This article was downloaded by: [Moskow State Univ Bibliote]

On: 12 October 2013, At: 10:39

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Machining Science and Technology: An International

Journal
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/lmst20

SIMULATED ANNEALING TO ESTIMATE THE OPTIMAL

CUTTING CONDITIONS FOR MINIMIZING SURFACE
ROUGHNESS IN END MILLING Ti-6Al-4V
a

Azlan Mohd Zain , Habibollah Haron & Safian Sharif

Department of Modelling and Industrial Computing, Faculty of Computer Science and

Information System , Universiti Teknologi Malaysia , UTM Skudai Johor, Malaysia
b

Department of Manufacturing and Industrial Engineering, Faculty of Mechanical

Engineering , Universiti Teknologi Malaysia , UTM Skudai Johor, Malaysia
Published online: 11 Mar 2010.

To cite this article: Azlan Mohd Zain , Habibollah Haron & Safian Sharif (2010) SIMULATED ANNEALING TO ESTIMATE THE
OPTIMAL CUTTING CONDITIONS FOR MINIMIZING SURFACE ROUGHNESS IN END MILLING Ti-6Al-4V, Machining Science and
Technology: An International Journal, 14:1, 43-62, DOI: 10.1080/10910340903586558
To link to this article: http://dx.doi.org/10.1080/10910340903586558

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

Machining Science and Technology, 14:4362

Copyright 2010 Universiti Teknologi Malaysia
ISSN: 1091-0344 print/1532-2483 online
DOI: 10.1080/10910340903586558

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

SIMULATED ANNEALING TO ESTIMATE THE OPTIMAL

CUTTING CONDITIONS FOR MINIMIZING SURFACE ROUGHNESS
IN END MILLING Ti-6Al-4V

Azlan Mohd Zain1 , Habibollah Haron1 , and Saan Sharif2

1
Department of Modelling and Industrial Computing, Faculty of Computer Science
and Information System, Universiti Teknologi Malaysia, UTM Skudai Johor, Malaysia
2
Department of Manufacturing and Industrial Engineering, Faculty of Mechanical
Engineering, Universiti Teknologi Malaysia, UTM Skudai Johor, Malaysia

This study presents the estimation of the optimal effect of the radial rake angle of the tool,
combined with cutting speed and feed in inuencing the surface roughness result. Studies on
optimization of cutting conditions for surface roughness in end milling involving radial rake
angle are still lacking. Therefore, considering the radial rake angle, this study applied simulated
annealing in determining the solution of the cutting conditions to obtain the minimum surface
roughness when end milling Ti-6Al-4V . Considering a set of experimental machining data, the
regression model is developed. The best regression model was considered to formulate the tness
function of the simulated annealing. It was recommended that the cutting conditions should be
set at highest cutting speed, lowest feed and highest radial rake angle in order to achieve the
minimum surface roughness of 0
1385 m. Subsequently, it was found that by using simulated
annealing, the minimum surface roughness was much lower than the experimental sample data,
regression modelling and response surface methodology technique by about 27%, 26% and
50%, respectively.
Keywords optimization, radial rake angle, regression, simulated annealing, surface
roughness

INTRODUCTION
In real machining applications, three conicting objectives are
often considered; these are the maximum production rate, minimum
operational cost and good quality of machining (Cus and Zuperl,
2006). Improvement in the quality could be indicated by referring to
one of the machining performance measures, surface roughness (Ra ).
Address correspondence to Azlan Mohd Zain, Department of Modelling and Industrial
Computing, Faculty of Computer Science and Information System, Universiti Teknologi Malaysia,
UTM Skudai Johor 81310, Malaysia. E-mail: azlanmz@utm.my

44

A. M. Zain et al.

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Several optimization techniques, which can be classied as conventional

and non-conventional (soft computing), could be effectively applied to
optimize the cutting conditions that affect the Ra value. The conventional
optimization techniques include Taguchi method, factorial technique,
and response surface methodology (RSM). Among the conventional
optimization techniques, RSM was mostly applied by researchers. Some of
the capabilities of RSM in optimizing cutting conditions during machining
may include (Mukherjee and Ray, 2006):
Useful
intrinsic
model-based
technique
when
sequential
experimentation is possible.
Suitable for solving problems where a lowerorder-polynomial regression
equation exists to establish the relationship between response and
decision variables at early stage of experimentation.
Although RSM works well in many different optimization problems,
there are a few limitations inherent in this technique, such as (Carlyle
et al., 2000; Castillo and Semple, 2000):
Increase in problem complexity when involving multiple objectives.
RSM is based on a series of experiments, and may not be feasible or cost
effective in many manufacturing situations.
Objective or response function needs to be continuously differentiated
for the determination of optimal cutting condition, which may not be
the case in many complex physical processes.
Some of the established soft computing techniques applied by previous
researchers in machining applications are simulated annealing (SA),
genetic algorithm (GA), tabu search (TS), ant colony optimization
(ACO), and particle swarm optimization (PSO) (Aggarwal and Singh,
2005; Mukherjee and Ray, 2006). One of the alternatives in using soft
computing is the application of SA in estimating the optimal cutting
parameters, particularly for the Ra value in end milling process. Some of
the capabilities of SA in optimizing machining conditions may include
(Khan et al., 1997; Tarng et al., 1995; Wang et al., 2005):
SA algorithm is easy to program and typically takes only a few hundred
lines of the computer code.
Implementation of a new problem often only takes very little
modications of the existing code.
SA algorithm does not need to calculate the gradient descent that is
required for most traditional optimization algorithms, hence can be
applied to various kinds of objectives and constraint functions.

