character of bit
strings, which might be randomly selected. The entire set
of these chromosomes incorporates a population and the
chromosomes evolve for the duration of numerous
iterations or generations.
22,23
The GA optimization has
been carried out with the use of MATLAB
®
software
with the quadratic model Equation (4) has been utilized
as an objective function. The procedures to conduct
optimization process, the population size 200, 0.8
crossover probability, uniform distribution of mutation
rate of 0.1, the stall generation as 100 and the stopping
criteria as best fitness were used. The ranges of the
machining parameters, objective function were fixed as
N
min
 N  N
max
; f
min
 f  f
max
;
ap
min
 ap  ap
max
Objective function = Minimize R
a
(N,f,ap)
3. RESULTS AND DISCUSSION
3.1 Main effects and the analysis of variance
In this study the evaluation of the main effects of the
machining parameters on the milling of magnesium alloy
for surface roughness were shown in Figure 2. The main
effect plot shows that the lower rate of speed gives
highest surface roughness value and better surface finish
can be achieved at a higher speed rate. In terms of feed
rate, the plot realizes that minimal surface roughness is
obtained at a lower rate of feed and while increasing feed
rate, the surface roughness is increased predominately. In
an analysis of the depth-of-cut parameter, it understands
that the parameter is insignificant in this operation and
M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373 369

there is minimal deviation occurring in the surface
roughness in spite of increasing or decreasing depth of
cut ranges. The effects plots recommend that the higher
level of cutting speed, middle level of feed rate and a
higher level of depth of cut provides smallest surface
roughness value.
The main objective of the ANOVA is to notice which
machining parameters significantly affect the quality
characteristics. In this experimental work, the ANOVA
Table 2 shows that the percentage contribution of the
model and the error were 96.76 and 3.24, respectively, it
realize that the experiment was done with significant
machining parameters. From this the first order of feed
rate and cutting speed has more contribution on surface
roughness than the depth of cut and the percentages were
38.43 and 36.03, respectively. In terms of second-order
interaction, the cutting speed has a higher contribution of
10.48 % compared to other factors. The ANOVA sug-
gests the feed rate has more contribution in end milling
operation than the cutting speed, it means the slightest
change of feed rate will affect the surface-roughness
bound values. The adequacy of the model is investigated
by the examination of residuals. The normal probability
plots of the residuals versus the predicted response and
Box–Behnken cubic model for the surface roughness
values are shown in Figure 3 and it reveals that the
residuals generally fall on a straight line, implying that
the errors are distributed normally.
Table 2: Analysis of variance for surface roughness
Source DF SS MS F
Contri-
bution
N 1 0.41405 0.414 55.58 36.02 %
f 1 0.44180 0.441 59.30 38.43 %
ap 1 0.00020 0.0002 0.03 0.02 %
N*N 1 0.12048 0.1130 15.18 10.48 %
f*f 1 0.02132 0.0207 2.79 1.85 %
ap*ap 1 0.00037 0.0003 0.05 0.03 %
N*f 1 0.10240 0.1024 13.74 8.91 %
N*ap 1 0.00810 0.0081 1.09 0.70 %
f*ap 1 0.00360 0.0036 0.48 0.31 %
Error 5 0.03725 0.0074 3.24 %
Total 14 1.14957 100.00 %
M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
370 Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373
Figure 4: The 3D surface plot for the effects of machining parameters
on the surface roughness Figure 2: Main effect plots for surface roughness
Figure 3: Normal plot and RSM Box-Behnken cubic design

3.2 The surface and contour-plot analyses
The RSM Box-Behnken experimental analysis was
carried out in order to get better visibility and the
analysis of the effects of machining parameters on the
output response surface roughness the contour plot and
3D surface plots were drawn and shown in Figure 4. The
contour plot and 3D surface plot of Figure 4a and 4d
indicate the higher value of the speed at 1250 min
–1
and a
lower rate of feed of about 75 mm/rev gives a better
surface finish, while increasing the feed rate with a lower
level of speed gives the highest value of the surface
roughness. In Figure 4b and 4e we realize that there is
no significant change while varying the depth of cut with
respect to the speed, but the output response would be
changed with respect to the variation of the speed values.
Figure 4c and 4f represent the end milling process,
while increasing feed rate gives predominantly effects on
the surface roughness in terms of higher roughness
values. The contour and 3D surface plots understand that
the speed and feed rate have the highest influence on the
output response. In this Box-Behnken experimental
design the investigation provides the correlation between
the machining parameters and the output response
surface roughness in terms of the quadratic equation.
