Sunteți pe pagina 1din 17

G Model

JMP-244; No. of Pages 17

ARTICLE IN PRESS
Journal of Manufacturing Processes xxx (2014) xxxxxx

Contents lists available at ScienceDirect

Journal of Manufacturing Processes


journal homepage: www.elsevier.com/locate/manpro

Modeling and optimization of turning duplex stainless steels


Rastee D. Koyee a, , Uwe Heisel a,1 , Rocco Eisseler a,2 , Siegfried Schmauder b,3
a
b

Institute for Machine Tools, Stuttgart University, Holzgartenstr.17, D-70174 Stuttgart, Germany
Institute for Materials Testing, Materials Science and Strength of Materials, Stuttgart University, Pfaffenwaldring 32, D-70569 Stuttgart, Germany

a r t i c l e

i n f o

Article history:
Received 14 January 2014
Received in revised form 25 March 2014
Accepted 9 May 2014
Available online xxx
Keywords:
Duplex stainless steels
Response surface methodology
Cuckoo search
TOPSIS
Operational sustainability index

a b s t r a c t
The attractive combination of high mechanical strength, good corrosion resistance and relatively low cost
has contributed to making duplex stainless steels (DSSs) one of the fastest growing groups of stainless
steels. As the importance of DSSs is increasing, practical information about their successful machining
is expected to be crucial. To address this industrial need, standard EN 1.4462 and super EN 1.4410 DSSs
are machined under constant cutting speed multi-pass facing operations. A systematic approach which
employs different modeling and optimization tools under a three phase investigation scheme has been
adopted. In phase I, the effect of design variables such as cutting parameters, cutting uids and axial length
of cuts are investigated using the D-Optimal method. The mathematical models for performance characteristics such as; percentage increase in radial cutting force (%Fr ), effective cutting power (Pe ), maximum
tool ank wear (VBmax ) and chip volume ratio (R) are developed using response surface methodology
(RSM). The adequacy of derived models for each cutting scenario is checked using analysis of variance
(ANOVA). Parametric meta-heuristic optimization using Cuckoo search (CS) algorithm is then performed
to determine the optimum design variable set for each performance. In the phase II, comprehensive
experiment-based production cost and production rate models are developed. To overcome the conict
between the desire of minimizing the production cost and maximizing the production rate, compromise
solutions are suggested using Technique for Order Preference by Similarity to Ideal Solution (TOPSIS).
The alternatives are ranked according to their relative closeness to the ideal solution. In the phase III,
expert systems based on fuzzy rule modeling approach are adopted to derive measures of machining
operational sustainability called operational sustainability index (OSI). Articial neural network (ANN)
based models are developed to study the effect of design variables on computed OSIs. Cuckoo search
neural network systems (CSNNS) are nally utilized to constrainedly optimize the cutting process per
each cutting scenario. The most appropriate cutting setup to ensure successful turning of standard EN
1.4462 and super EN 1.4410 for each scenario is selected in accordance with conditions which give the
maximum OSI.
2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

1. Introduction
Duplex stainless steels (DSSs) are chromiumnickel
molybdenumiron bi-phased alloys in which the proportions
of the constituent elements enable the optimization of the balance
of the volume fractions of austenite and ferrite [1]. In many ways,
the duplex stainless alloys represent a best of both worlds in
combining traits from the austenitic and ferritic alloys. They have

Corresponding author. Tel.: +49 711 685 84558; mobile: +49 163 9187997.
E-mail addresses: rastee.ali@ifw.uni-stuttgart.de (R.D. Koyee),
siegfried.schmauder@imwf.uni-stuttgart.de (S. Schmauder).
1
Tel.: +49 711 685 83860; fax: +49 711 685 73860.
2
Tel.: +49 711 685 83876; fax: +49 711 685 83858.
3
Tel.: +49 711 685 62556; fax: +49 711 685 62635.

a structure that is roughly a 50% mix of austenite and ferrite; thus,


two hardness materials with different hardnesses coexist side by
side. The presence of ferrite however, means that DSSs have a
ductile brittle transition temperature. During machining process,
the tool will alternate cutting between soft and hard grains of
the duplex structure, leading to an automatic tendency to initiate
chatter in the cutting system. They possess quite a bit higher
mechanical strength and lower ductility than standard austenitic
stainless steels. Therefore, DSSs are expected to be more difcult
to machine than the standard austenitic stainless steels of similar
corrosion resistance [2].
In the last ten years, several studies on machining of stainless
steels have been conducted [38]. The cutting process modeling
by response surface methodology (RSM) using statistical design of
experiments based on D-optimal design is proved to be an efcient
modeling tool [911]. Researchers have considered the machining

http://dx.doi.org/10.1016/j.jmapro.2014.05.004
1526-6125/ 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model
JMP-244; No. of Pages 17