Simulated Annealing to Estimate

45

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

SA algorithm can be used to determine the global minimum more

efciently instead of trapping in a local minimum where the objective
function has surrounding barriers.
SA search is independent of initial conditions.
Despite the numerous capabilities of SA, its application in optimization
of cutting conditions for various machining performances was given less
attention by researchers. SA was used to optimize the cutting parameters
for multi-pass milling process (Wang et al., 2005). Juan et al. (2003) applied
SA to the polynomial network to determine the optimal cutting parameters
for minimum production cost in high speed machining (HSM) SKD61 tool
steels. SA was also considered in optimization of machining conditions for
minimum production of spur gears (Sankar and Ponnambalan, 2008) and
for wire electrical discharge machining (Tarng et al., 1995).
Considering the capabilities of SA technique in the machining optimization problem, an effort is undertaken to estimate the best combination
of cutting conditions for the Ra value in end milling titanium alloy,
Ti-6Al-4V. The expectation of the study is to predict the minimum Ra value
as compared to the experimental results and the RSM results that were
reported by Mohruni (2008). Three machining variables were investigated;
they are cutting speed, feed and radial rake angle. Generally, the Ra value is
inuenced by many factors such as cutting parameters, cutting phenomena,
workpiece properties and type of cutting tools. Recent studies on the
optimization of the various inuenced factors for Ra are reported by many
researchers such as in turning (Grzesik and Brol, 2009; Karayel, 2009; Rafai
and Islam, 2009; Salgado et al., 2009) and end milling (Fan and Loftus,
2009; Terai et al., 2009; Vijayaraghavan et al., 2009; Zain et al., 2009).
As shown in Figure 1, the tool geometry is one of the signicant
parameters that affect the Ra . The various types of tool angle in end

FIGURE 1 Parameters that affect Ra (Benardos and Vosnaikos, 2003).

46

A. M. Zain et al.

milling are the axial rake angle, radial rake angle, helix angle, etc. It was
reported that the optimization of cutting conditions for Ra in end milling
involving radial rake angle is still lacking, in particular when dealing with
titanium alloys. As such optimization of the cutting conditions, which
include radial rake angle, combined with cutting speed and feed, for the
Ra in end milling of Ti-6Al-4V can be considered as a new contribution to
the machining research.

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

METHODOLOGY
This study involved four phases in order to obtain the optimal cutting
conditions that minimize the Ra value during end milling Ti-6Al-4V using
various carbide tools. The phases are:
(1) Evaluate the real experimental data involving the various cutting
conditions (cutting speed, feed and radial rake angle) that lead to the
Ra results. For this purpose, the machining experiment in end milling
titanium alloy (Ti-6Al-4V) using various cutting tools is referred.
(2) Develop the machining model to describe the relationship between
independent machining variables and dependent machining variables
(surface roughness) by using the regression technique. By the t test,
the best regression model is determined to be the tness function of
the SA optimization.
(3) Determine the optimal set of cutting conditions using the SA
algorithm which leads to the minimum (lower) value of Ra .
(4) Evaluate the SA optimization solution by calculating the relative
error percentage, where the minimum Ra value at the optimal
cutting conditions will be compared to the minimum Ra value of the
experimental data, regression model and RSM as reported by Mohruni
(2008).

FIGURE 2 Flow of searching for optimum cutting conditions.

47

Simulated Annealing to Estimate

TABLE 1

Chemical composition and mechanical properties of Ti-6Al-4V

Chemical Compositions

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Al
V
Fe
C
Mo
Mn
Si
Ti

6.37
3.89
0.16
0.002
<0
01
<0
01
<0
01
Balance

Mechanical properties
Tensile strength (MPa)
Yield strength (MPa)
Elongation 5D (%)
Reduction in area (%)
Density (g/cm3 )
Modulus of elasticity tension (GPa)
Hardness (Hv)
Thermal conductivity (W/mK)