The quadratic response model provides the linear and
interaction of the machining parameters N, f and ap
influencing the surface roughness were expressed in
Equation (2) and it could be used to predict the surface-
roughness values for the anticipated machining para-
meters and displayed in Table 1. The RSM predicted
values are closer to the experimental values and the
higher level of speed at 1250 min
–1
, the medium level of
feed rate at 100 mm/rev and the higher level of depth of
cut at 0.5 mm provides a better surface finish of about
1.121 μm, it shows that the development and evaluation
of the optimum machining parameters in the RSM
technique proves the ability of optimization.
R
a
= 2.98 + 0.00797·N + 0.0063·f + 1.35·ap –
0.000003·N·N + 0.00012·f·f–1· ap·ap – 0.00026·N·f –
0.00280·N·ap + 0.12·f·ap (4)
3.3 The evolutionary optimization analysis
The ANN model provides the performance of the
network by an evaluation of the correlation coefficient
between the output values and the target values for the
given test data and shown in Figure 5. The evaluation of
the network provides the correlation coefficient as 0.99
and it suggests that a good correlation has been obtained
in between the experimental and predicted values for this
ANN model. The average relative error in between the
experimental and predicted values is 1.52 %; it confirms
that the well-trained network has good accuracy in
predicting the surface-roughness values. Figure 6 shows
the error percentage chart of the predicted surface
roughness values in terms of RSM, ANN and compared
with experimental values and it resembles the average
relative error found in terms of RSM and ANN models
are 2.40 %, 1.52 %, respectively. The ANN model pro-
vides the predicted values are very close to the experi-
mental values and it suggests a better surface-finish
value of 1.11 μm with the optimal combination of
machining parameters N at 1250 min
–1
, f at 100 mm/rev
and ap at 0.5 mm. In this end milling optimization study
the ANN model provides very close values of the output
response to the experimental than the RSM model, and it
proves the ability of training the model is more efficient.
The GA optimization has been carried out with
coupling the result of ANN based on the genetic para-
meters above and the best fitness was attained for the
given machining parameters ranges and shown in Fig-
ure 7. It shows that minimization of the surface
roughness objective function has been achieved as
1.0926 μm with the optimized values of the cutting
M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373 371
Figure 6: Error percentage chart of experimental vs. predicted values
Figure 5: Artificial neural network: training and target regression
model

parameters cutting speed N at 1250 min
–1
, feed rate f at
82.49 mm/rev and depth of cut ap at 0.497 mm. The con-
firmation test was carried out based on the recommen-
dation from ANN and GA and the results shown in
Table 3. It shows that the ANN and GA model brings
good results with a low error percentage and both me-
thods prove they are more effective in parameter
optimization.
Table 3: Confirmation test
Machining parameters Surface roughness
Error N
(min
–1
)
f
(mm/rev)
ap
(mm)
Predicted Confir-
mation
test
ANN GA
1250 100 0.5 1.11 - 1.15 3.6 %
1250 82.49 0.49 - 1.09 1.12 2.75 %
4 CONCLUSIONS
This experimental work suggests the application of
evolutionary approaches to optimizing the machining
parameters in the milling of the modern material
magnesium alloy AZ91D. The following conclusions can
be drawn:
• The Box-Behnken experimental design provides the
better regression value of 96.76; it proves the most
efficient design in the design of experiment. The
outcome of the ANOVA suggests the cutting speed
and feed rate have more of a contribution in this
experiment, i.e., 36.02 % and 38.43 %, respectively.
• The predicted values in the ANN model are very
close to the experimental values and based on the
error analysis, it indicates that the ANN model is a
more effective approach than the RSM. The genetic
algorithm suggests the minimal surface roughness as
being 1.0926 μm.
• From this research it is clear that the ANN and GA
model delivers the output responses with a lower
error percentage and proves that they are more
efficient methodologies for the optimization of ad-
vanced machining conditions.