ARTICLE IN PRESS
R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

of metals with multiple performance characteristics and application of one or more multiple attribute decision making (MADM)
techniques to optimize the process parameters [1218]. Considering the drawbacks of traditional optimization techniques, attempts
are being made to optimize the multi-pass machining problem using evolutionary optimization techniques [1924]. Logical
fuzzy reasoning has been applied to multiple output optimization of machining processes [2527]. As investigated by previous
researchers, the articial neural network (ANN) is a powerful tool
for dealing with the complex nature of the cutting process [2830].
A number of attempts have also been made to study the machining of DSSs experimentally [3135]. For instance, Nomani J. and
his coworkers have conducted machinability tests on duplex alloys
SAF 2205 and SAF 2507, while employing austenite stainless steel
316L as a benchmark during drilling. Both duplex alloys displayed
poorer machinability responses, with 2507 being worst [36]. De
Oliveira et al. studied the turning operation of SAF 2507, and its
inuence on the alloys corrosion resistance in practical applications. The results indicate that turning with PVD-coated inserts
under high-pressure cooling resulted in long tool lives, good workpiece roughness and high corrosion resistance of the material after
machining. The most frequent wear mechanism found during the
tests was notch wear, while the main tool wear mechanism was
attrition [37]. Selvaraj et al. are optimized dry turning parameters
of two different grades of nitrogen alloyed duplex stainless steel by
using Taguchi method. Their results revealed that the feed rate is
the more signicant parameter inuencing the surface roughness
and cutting force. The cutting speed was identied as the more
signicant parameter inuencing the tool wear [38]. Most recent
works of Krolczyk et al., in their fourth contribution, identied
microhardness of surface integrity (SI) after turning with wedges of
coated sintered carbide. The investigation included microhardness
analyses in dry and wet machining of duplex stainless steel. It has
been shown that wet cutting leads to the decrease of SI hardening depth, while increasing the rounded cutting edge radius of the
wedge increases the maximum microhardness values and the hardening depth [39]. Finally, some recently conducted studies have
focused on machining technologies in order to achieve sustainable
development objectives [4044].
In majority of the aforesaid past researches, very few efforts
have been made toward application of recent modeling and optimization techniques for optimizing the machining of DSSs. In order
to provide a scientic approach to select the most proper cutting
condition for machining DSSs, this paper presents the application of
relatively new statistical techniques in experimental design, articial intelligence, MADM and recently developed meta-heuristic
optimization algorithms in the context of a three phase investigation. The owchart of the study is shown in Fig. 1.

D- Optimal design
Categorical factors: Cutting Fluid Type

vc
m/min

mm/rev

Lc

ap

mm

mm

Response Surface Modeling

VB

%F r

mm

Pc
W
Fuzzy
Logic

Machine M. Time- C. ToolOccup.


Related
Related
Time
Cost
Cost
(tbB)
K1
K3

MADM
(TOPSIS)

OSI

Ranking
Preference

Optimum
Cutting set

Lc ap f

vc

Adjust Network
via Taguchi

OSI predict.
No

min(MSE)?

OSI actual.

Yes
Inputs

Neural Network
Model
Outputs

CSNNS

Feasible region C1

Feasible
region C2
Solution

Feasible
region C3

Fe a sible
re gion C4

Optimized
Cutting
Parameters
Fig. 1. Flow chart of the study.

2. Materials and methods


2.1. Workpiece material
Although there are many varieties of duplex stainless steels
being cut in industry, the current research selected two common types of these materials for machining investigation. Standard
duplex EN 1.4462 and super duplex EN 1.4410 stainless steel rods
of diameter 55 mm were used for all the turning experiments.
The chemical composition and mechanical properties of workpiece
materials are listed in Table 1.

experiments. Coated carbide inserts with ISO code of CNMG


120408-QM 2025 are used in this study. The QM designation has
a sharp positive edge and open chip breaker which promotes
low cutting force and more stable cutting operation. The inserts
had a multilayer CVD coating (TiN/MtTiCN/Al2 O3 ) on a cemented
carbide substrate and were mounted on a right hand style PCLNL2525M-12 ISO tool holder with a tool geometry feature as follow:
including angle = 80 , back rake angle = 6 , approach angle = 95
and clearance angle = 5 .
2.3. Experimental design

2.2. Machine tool and cutting tool


CNC lathe CTX 420 Linear V5 with maximum drive power
25 kW and a speed range of 357000 rpm is used to perform the

Optimal design methods use a single criterion in order to


construct designs for RSM; this is especially relevant when tting second-order models. D- and A-optimality criteria provide a

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

Table 1
Workpiece materials properties.