9601270
820
8
25
4.42
100130
330370
7

Figure 2 illustrates the ow of searching for optimum cutting

conditions for the four phases listed here.
CASE STUDY: EXPERIMENTAL SAMPLE DATA
The experimental data of the Ra values conducted by Mohruni (2008)
during end milling Ti-6Al-4V was considered in this study. The workpiece
was an annealed alpha-beta titanium alloy, Ti-6Al-4V. The mechanical
properties and chemical composition of the Ti-6Al-4V are illustrated in
Table 1. Three types of end mills were used, namely uncoated carbide
(WC-Co) and two TiAlN coated carbide tools which include PVD-TiAlN
coated carbide tool and PVD with enriched Al-content TiAlN coated
carbide tools, also called Supernitride coating (SNTR ). The composition
and properties of these cutting tools are shown in Table 2.
Experimental Design
According to the design of the experiment (DOE) for three
independent variables, the coding variables for the end milling Ti-6Al-4V
are illustrated in Table 3. From this table, the coded variables used in
the 23 -factorial design are only for levels 1, 0 and +1. The variables
TABLE 2

Properties of the cutting tool used in the experiments

Tool type

Substrate (wt%)
Properties
Coating

WC
Co
Grade
Grain size (m)
Process
Coating thickness
Film composition (mol-%AIN)

WC-Co
94
6
K30
0.5

TiAlN coated

Supernitride coated

94
94
6
6
K30
K30
0.5
0.5
PVD-HIS
PVD-HIS
Monolayer (34 m) Multilayer (18 m)
Approx. 54
Approx. 6567

48

A. M. Zain et al.

TABLE 3

Levels of independent variables and coding identication

Level in coded form
Units

1
4142

+1

+1
4142

m/min
mm/tooth

124.53
0.025
6.2

130.00
0.03
7.0

144.22
0.046
9.5

160.00
0.07
13.0

167.03
0.083
14.8

Independent Variables

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Cutting speed, v
Feed rate, f
Radial rake angle, 

that were kept constant were axial depth of cut = 5 mm and radial depth
of cut = 2 mm. The end milling trials were conducted on a CNC MAHO
700S machining centre under wet condition. The specication of the CNC
machine is given in Table 4. The Ra value of the machined workpiece
was measured using a Taylor Hobson Surftronic +3 instrument. Five
measurements were conducted at the location of the length of cut on the
workpiece and the average Ra value was recorded.
Experimental Results
A total of 24 experimental trials were executed based on eight data of
two levels DOE 2k full factorial, four centre and twelve axial points. All Ra
values were collected during actual machining for the three type of cutting
tools, and results are shown in Table 5.
REGRESSION MODELING
Normally, the Ra value in end milling is commonly expressed
mathematically as follows:
Ra = cv k f l m 

(1)

where Ra is the measured surface roughness in m, v is the cutting speed

in m/min, f is the feed in mm/tooth,  is radial rake angle in and c, k, l ,
m are the model parameters to be estimated using the experimental data.
TABLE 4

Specication of the CNC machine

Brand
Model
Electrical data (Motor)
No. of axes
Tool capacity
Spindle speed
Controller

CNC Flexible Machining Cell

MAHO 700S 5 Axis
3 300 V 50 Hz
5
60
206300 rpm
Philip 432

49

Simulated Annealing to Estimate

TABLE 5

Ra values for real machining experiments

Setting values of experimental cutting
conditions

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

No. Data source v (m/min) f (mm/tooth)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

DOE 2k

Centre

Axial

130
160
130
160
130
160
130
160
144
22
144
22
144
22
144
22
124
53
124
53
167
03
167
03
144
22
144
22
144
22
144
22
144
22
144
22
144
22
144
22

0
03
0
03
0
07
0
07
0
03
0
03
0
07
0
07
0
046
0
046
0
046
0
046
0
046
0
046
0
046
0
046
0
025
0
025
0
083
0
083
0
046
0
046
0
046
0
046

Ra (minimum)

 ()
7
7
7
7
13
13
13
13
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
6
2
6
2
14
8
14
8

Experimental Ra value (m)

Ra_uncoated Ra_TiAlN

Ra_SNTR

0.365
0.256
0.498
0.464
0.428
0.252
0.561
0.512
0.464
0.444
0.448
0.424
0.328
0.324
0.236
0.240
0.252
0.262
0.584
0.656
0.304
0.288
0.316
0.348

0.32
0.266
0.606
0.476
0.260
0.232
0.412
0.392
0.324
0.380
0.460
0.304
0.360
0.308
0.340
0.356
0.308
0.328
0.656
0.584
0.300
0.316
0.324
0.396

0.284
0.196
0.668
0.624
0.280
0.190
0.612
0.576
0.329
0.416
0.352
0.400
0.344
0.320
0.272
0.288
0.230
0.234
0.640
0.696
0.361
0.360
0.368
0.360

0.236

0.232

0.190

To develop the regression model for estimating the Ra value, the

mathematical model given in Equation (1) is linearized by performing a
logarithmic transformation as follows:
ln Ra = ln c + k ln v + l ln f + m ln  + ln 

(2)

Equation (2) can be written as:

y = b0 x0 + b1 x1 + b2 x2 + b3 x3 + 

(3)

where y is the logarithmic value of the experimental Ra , x0 = 1 is a

dummy variable, x1 , x2 and x3 are the cutting condition values (logarithmic
transformations) of v, f and  respectively,  is the logarithmic
transformation of experimental error  and b0 , b1 , b2 and b3 are the model
parameters to be estimated using the experimental data.
Equation (3) can also be written as follows:
y = y  = b0 x0 + b1 x1 + b2 x2 + b3 x3

(4)

50

A. M. Zain et al.

TABLE 6

Values of Coefcients for Uncoated Tool

Unstandardized coefcients

Independent variable
1

(Constant)
SPEED
FEED
RAKE_ANGLE
(Constant)

Std. error

0.451
2
67E -03
5.671
4.60E-03
0.386

0.175
0.001
0.811
0.005
0.025

Standardized coefcients
Beta
0
277
0.805
0.097

Sig.