5 REFERENCES
1
H. Friedrich, S. Schumann, Research for a "New age of magnesium"
in the automotive industry, Journal of Materials Processing Tech-
nology, 117 (2001), 276–281, doi:10.1016/S0924-0136(01)00780-4
2
Eva María Rubio, María Villeta, Diego Carou, Adolfo Saa, Compa-
rative analysis of sustainable cooling systems in intermittent turning
of magnesium pieces, International Journal of Precision Engineering
Manufacturing, 15 (2014) 5, 929–940, doi:10.1007/s12541-014-
0419-5
3
I. Mukherjee, P. Kumar Ray, A review of optimization techniques in
metal cutting processes, Computers & Industrial Engineering, 50
(2006), 15–34, doi:10.1016/j.cie.2005.10.001
4
A. J. Makadia, J.I. Nanavati, Optimisation of machining parameters
for turning operations based on RSM, Measurement, 46 (2013),
1521–1529, doi:10.1016/j.measurement.2012.11.026
5
K. Shi, D. Zhang, J. Ren, Optimization of process parameters for
surface roughness and microhardness in dry milling of magnesium
alloy using Taguchi with grey relational analysis, International
Journal of Advanced Manufacturing Technology, 81 (2005), 645–65,
doi:10.1007/s00170-015-7218-8
6
N. Haq, P. Marimuthu, R. Jeyapaul, Multi response optimization of
machining parameters of drilling Al/SiC metal matrix composite
using grey relational analysis in the Taguchi method, International
Journal of Advanced Manufacturing Technology, 37 (2008),
250–255. doi:10.1007/s00170-007-0981-4
7
A. M. Zain, H. Haron, S. Sharif, Prediction of surface roughness in
the end milling machining using artificial neural network, Expert
Systems with Applications, 37 (2010) 2, 1755–1768, doi:10.1016/
j.eswa.2009.07.033
8
J. Kopac, P Krajnik, Robust design of flank milling parameters based
on grey-Taguchi method, Journal of Material Processing Technology,
191 (2007) 1–3, 400–403
9
J. A. Ghani, I. A. Choudhury, H. H. Hassan, Application of Taguchi
method in the optimization of end milling parameters, Journal of
Materials Processing Technology, 145 (2004), 84–92. doi:10.1016/
S0924-0136(03)00865-3
10
J. S. Pang, M. N. M. Ansari, O. S. Zaroog, M. H. Ali, S. M. Sapuan,
Taguchi design optimization of machining parameters on the CNC
end milling process of halloysite nanotube with aluminium rein-
forced epoxy matrix (HNT/Al/Ep) hybrid composite, Housing and
Building National Research Center Journal, 10 (2014), 138–144.
doi:10.1016/j.hbrcj.2013.09.007
11
E. Kuram, B. Ozcelik, Multi-objective optimization using Taguchi
based grey relational analysis for micro-milling of Al 7075 material
with ball nose end mill, Measurement, 46 (2013), 1849–1864,
doi:10.1016/j.measurement.2013.02.002
12
Montgomery D C., Design and analysis of experiments, John Wiley
and Sons, New York, 1991
13
P. J Ross, Taguchi Techniques for Quality Engineering, Tata Mc
Graw Hill, (2005)
14
R. H. Myers, D. C. Montgomery, C. M. Anderson-Cook, Response
Surface Methodology: Process and Product Optimization Using
Designed Experiments, John Wiley & Sons, 2016
15
I. Asiltürk, S. Neseli, Multi response optimization of CNC turning
parameters via Taguchi method-based response surface analysis,
Measurement, 45 (2012), 785–794, doi:10.1016/j.measurement.
2011.12.004
16
H. El-Mounayri, H. Kishawy, J. Briceno, Optimization of CNC ball
end milling: a neural network-based model, Journal of Materials
Processing Technology, 166 (2005) 1, 50–62, doi:10.1016/j.jmat-
protec.2004.07.097
M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
372 Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373
Figure 7: Best fitness of surface roughness on GA

17
R. K. Dutta, S. Paul, A. B. Chattopadhyay, The efficacy of back pro-
pagation neural network with delta bar delta learning in predicting
the wear of carbide inserts in face milling, International Journal of
Advanced Manufacturing Technology, 31 (2006) 5–6, 434–442,
doi:10.1007/s00170-005-0230-7
18
N. R. J. Hynes, R. Kumar, J. A. J. Sujana, Optimum bushing length
in thermal drilling of galvanized steel using artificial neural network
coupled with genetic algorithm, Mater. Tehnol., 51 (2017) 5, 813,
doi:10.17222/mit.2016.290
19
H. El-Mounayri, J. F. Briceno, M. Gadallah, A new artificial neural
network approach to modeling ball-end milling, International Journal
of Advanced Manufacturing Technology, 47 (2010) 5–8, 527–534,
doi:10.1007/s00170-009-2217-2
20
M. Durairaja, D. Sudharsunb, N. Swamynathan, Analysis of Process
Parameters in Wire EDM with Stainless Steel using Single Objective
Taguchi Method and Multi Objective Grey Relational Grade,
Procedia Engineering, 64 (2013), 868–877, doi:10.1016/j.proeng.
2013.09.163
21
F. Cus, J. Balic, Optimization of cutting process by GA approach,
Robotics and Computer Integrated Manufacturing, 19 (2003),
113–121, doi:10.1016/S0736-5845(02)00068-6
22
R. Q. Sardin, P. Reis J., P. Davim, Multi-objective optimization of
cutting parameters for drilling laminate composite materials by using
GA, Composites Science and Technology, 66 (2006), 3083–3088,
doi:10.1016/j.compscitech.2006.05.003
23
K. Deb, Multiobjective Optimization Using Evolutionary Algo-
rithms, New York: Wiley (2001)
M. PRADEEPKUMAR et al.: EVALUATION OF THE SURFACE INTEGRITY IN THE MILLING OF A MAGNESIUM ALLOY ...
Materiali in tehnologije / Materials and technology 52 (2018) 3, 367–373 373