Mechanical properties
Yield strength (MPa)
Tensile (MPa)
Hardness (BHN)
Elongation (%)

EN 1.4462

EN 1.4410

0.018
22.42
5.44
3.12
0.84
0.37
0.18
0.025
0.0033

0.015
24.92
6.91
4.06
0.75
0.25
0.3
0.021
0.0007
Fig. 2. Constant cutting speed facing operation.

514
737
212
41

579
826
236
40

measure of the variance of the model coefcients through the


moment matrix, M = X X/N where X is the model matrix and N is the
number of runs in the design. A D-optimal design is one that maximizes the determinant of M, equivalently minimizing the volume
of the condence region on the model coefcients [45]. It is suitable
for conditions when categorical factors are desired to be included
in experimentation plans, models, the full quadratic models have
to be obtained and existing designs have to be augmented. As
with completely randomized designs, computer generated designs
are available to optimize based on various criteria. If interest lies
primarily in estimating the machining parameters and generate
exible designs that meet the needs of particular experimental
design situations then the D-optimal criterion can be well adapted
to the present study. Based on practical considerations, numerical
parameters such us cutting speed (vc ), feed rate (f), depth of cut
(ap ) and total length of cut (Lc ), and categorical parameter such
as process conditions have been selected as input factors to the
experimental design. Table 2 presents the low and high levels of
the involved numerical and categorical factors. For the wet cutting
tests, the cutting area was ooded by a synthetic soluble cutting
uid with a cutting uidwater ratio of 1:20.
Using of the D-optimal design is often involved with insufcient dispersion due to fewer degrees of freedom; therefore, the
design has been bulked up to total 48 runs, i.e., 24 for each category.
Design-Expert V8 statistical package has been used to obtain the
D-optimal plan. The program has identied 113 candidate points
for each of the 2 the process conditions for a total of 226, from
which 36 points have been selected as a minimum for the expected
non-linear model. Eight more runs with unique factor combinations to test lack of t. Finally, 4 of the 44 points already identied
have been replicated. The workpieces were multi-pass faced at constant cutting speed with rotational speed limit of 3500 rpm. Facing
at constant cutting speed optimizes tool life, provides consistent
workpiece nish, and optimizes machining time. An example of
the performed cutting operation is shown in Fig. 2.

Table 2
Process parameters.
Numerical factor levels

Low level

High level

Cutting speed (m/min)


Feed rate (mm/rev)
Depth of cut (mm)
Length of cut (mm)
Process condition

75
0.1
0.5
3
Wet

200
0.25
1.5
12
Dry

2.3.1. Performance characteristics


The cutting force measurement was carried out in each cutting
pass using Kistler type 9129A piezo-electric dynamometer, which
was connected to a charged Kistler type 5070A amplier and personal computer through an analog to digital converter card. To
record the force softly, Windows software for data acquisition and
evaluation DynoWare type 2825A was installed. The force signals
were analyzed to determine the most sensitive force component
to the cutting time. Cutting force in radial direction was seen more
sensitive than the other two components in axial and tangential
directions. Therefore, a new parameter which accounts for percentage increase in radial cutting force has been calculated using the
formula:
%Fr =

Frmax Frinit

(1)

Frinitial

where Frmax . is the maximum and Frinit . is the initial radial cutting
forces in each experimental run. In Fig. 3, a typical radial force signal
of wet cutting EN 1.4462 at vc = 200 m/min, f = 0.25 mm/rev, ap = 1.5,
Lc = 12 mm is shown, indicative of the increasing trend when the
total length of cut has been reached. The average radial cutting
force in constant cutting speed area in every cutting pass per cutting run is calculated and the cutting pass with maximum value is
designated. In order to compute all increasing percentages in the
study, a total of 236 cutting passes per category and 472 cutting
passes per material have to be similarly analyzed.
One of the most important items related to cleaner production consists in reducing energy consumption in order to cut down
carbon emissions associated to energy generation. Therefore, the
association of environmental footprint in view of minimization of
motor power per each experimental run was considered. The power
consumed by the motor of a machine tool is composed of an effective and an idle component. Because of its proportionality to the
torque emitted by the motor, the effective power is often used as
a signal within the control system for quantifying the motor load.

600

Radial cutting force (N)

Chemical composition %weight


C
Cr
Ni
Mo
Mn
Si
N
P
S

500
400
300
200
100
0
10

20

30

40

50

60

70

Cutting time (sec)


Fig. 3. Typical radial cutting force increasing trend.