2
582
2
407
6
994
0
842
15
627

0
018
0
026
0
0
41
0

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Dependent variable: UNCOATED

where y is the logarithmic value of the predictive (estimated) Ra .

Subsequently, this equation will be proposed as the tness function of the
optimization solution.
Regression Model for Each Cutting Tool
Based on the actual experimental data given in Table 5, regression
models for each cutting tool given in Equation (4) are developed using
the SPSS software. The values of coefcients for the model parameters
of uncoated, TiAlN coated and SNTR coated cutting tools are given in
Tables 68, respectively.
By transferring the values of coefcients for each cutting tool from
Tables 68 into Equation (4), the regression model for each cutting tool
can be written as follows:

TABLE 7

uncoated = 0
451 0
00267x1 + 5
671x2 + 0
0046x3

y1 = R

(5a)

TiAlN = 0
292 0
000855x1 + 5
383x2 0
00553x3

y2 = R

(5b)

SNTR = 0
237 0
00175x1 + 8
693x2 + 0
00159x3

y3 = R

(5c)

Values of coefcients for TiAlN coated tool

Unstandardized coefcients

Independent variable
1

(Constant)
SPEED
FEED
RAKE_ANGLE
(Constant)

Standardized coefcients

Std. error

Beta

Sig.

0.292
8
55E-04
5.383
5
53E-03
0.375

0.158
0.001
0.731
0.005
0.022

0
098
0
843
0
129

1
85
0
854
7
36
1
122
16
771

0
079
0
403
0
0
275
0

Dependent Variable: TiAlN COATED

51

Simulated Annealing to Estimate

TABLE 8

Values of coefcients for SNTR coated tool

Unstandardized coefcients

Independent variable
1

(Constant)
SPEED
FEED
RAKE_ANGLE
(Constant)

Standardized coefcients

Std. error

Beta

Sig.

0.237
1
75E-03
8.693
1
59E-03
0.392

0.116
0.001
0.539
0.004
0.032

0
14
0
954
0
026

2
042
2
368
16
143
0
437
12
261

0
055
0
028
0
0
667
0

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Dependent variable: SNTR COATED.

Equations (5a)(5c) are then employed to calculate the predicted Ra

values, and the results are summarized in Table 9.
Consequently, scores of Ra values of the experimental data (given in
Table 5) and the predicted values of the regression model (given in

TABLE 9

Ra predicted values of regression modelling

Setting values of experimental cutting
conditions

No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

Data source

DOE 2k

Centre

Axial

Ra predicted values (m)

v (m/min)

f (mm/tooth)

 ()

Runcoated

RTiAlN

RSNTR

130
160
130
160
130
160
130
160
144
22
144
22
144
22
144
22
124
53
124
53
167
03
167
03
144
22
144
22
144
22
144
22
144
22
144
22
144
22
144
22

0
03
0
03
0
07
0
07
0
03
0
03
0
07
0
07
0
046
0
046
0
046
0
046
0
046
0
046
0
046
0
046
0
025
0
025
0
083
0
083
0
046
0
046
0
046
0
046

7
7
7
7
13
13
13
13
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
9
5
6
2
6
2
14
8
14
8

0.306
0.226
0.533
0.453
0.334
0.254
0.561
0.481
0.370
0.370
0.370
0.370
0.423
0.423
0.310
0.310
0.251
0.251
0.580
0.580
0.355
0.355
0.395
0.395

0.304
0.278
0.519
0.493
0.270
0.245
0.486
0.460
0.364
0.364
0.364
0.364
0.381
0.381
0.344
0.344
0.251
0.251
0.563
0.563
0.382
0.382
0.334
0.334

0.259
0.207
0.607
0.554
0.250
0.197
0.597
0.545
0.369
0.369
0.369
0.369
0.404
0.404
0.329
0.329
0.187
0.187
0.691
0.691
0.374
0.374
0.361
0.361

0.226

0.245

0.187

Ra (minimum)

52

A. M. Zain et al.

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Table 9) for each cutting tool are compared, as shown graphically

in Figure 3.
As illustrated in Figure 3, similar pattern of the surface roughness
proles between experimental results and regression models were
observed for all type of cutting tools. Therefore, it can be suggested that
the regression models showed a good estimation of the predicted Ra values
for all tools. From the statistical analysis, the best regression model is
considered to be the tness function for the SA optimization.