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model
JMP-244; No. of Pages 17

ARTICLE IN PRESS
R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

External effective-power modules are installed in machine tools


between the frequency converter and the motor. To read the current, hall sensors are used that measure the current along with its
phasing via the magnetic eld surrounding the conductor in a circular shape. The effective power is obtained from these using Eq.
(2) [46].
Pe =

3.U.I. cos

(2)

where U is the line voltage in volts, I is the line current in ampere


and cos is the power factor. The average Pe values of cutting
passes were added and later divided by total number of passes to
determine the average effective cutting power consumption in each
experiment run.
At the end of each experimental run the tool wear was measured
using an optical microscope and the tool life was estimated based
on ank wear criterion of VBmax = 600 m on the major cutting
edge. During cutting tests the metal chips posed great challenges
to the machine tool and the workpiece. To deal with this problem, the chips were collected after each machining trial and later
analyzed. Chip volume ratios (R) were then calculated by; (a) measuring the mass of the sample chip, (b) dividing the mass by density
of the steels hence calculating the volume of the cut material prior
to machining, (c) gently placing the chips inside plastic bags, (d)
gently vacuum packaging which removed air from the package
prior to sealing, (e) measuring the volume of the gently packed
chips using the water displacement procedure and (f) dividing the
volume of the gently packed chips by volume of the cut material
prior to machining. The ratio corresponds to the intended R. The
experimental results as shown in Table 3 were utilized for the next
analyses.

2.3.2. RSM modeling of performance characteristics


Response surface methodology (RSM) is a collection of statistical and mathematical techniques useful for developing, improving,
and optimizing processes. It also has important applications in the
design, development, and formulation of new products, as well as
in the improvement of existing product designs. Successful use
of RSM is critically dependent upon the experimenters ability to
develop a suitable approximation for response function. Usually,
a low-order polynomial in some relatively small region of the
independent variable space is appropriate. In many cases, either
a rst-order or a second-order model is used. Often the curvature in the true response surface in machining experimentations
is strong enough that the rst-order model (even with the interaction term included) is inadequate. A second-order model will likely
be required in these situations. The second-order model is widely
used in response surface methodology for their exibility, easiness
in estimating the constant parameters, and indications of considerable practical experiences which conrm that second-order models
work well in solving real response surface problems [47]. However,
when the entire design space is used to develop the models, then,
second-order models, modied second order models, cubic models, modied cubic models, response transformations and model
design reductions were seen not satisfactory to capture accurately
the highly non-linear relations between the cutting variables and
performance characteristics. Therefore, an intermediate and reasonable solution were found for all the cases when second-order
non-linear mathematical models in terms of natural variables are
selected to predict the performance characteristics (Y) for each condition category in separate, which were of the following form:
Y = b0 + b1 vc + b2 f + b3 ap + b4 Lc + b11 v2c + b22 f 2 + b33 a2p + b44 Lc2
+ b12 vc f + b13 vc ap + b14 vc Lc + b23 fap + b24 fLc + b34 ap Lc

(3)

The values of the regression coefcients as dened above are


determined using the following formula:
b = (X T X)

XTY

(4)
XT ,

transpose of the
where b, matrix of parameter estimates;
independent variables matrix (X); Y, matrix of the performance
characteristic.
2.4. Cuckoo search
Cuckoo Search is one of the latest nature-inspired metaheuristic algorithms, developed by Xin-She Yang and Suash Deb
[48]. Cuckoo Search is inspired by lifestyle of a bird family called
cuckoo. Specic egg laying and breeding of cuckoos is the basis
of this optimization algorithm. Like other evolutionary algorithms,
the proposed algorithm starts with an initial population of cuckoos. These initial cuckoos have some eggs to lay in some host birds
nests. Some of these eggs which are more similar to the host birds
eggs have the opportunity to grow up and become a mature cuckoo.
Other eggs are detected by host birds and are killed. The grown eggs
reveal the suitability of the nests in that area. The more eggs survive in an area, the more prot is gained in that area. So the position
in which more eggs survive will be the term that cuckoo optimization algorithm is going to optimize [49]. For simplicity in describing
Cuckoo Search the following idealized assumptions are made:
a. Each cuckoo lays one egg at a time, and dumps its egg in a randomly chosen nest.
b. The best nest with high-quality eggs will be carried over to the
next generation.
c. The number of available host nests is xed, and the egg laid by
a cuckoo is discovered by the host bird with a probability Pa
between 0 and 1. In this case, the host bird can either get rid
of the egg, or simply abandon the nest and build a completely
new nest.
3. Results and discussions
3.1. Effect of independent cutting variables on performance
characteristics
The values of regression coefcients in natural form and various statistics about the statistical validity of the developed models
2
at a 95% condence interval are given in Table 4. The R2 and Radj
.
values indicate that the models t the data well. As per analysis of
variance (ANOVA) technique, since the calculated value of Fratio of
all developed second-order models are greater than the standard
tabulated value of the table (F14,9,0.05 = 3.03), then the models are
considered adequate within the condence limit. The adequate
precision (Aprec. ) for all models are greater than 4, which indicate
an adequate signal to noise ratio, thus the models can be used to
navigate the design space.
3.1.1. Effect of independent cutting variables on the percentage
increase in radial cutting force (%Fr )
In phase I, the developed RSM models were utilized to study
the interaction effects of selected independent variables on %Fr .
To analyze the interaction effects, three dimensional plots were
generated considering two parameter at a time while the other
parameters are held constant at their respective center levels. These
interaction plots are presented in Fig. 4a. From which, the following
observations can be made:
a. For a given depth and length of cut, %Fr approaches its minimum value as cutting speed and feed rate increase to certain

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

EN 1.4462

vc (m/min)

f (mm/r)

ap (mm)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

200
75
75
200
200
200
75
137.5
75
200
75
200
200
200
75
75
75
200
200
200
75
137.5
137.5
75
75
137.5
200
106.25
137.5
200
75
200
200
75.0
137.5
200
75
168.75
75
200
200
137.5
200
75
75
75
75
200

0.25
0.25
0.25
0.25
0.25
0.10
0.17
0.17
0.25
0.10
0.10
0.25
0.10
0.10
0.25
0.10
0.25
0.25
0.25
0.25
0.10
0.17
0.17
0.25
0.10
0.10
0.10
0.21
0.10
0.25
0.17
0.17
0.17
0.10
0.25
0.10
0.10
0.14
0.17
0.10
0.25
0.25
0.10
0.25
0.10
0.25
0.10
0.17

1.5
1.5
1.5
0.5
0.5
0.5
1
1.5
1
1.5
1.5
0.5
1.5
0.5
1.5
0.5
0.5
1.5
1
1.5
1.5
1
1
0.5
0.5
1
1.5
0.75
1.5
1.5
1.5
1
1.5
0.5
1.5
0.5
1
0.75
0.5
0.5
0.5
0.5
1
0.5
0.5
0.5
1.5
0.5

Lc (mm)
12
7.5
3
12
3
12
7.5
7.5
3
3
12
12
12
3
12
3
3
3
12
3
3
3
12
3
12
7.5
7.5
7.5
3
3
3
7.5
12
7.5
12
12
12
7.5
12
12
7.5
3
3
12
3
12
12
3

EN 1.4410

Cutting cond.

%Fr

Pe (W)

VBmax (m)

%Fr

Pe (W)

VBmax (m)

Wet
Dry
Wet
Wet
Wet
Wet
Wet
Wet
Dry
Wet
Wet
Wet
Wet
Wet
Wet
Dry
Wet
Dry
Dry
Wet
Wet
Wet
Wet
Wet
Wet
Wet
Dry
Dry
Dry
Dry
Dry
Wet
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Wet
Wet
Dry
Dry
Dry

21.777
18.382
9.660
26.401
12.545
10.383
15.900
5.609
15.258
11.280
19.086
26.401
16.737
11.791
20.717
20.418
14.300
25.026
36.770
10.733
14.310
2.265
10.520
14.300
25.617
12.708
26.499
15.115
10.559
27.016
10.830
14.471
28.663
21.113
16.158
55.258
32.324
19.217
24.680
55.988
37.156
13.758
24.212
25.459
17.480
29.590
17.597
18.255

4236.490
1815.024
1000.050
3489.350
3234.730
3545.790
923.635
1904.170
958.954
3143.700
968.770
3487.520
3294.360
3155.320
1306.680
912.855
912.033
4556.128
4621.133
4072.790
793.157
1781.650
1616.510
913.133
838.760
1592.840
3634.899
1558.224
1946.809
4556.128
1206.554
3672.370
3330.270
996.490
2718.074
3293.114
1146.716
2197.682
959.551
3313.114
3557.210
2240.772
3207.704
998.866
759.173
1052.569
1051.178
3232.602

193.784
117.880
70.967
211.677
109.044
93.154
101.364
75.312
98.648
93.781
154.679
211.693
131.510
137.022
134.271
130.369
91.209
178.543
230.010
93.302
114.375
59.643
89.178
88.608
146.160
82.713
167.532
97.768
79.498
180.683
71.187
112.962
180.719
234.622
124.200
448.002
203.001
127.007
156.435
446.414
232.355
109.378
152.577
122.391
132.548
216.367
113.056
119.098

32.025
159.015
116.754
105.655
96.068
105.939
99.369
91.591
24.016
59.867
97.277
106.946
63.959
101.368
128.988
88.146
98.674
33.886
17.753
21.022
92.396
50.501
55.124
97.675
74.074
69.125
121.988
69.057
115.750
34.013
139.880
48.026
29.875
111.258
65.049
119.785
104.255
98.879
126.020
121.665
116.846
56.322
64.148
107.104
70.778
119.995
112.556
69.339