FIGURE 3 Comparison of experimental and regression values of Ra . (a) uncoated cutting tool,
(b) TiAlN coated cutting tool, and (c) SNTR coated cutting tool.

53

Simulated Annealing to Estimate

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Determination of the Best Regression Model

In order to choose the appropriate regression model as the tness
function in the optimization module, the paired-sample t -test was
conducted, and the results are summarized in Tables 10 and 11.
Table 10 shows that all three pairs of experimental and regression
modeling data are positively correlated, r (N = 24) = 0
857 for pair 1,
r (N = 24) = 0
859 for pair 2, and r (N = 24) = 0
965 for pair 3.
Results in Table 11 indicate that the mean Ra value for pair 1
increased from the experimental result to the uncoated regression model
by 0
0000833, t (23) = 0
007, p = 0
995. The 95% condence interval
ranges from 0
0264 to 0.0263 (including zero). Therefore, the two
means of experimental result and regression model results are not
signicantly different from each other. The mean Ra value for pair 2 also
increased from the experimental result to the TiAlN regression model
by 0
000542, t (23) = 0
047, p = 0
963. The 95% condence interval
ranges from 0
02430.0232 (including zero), which also proves that the
two means are not signicantly different from each other. By looking at
the pair 3 in Table 11, it can be seen that the mean Ra value however
reduced from the experimental result to the SNTR regression model by
0.000667, t (23) = 0
079, p = 0
938. The 95% condence interval ranges
from 0
01680.0181 (including zero). Thus, the two means too are not
signicantly different from each other.
As a conclusion, it could be summarized that the SNTR coated cutting
tool has shown the highest positive correlation and is the only pair that
indicated a reduction in the mean Ra value from the experimental result.
Thus, it can be suggested that the predicted Ra equation of SNTR coated
tools as given in Equation (5c) is the best regression model and it is
proposed to be the tness function in the SA optimization.
SA OPTIMIZATION
SA is a random search technique that is able to escape local
optima using a probability function (Kirkpatrick et al., 1983). Based on the
TABLE 10

Pair 1
Pair 2
Pair 3

Statistics and correlations for paired samples

Variable

Mean

Std.
deviation

Std. error
mean

EXP_UNCO
REG_UNCO
EXP_TiAlN
REG_TiAlN
EXP_SNTR
REG_SNTR

0.38558
0.38567
0.37533
0.37587
0.39167
0.39100

24
24
24
24
24
24

0.12088
0.10363
0.10964
9.42E-02
0.1565
0.1509

2.47E-02
2.12E-02
2.24E-02
1.92E-02
3.19E-02
3.08E-02

Correlation
(Pearson) Sig.
0.857

0.000

0.859

0.000

0.965

0.000

54

Pair 1
Pair 2
Pair 3

TABLE 11

EXP_UNCO - REG_UNCO
EXP_TiAlN - REG_TiAlN
EXP_SNTR - REG_SNTR

Pair

Paired samples test

Std.
deviation
6.24E-02
5.62E-02
4.13E-02

Mean
8
33E-05
5
42E-04
6.67E-04
1.27E-02
1.15E-02
8.43E-03

Std. error
Mean

Paired Differences

2
64E-02
2
43E-02
1
68E-02

Lower
2.63E-02
2.32E-02
1.81E-02

Upper

95% Condence Interval

of the Difference

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

-0.007
0
047
0.079

23
23
23

Df

0.995
0.963
0.938

Sig.
(2-tailed)

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Simulated Annealing to Estimate

55

iterative improvement, the SA algorithm is a heuristic method with the

basic idea of generating random displacement from any feasible solution.
This process accepts not only the generated solutions, which improve the
tness function but also those which do not improve it with the probability
function; a parameter depending on the tness function (Swarnkar and
Tiwari, 2004).
SA is a method for solving unconstrained and bound-constrained
optimization problems. It models the physical process of heating a material
and then slowly lowering the temperature to decrease defects, thus
minimizing the system energy. At each iteration of the SA algorithm, a
new point is randomly generated. The distance of the new point from
the current point, or the extent of the search, is based on a probability
distribution with a scale proportional to the temperature. The algorithm
accepts all new points that lower the objective, but also, with a certain
probability points that raise the objective. By accepting points that raise
the objective, the algorithm avoids being trapped in a local minimum, and
is able to explore globally for more possible solutions.
An annealing schedule is selected to systematically decrease the
temperature as the algorithm proceeds. As the temperature decreases,
the algorithm reduces the extent of its search to converge to a minimum.
An important part of the SA process is how the inputs are randomized.
The randomization process takes the previous input values and the current
temperature as inputs. The input values are then randomized according to
the temperature. A higher temperature will result in more randomization;
a lower temperature will result in less randomization. There is no specic
method dened by the SA algorithm for how to randomize the inputs.
The exact nature by which this is done often depends upon the nature
of the problem being solved. Figure 4 illustrates the ow on how the SA
technique operates in order to search the optimal solution.
SA Optimization Solution
The target of the optimization process in this study is to determine the
optimal values of the cutting conditions that lead to the minimum value
of Ra . To formulate the optimization problem, the best regression model
which is proposed in Equation (5c) is taken to be the tness function of
the optimization solution and is expressed as follows:
Minimize Ra (v, f , ) = min(0
237 0
00175v + 8
693f 0
00159)