30.187
20.390
10.415
32.821
16.319
13.734
15.100
10.485
11.977
14.307
23.861
32.821
19.961
21.276
20.597
24.504
13.590
31.561
41.417
13.953
17.752
8.102
12.997
13.075
22.044
11.703
30.389
19.021
14.382
30.553
13.057
16.727
35.530
26.133
20.386
61.414
11.022
23.357
29.416
61.414
42.082
22.933
26.892
18.372
20.114
34.714
23.056
22.813

4451.450
2222.079
1063.860
3566.580
3467.240
3887.250
1011.250
2018.540
980.695
3356.980
1100.785
3586.330
3489.570
3489.670
1512.830
831.534
998.238
5748.831
5493.925
4189.540
842.735
1884.760
1745.240
985.853
996.367
1606.870
3848.593
1547.884
2052.213
5739.279
1332.649
4016.450
3622.993
1027.255
2812.954
3579.649
1071.263
2201.888
913.944
2934.255
3557.355
1971.538
3514.131
1208.460
845.423
955.323
1075.786
3447.930

186.874
134.668
85.757
225.957
111.492
93.819
137.964
57.276
92.118
98.180
164.205
225.972
145.644
103.697
176.426
160.162
124.067
203.640
284.183
95.447
122.340
29.152
95.351
124.555
218.835
114.160
216.633
126.149
97.116
197.442
88.758
128.364
234.047
170.221
134.641
579.189
81.977
153.066
190.451
586.846
268.086
150.439
185.018
215.834
151.705
205.826
153.231
153.748

37.488
157.925
141.096
115.255
100.109
114.013
105.846
111.865
71.809
66.079
106.946
117.567
76.158
123.895
155.125
91.102
113.855
39.962
29.796
25.092
105.448
75.026
82.567
112.115
81.368
83.558
119.588
86.653
123.736
38.267
147.997
60.467
57.999
155.727
30.174
138.548
122.984
117.180
145.013
139.975
135.383
84.121
91.980
118.025
80.757
128.617
140.951
65.265

ARTICLE IN PRESS

Independent variables

G Model

JMP-244; No. of Pages 17

Exp. No

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

Table 3
Experimental results.

G Model

JMP-244; No. of Pages 17

Coef.

EN 1.4462

EN 1.4410

%Fr

VBmax

Pe

%Fr

Pe

VBmax

Dry

Wet

Dry

Wet

Dry

Wet

Dry

Wet

Dry

Wet

Dry

Wet

Dry

Wet

Dry

70.6
0.59
386.
9.22
1.58
0.00
931
6.29
0.10
0.38
0.03
0.00
22.1
5.81
0.02

69.4
0.48
435
0.74
1.70
0.00
1235
1.34
0.08
0.01
0.03
0.01
31.8
5.36
1.64

1700.9
20.63
1384.5
811.25
61.81
0.13
3539.4
88.09
3.74
11.96
1.78
0.08
3421.3
16.90
3.13

2165.7
5.30
23259
0.52
46.41
0.07
50,842
8.97
0.54
36.67
0.23
0.06
3590.9
97.45
36.41

467.2
3.03
1550
69.37
10.37
0.01
2176
27.36
0.08
3.88
0.06
0.01
73.52
48.25
3.56

604.2
1.91
4341.5
257.56
19.05
0.01
11,795
103.39
0.44
0.95
0.16
0.07
580.6
58.30
14.68

29.8
0.67
412.3
144.9
13.30
0.00
95.65
118.20
0.90
2.46
0.65
0.00
117.5
5.89
0.02

88.9
0.32
401.5
173.5
31.54
0.00
737.5
166.96
1.60
1.27
0.65
0.02
93.95
26.88
7.57

75.8
0.51
253.4
10.17
1.94
0.00
358.9
4.15
0.03
0.62
0.01
0.00
12.25
8.00
0.58

22.3
0.21
160.1
21.75
2.01
0.00
425.8
9.63
0.11
0.28
0.04
0.00
35.85
4.00
0.60

1765.7
25.55
5243
1274.9
59.15
0.16
21,964
236.10
2.81
3.69
1.52
0.04
4036.3
9.61
11.06

2013.1
18.75
14,453
1963.4
181.15
0.06
8438.1
869.17
9.98
25.30
3.76
0.42
3235.8
360.96
54.40

576.9
4.70
3123
65.42
13.86
0.01
7473.3
47.89
0.88
3.24
0.28
0.01
176.9
46.71
0.07

176.2
1.64
1265.2
171.8
15.86
0.01
3365
76.09
0.87
2.21
0.33
0.04
283.2
31.60
4.76