(6)

The minimization of the tness function value of Equation (6) is subjected

to the boundaries (limitations) of the cutting conditions. The range of

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

56

A. M. Zain et al.

FIGURE 4 Flowchart for optimization using SA.

values of experimental cutting conditions in Table 3 is selected to present

the limitations of the optimization solution and is given as follows:
124
53 v 167
03

(7a)

0
025 f 0
083

(7b)

6
2  14
8

(7c)

Since Equation (5c) that relates to the SNTR cutting tool was taken
as the tness function of the optimization solution, therefore, the cutting
conditions that lead to the minimum Ra (minimum) of its regression
model will be chosen to be the initial points for the SA solution and are

Simulated Annealing to Estimate

57

Initial point of v = 144
22

(8a)

Initial point of f = 0
025

(8b)

Initial point of  = 9
5

(8c)

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

given as follows:

Basically, to obtain the optimal solutions, some criteria must be

considered by the SA algorithm as listed in Table 12. By using the tness
function formulated in Equation (6), the limitations of cutting conditions
formulated in Equations (7a)(7c), the initial points formulated in
Equations (8a)(8c), and the SA parameters given in Table12, the Matlab
Optimization Toolbox are applied to nd the minimum values of Ra at the
optimal points. The results of the Matlab Optimization Toolbox are given
in Figures 5 and 6.
As shown in Figure 5, it was observed that the minimum Ra value is
0
1385 m. The set values of optimal cutting conditions that lead to the
minimum Ra value are 167.03 m/min for cutting speed, 0.025 mm/tooth
for feed and 14
797 for radial rake angle. It is also indicated that the
optimal solution is obtained at the 3320th iteration of the SA algorithm.
Figure 6 shows a plot of the tness function value at each iteration. It was
indicated that the best tness value is 0
1385 m.
Evaluation of the SA Result
To evaluate the SA result, the issues concerned in this study are:
(1) The Ra value that estimated by SA is expected to be lower than the
minimum Ra value of the experimental and regression model.
(2) The values of the optimal cutting conditions that lead to the best
tness function which is obtained at the last iteration of SA is expected
to be within the range of the minimum and maximum values of the
level of independent variables (given in Table 3).
TABLE 12
solution

Combination of SA parameter rates leading to the optimal

Parameters

Setting Value / Function Type

Annealing function
Reannealing interval
Temperature update function
Initial temperature
Acceptance probability function
Data type

Boltzmann annealing
100
Exponential temperature
100
Simulated annealing acceptance
Double

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

58

A. M. Zain et al.

FIGURE 5 Results of the Matlab optimization toolbox.

For the rst issue, with reference to Table 5, the minimum Ra value
among all the cutting tools for the actual machining is 0
190 m which
was obtained from SNTR tool. Similarly as shown in Table 9, the minimum
Ra value among all the cutting tools for the regression model is 0
187 m
(Ra value of SNTR cutting tool). Figure 5 shows that the minimum Ra value
of the SA technique is 0
1385 m. Therefore, it can be suggested that the

FIGURE 6 Plot functions of the best tness.

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Simulated Annealing to Estimate

59

SA technique managed to generate the minimum Ra value as compared to

the result of the experimental and regression model.
With regards to the second issue, and referring to Table 3, the
minimum and maximum values of the independent variables are
124.53167.03 m/min for cutting speed, 0.0250.083 mm/tooth for feed
and 6.214
8 for the radial rake angle. From Figure 5, the optimal results
that were estimated by the SA technique are 167.03 m/min for cutting
speed, 0.025 mm/tooth for feed and 14
797 for the radial rake angle.
Since the optimal values for each cutting condition are within the ranges
of minimum and maximum values of the independent variables, it can be
stated that the minimum (best) tness function of the Ra value could be
obtained when it is used in real machining condition.
Theoretically, to validate the result of optimal cutting conditions that
are obtained from SA technique, these values need to be transferred into
the best regression model equation. Equation (5c), which is taken as the
tness function of the optimization SA solution, is used to validate the
optimal cutting conditions. With x1 , x2 , and x3 as the optimal solution for
cutting speed, feed and radial rake angle respectively, the solution is given
as follows:

SNTR = 0
237 0
00175x1 + 8
693x2 0
00159x3
y3 = R
= 0
237 0
00175(167
03) + 8
693(0
025) 0
00159(14
797)
= 0
1385
By transferring the optimal values of SA into the best regression model,
the predicted Ra value obtained is 0
1385 m. Next, this value is compared
to the minimum tness function value of the SA technique. As shown in
Figure 5, the minimum tness function value of the SA is also 0
1385 m
which is exactly the same as the result of the transformation process. This
can be taken as the indicator that the same result (Ra = 0
1385 m) might
be obtained when the optimal set of cutting conditions that was estimated
by using the SA technique are used in the real machining process.
Comparison of the Result: SA vs. RSM
Part of the interest of this study is to compare the result of the SA
with the result of Mohrunis (2008) work. The latter study dealt with the
conventional technique known as the RSM to observe the optimal cutting
conditions when end milling Ti-6Al-4V to determine the minimum value
of Ra . In order to compare the results of the SA with the RSM, the values
of the cutting condition level, noted as 1
4142, 1, 0, +1 and +1
4142 as
given in Table 3, are classied as the lowest, lower, medium, high, highest
scales as stated in Table 13. Then, by using the scale in Table 13, the

60

A. M. Zain et al.

TABLE 13

Classication of cutting conditions scale for comparing the optimal result

Classication of cutting conditions scale

Independent variables

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Cutting speed, v
Feed rate, f
Radial rake angle, 

Units

Lowest

Lower

Medium

High

Highest

m/min
mm/tooth

124
53
0
025
6
200

130
00
0
03
7
00

144
22
0
046
9
500

160
00
0
07
13
00

167
030
0
083
14
800

comparison between the SA and RSM output is summarized as shown in

Table 14.
From Table 14, it was observed that the SA technique outperforms the
RSM solution with respect to the minimum Ra value. At high cutting speed,
medium feed and lower radial rake angle, the lowest Ra value estimated by
RSM was 0
277 m. However, results from SA indicated that the lowest Ra
value of 0
1385 m was obtained at highest speed, lowest feed and highest
radial rake angle when end milling Ti-6Al-4V.
DISCUSSION AND CONCLUSION
This study reports on the application of the SA technique in order
to estimate the optimal solutions of cutting conditions that lead to a
minimum Ra value. The machining experiment carried out by Mohruni
(2008), which dealt with the radial rake angle, combined with cutting
speed and feed in end milling was referred to as the case study. The
workpiece used in the experiment was an annealed alpha-beta titanium
alloy, Ti-6Al-4V. The cuttings tools used were uncoated, TiA1N coated and
SNTR coated carbide tools. The regression model equation of SNTR coated
tool has been selected to be the tness function equation for the SA
optimization. The results of the SA optimization can be summarized in
Table 15. It was found that SA is an effective technique for estimating
the minimum Ra values as compared to the experimental and regression
modeling results. It has also been discovered that the optimal value for
each of the cutting conditions recommended by the SA which lead to the
TABLE 14

Comparison the optimal cutting condition results of SA and RSM

Optimal

Level

Optimal

Level

Optimal

Level

The best
predicted
point of Ra

167.03
160.00

Highest
High

0.025
0.044

Lowest
Medium

14.797
7

Highest
Lower

0.1385
0.277

Cutting speed, v
Technique
SA
RSM (Mohruni, 2008)

Feed rate, f

Radial rake angle, 

61

Simulated Annealing to Estimate

TABLE 15

Summary of the SA result

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

Consideration
Variables
factors

Issue 2:
Issue 1: The
best point of Optimal v, f
and  of SA
Ra

Ra

Experimental result

0.190

Regression result
SA result
Required range: 124.53 to 167.03

0.187
0.1385

167.03

f


Required range: 0.025 to 0,083

Required range: 6.2 to 14.8

0.025
14.797

Remarks
Compared to the
experimental and
regression, the SA
gave the lowest
value of Ra value.

All optimal points of

v, f and  of the
SA were within the
required range values.

minimum Ra values satises the range of minimum and maximum coded

value of the experimental design.
The target of the optimization process is also to determine the optimal
values of decision variables that could lead to the minimum Ra value.
Therefore, with 167.03 m/min for cutting speed, 0.025 mm/tooth for feed
and 14
797 for the radial rake angle, the best Ra value obtained was
0
1385 m, and the percentage ratio of SA to decrease the minimum Ra
is calculated. When compared to the lowest Ra of the experiment data
(0
190 m), the regression model (0
187 m) and the RSM technique
(0.277), it was found that the SA technique reduce the Ra values by about
27%, 26% and 50%, respectively. Similar study of using SA was conducted
by Juan et al. (2003) when they considered the cost minimization in the
machining performance during high-speed milling SKD61. As such SA is
suitable to be used as one of the optimization tool in evaluating machining
performance.
An issue which can also be highlighted is related to the type of cutting
tool proposed to be used in the real machining. The best minimum value
of Ra is recommended by the experiment sample data and regression
model. It was clear from this study that the best minimum Ra values came
from the same source, which is the SNTR coated cutting tool. Therefore,
the SNTR coated cutting tool is recommended for end milling Ti-6Al-4V.
ACKNOWLEDGMENTS
The authors wish to thank the Research Management Center, UTM
and Ministry of Science, Technology and Innovation of Malaysia (MOSTI)
for nancial support through the eScienceFund Grant No. 79318.