16.31
1.56
732.1
143.2
6.91
0.00
666.4
119.06
0.58
3.27
0.74
0.00
77
9.78
0.52

76.65
0.71
550.24
74.7
6.90
0.00
1463.4
33.09
0.38
0.96
0.14
0.02
123.2
13.74
2.07

ANOVA
R2
2
Radj

0.96
0.90

0.97
0.93

0.99
0.98

0.98
0.95

0.92
0.80

0.98
0.95

0.97
0.91

0.93
0.83

0.92
0.80

0.95
0.87

0.99
0.97

0.97
0.93

0.96
0.89

0.97
0.92

0.97
0.93

0.94
0.85

RMSE
Aprec.
Fratio.
Reg. DF
Err. DF

2.09
14.9
15
14
9

18.6
9.82
7.57
14
9

21.4
22.91
32.9
14
9

8.33
15.57
17.8
14
9

16.8
9.66
8.89
14
9

3.01
9.82
7.62
14
9

4.86
12.67
12.4
14
9

212
21.59
61.3
14
9

423
14.67
21.9
14
9

17.4
14.53
14.5
14
9

37
17.33
19.2
14
9

8.27
19.1
21.6
14
9

b0
b1
b2
b3
b4
b11
b22
b33
b44
b12
b13
b14
b23
b24
b34

3.2
17.7
22.3
14
9

179
24.5
79.9
14
9

274
17.1
34.8
14
9

RMSE, root mean squared error; Reg. DF, regression degrees of freedom; Err. DF, error degrees of freedom.

16.1
9.465
10.6
14
9

ARTICLE IN PRESS

Wet

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

Table 4
Models coefcients estimates and ANOVA.

G Model
JMP-244; No. of Pages 17

ARTICLE IN PRESS
R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

Fig. 4. Interaction effects of independent cutting variables on the performance characteristics.

specied limit. Beyond that limit it starts to increase again. When


the materials are cut dry, lower cutting speed and feed rates are
preferred to minimize %Fr . The values of obtained %Fr when EN
1.4410 are cut were higher than EN 1.4462 due to the generally
higher mechanical strength of EN 1.4410 and fewer weight percentage of assisting machinability elements such as sulfur and
phosphorus, see Table 1.
b. For a specied cutting speed and feed rate, depth of cut values
of (1.251.5 mm) were seen to give the minimum %Fr , as they
reduce the ploughing effect, produce friendlier-to-machine chips
and exert a more stable cutting process. Generally, lower values
of depths of cuts are recommended when process conditions are
dry and/or super DSS EN 1.4410 is machined. Values lower than
tool nose radius of 0.8 mm have to be avoided.
c. When cutting speed, feed rate and depth of cut are set at a specied value, %Fr is expected to increase with the increase of total
length of cut. However, the slopes of %Fr versus total length of
cut were seen steeper at higher levels of cutting speeds.

3.1.2. Effect of independent cutting variables on the effective


cutting power (Pe )
Effect-power measurement systems are characterized by the
fact that they do not affect the mechanical properties of the
machine tool. The machining torque can be measured during the
operation without the integration of external sensors into the
electric ux of the machine. The primary area of application is
the recognition of tool fractures and collisions in the workspace.
Sufciently large force changes are necessary for efcient wear
monitoring. When monitoring cutting processes, usually external
effective-power measurement tools are used, often with associated
evaluation software and visualization unit [46].

Fig. 4b illustrates the interaction effects of the independent variables on the effective cutting power. It is seen from this collective
gure that:
a. Effective cutting power is most sensitive to cutting speed variations when other parameters are kept constant. Maximum and
minimum consumptions in effective cutting power are seen at
wet cutting of EN 1.4462 and dry cutting of EN 1.4410 respectively.
b. Effective cutting power is less sensitive to the length of cut variations as far as the cutting tool does not suffer the catastrophic
chipping in the cutting edges.
c. Minimum consumption in effective cutting power results when
the feed rate is in the range 0.1250.175 mm/rev of dry cutting
and 0.10.15 mm/rev of wet cutting.
d. Two-factor interaction models were seen accurate enough to
explain the interaction effects of cutting parameters on the effective cutting power.
3.1.3. Effect of independent cutting variables on the maximum
tool ank wear (VBmax )
Tool wear has a remarkable inuence in tool life, cutting forces,
vibration, quality of the machined surface and its dimensional accuracy, and consequently, the economics of cutting operations. In this
study off-line modeling of the maximum ank wear lands (VBmax )
are described as functions of independent cutting variables. Fig. 4c
depicts the interaction contour plots of VBmax as per RSM models
tabulated in Table 4. It can be found that:
a. During the course of the dry cutting of DSSs, an increased built
up edge formation on the tool, obvious inferior chip morphologies, more aggressive notch wear on the major and minor cutting