62

A. M. Zain et al.

Downloaded by [Moskow State Univ Bibliote] at 10:39 12 October 2013

REFERENCES
Aggarwal, A.; Singh, H. (2005) Optimization of machining techniquesa retrospective and
literature review. Sadhana Journal, 30: 699711.
Benardos, P.G.; Vosnaikos, G.C. (2003) Predicting surface roughness in machining: a review.
International Journal of Machine Tools & Manufacture, 43: 833844.
Carlyle, W. M.; Montgomery, D. C.; Runger, G. C. (2000) Optimization problem and method in
quality control and improvement. Journal of Quality Technology, 32(1): 117.
Castillo, E.D.; Semple, J. (2000) Discussion on Carlyle W. M., Montgomery D. C., Runger G.
Optimization problem and method in quality control and improvement. Journal of Quality
Technology, 32(1): 2023.
Cus, F.; Zuperl, U. (2006) Approach to optimization of cutting conditions by using articial neural
networks. Journal of Materials Processing Technology, 173: 281290.
Fan, X.; Loftus, M. (2009) A simplied cutting force model for light-cut-ball-end milling
applications. Machining Science and Technology, 13(1): 5275.
Grzesik, W.; Brol, S. (2009) Wavelet and fractal approach to surface roughness characterization
after nish turning of different workpiece materials. Journal of Materials Processing Technology,
209: 25222531.
Juan, H.; S. Yu, F.; Lee, B.Y. (2003) The optimal cutting-parameter selection of production cost in
HSM for SKD61 tool steels. International Journal of Machine Tools & Manufacture, 43: 679686.
Karayel, D. (2009) Prediction and control of surface roughness in CNC lathe using articial neural
network. Journal of Materials Processing Technology, 209: 31253137.
Khan, Z.; Prasad, L.B.; Singh, T. (1997) Machining condition optimization by genetic algorithms
and simulated annealing.Computers Operations Research, 24(7): 647657.
Kirkpatrick, F.; Gelatte, C.D.; Vecchi, M.P. (1983) Optimization by simulated annealing. Science, 220:
671780.
Mohruni, A.S. (2008) Performance evaluation of uncoated and coated carbide tools when end milling of
titanium alloy using response surface methodology. Thesis (PhD). Universiti Teknologi Malaysia,
Skudai, Johor, Malaysia.
Mukherjee, I.; Ray, P.K. (2006) A review of optimization techniques in metal cutting processes.
Computer & Industrial Engineering, 50: 1534.
Rafai, N.H.; Islam, M.N. (2009) An investigation into dimensional accuracy and surface nish
achievable in dry turning. Machining Science and Technology, 13(4): 571589.
Salgado, D.R.; Cambero, I.; Marcelo, A.; Alonso, F.J. (2009) Surface roughness prediction based
on the correlation between surface roughness and cutting vibrations in dry turning with
TiN-coated carbide tools. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of
Engineering Manufacture, DOI 10.1243/09544054JEM1508.
Sankar, S.S.; Ponnambalan, S.G. (2008) Optimisation of gear milling process using aggravated
simulated annealing algorithm. International Journal of Advanced Mechatronic System, 1(1): 5362.
Satishkumar, S.; Asokan, P. (2008) Selection of optimal conditions for CNC multitool drilling system
using non-traditional techniques. International Journal of Machining and Machinability of Materials,
3(1/2): 190207.
Swarnkar, R.; Tiwari, M.K. (2004) Modeling machine loading problem of FMSs and its solution
methodology using a hybrid tabu search and simulated annealing-based heuristic approach.
Robotics and Computer-Integrated Manufacturing, 20: 199209.
Tarng, Y.S.; Ma, S.C.; Chung, L.K. (1995) Determination of optimal cutting parameters in wire electrical
discharge machining. International Journal of Machine Tools & Manufacture, 35(12): 16931701.
Terai, H.; Asao, T.; Mizugaki, Y. (2009) Geometric simulation of ball end milling and application for the
precise machining, International Journal of Machining and Machinability of Materials, 6(1/2): 8394.
Vijayaraghavan, A.; Hoover, A.M.; Hartnett, J.; Dornfeld, D.A, (2009) Improving end milling surface
nish by workpiece rotation and adaptive toolpath spacing. International Journal of Machine Tools
& Manufacture, 49: 8998.
Wang, Z.G.; Rahman, M.; Wong, Y.S.; Sun, J. (2005) Optimization of multi-pass milling using parallel
genetic algorithm and parallel genetic simulated annealing. International Journal of Machine
Tools & Manufacture, 45: 17261734.
Zain, A.M.; Haron, H.; Sharif, S. (2009) Application of GA to optimize cutting conditions for
minimizing surface roughness in end milling machining process. Expert Systems with Applications,
doi:10.1016/j.eswa.2009.12.043.