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

8
48

EN 1.4462
6
43

EN 1.4410

Alternative Rank

37
31
25
19
13
7
1
1

11

13

15

17

19

21

23

2
25

27

29

31

33

35

37

39
9

41

43

45

4 7 48

Alternative
v No.
Fig. 5. Alternatives ranks.

edges and inferior surface qualities were noticed. Therefore,


when cutting DSSs, generous amount of cutting uid is always
recommended.
b. Wet cutting at cutting parameter ranges of vc 100160 m/min,
f 0.150.25 mm/rev, ap 0.751 mm for machining EN
1.4462
and
vc 120160 m/min, f 0.150.20 mm/rev,
ap 1.251.5 mm for machining EN 1.4462 were seen optimum
in minimizing the tool ank wear.
c. Dry machining is possible when cutting parameters are set
appropriately. The settings include; machining of EN 1.4462
at vc 75110 m/min, f 0.150.2 mm/rev, ap 11.5 mm
and maximum Lc /ap 6, and machining of EN 1.4410 at
vc 7590 m/min, f 0.1250.175 mm/rev, ap 0.81.35 mm
and maximum Lc /ap 4.
d. Direct linear correlations between %Fr and VBmax with reasonable
coefcients of determinations (R2 ) were noticed:
Cutting of DSS EN 1.4462
Wet: R2 = 0.70
VBmax = 5.4198 (%Fr ) + 35.149

(5)

Dry: R2 = 0.92
VBmax = 7.7696 (%Fr ) 18.549

(6)

Cutting of DSS EN 1.4410


Wet: R2 = 0.71
VBmax = 6.5767 (%Fr ) 16.793
Dry:

R2

(7)

= 0.93

VBmax = 9.918 (%Fr ) 63.494

(8)

3.1.4. Effect of independent cutting variables on the chip volume


ratio (R)
The chip shapes are assessed according to two criteria: transportability and danger for the machine operator. It is no problem
to move short, broken chips, such as fragmented spiral chips, in containers. By contrast, this is impossible for ribbon chips, since they
always demand special treatment (breaking in the chip breaker
or briquetting) in order to make them ready for transport. In
a plant with automated manufacturing equipment, where many
chips occur, these treatment procedures are very expensive. Consequently, as an alternative, always the aim is to produce chip forms
that can be handled easily. Since long ribbon chips and entangled

chips, whose edges are very sharp, could possibly endanger the
machine operator and cause safety risks [50]. In this study, utilizing the chip volume ratio R, the spatial requirement for the chips
is considered. Each chip form is assigned to a chip volume ratio
R, which denes by what factor the transport volume needed for
the specic chip form exceeds the intrinsic material volume of the
chip. According to these assignments, R 100 for snarled and ribbon chips, 60 R 100 for coiled, at helical and cylindrical helical
chips, 30 R 60 for short coiled chips, 10 R 30 for spiral chips
and 3 R 10 for short chip particles. It is clear from Fig. 5d that;
a. Higher values of cutting speeds and feed rates are necessary to
minimize the chip volume ratio.
b. In wet cutting of DSSs, friendlier-to-machine chip forms were
produced because of the less encountered friction in the contact
area between the chip and rake face.
c. Ribbon and snarled chip forms were common when the
metals are machined at feed rate and depth of cuts lower
than 0.15 mm/rev and 1 mm, respectively. At intermediate feed
ranges, the produced chips were rather of at-helical and
cylindrical-helical forms. The chip forms were rather short coiled
chips at higher feed ranges.
d. Under the same machining conditions which might produce
continuous chips, different chip forms were possible due to continuous entanglement of chips especially when Lc /ap 6.
e. Due to the higher proof and tensile strength, and, lower sulfur and
phosphor constituents to assist chip breaking, R values of cutting
EN 1.4410 were generally higher than of cutting EN 1.4462.
f. The tendency to enhance chip segmentation is expected to rise
when the chip curl radius is reduced, the coolant pressure is
increased and the coolant restricted to smaller area of the chip.
3.2. Parametric optimization of performance characteristics
The rst optimization process in this study is formulated as follows:

3.2.1. Objective function


Formulation of optimization model is one of the most important
tasks in optimization process. The type of optimization modeling
techniques used to express the objective function determines its
accuracy and the possibility of reaching a global optimum solution. The developed response surface models, expressed by Eq. (3)
and presented previously in Table 4 are used as both objective and
constraint functions.

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

Table 5
Optimization results.
Material

Process
condition

vc (m/min)

ap (mm)

Lc (mm)

%Fr

Pe (W)

VBmax (m)

Minimization of percentage increase in radial cutting force


EN
144.8322
0.1582
Wet
1.4462
Dry
107.1897
0.1833
EN
140.7331
0.1238
Wet
1.4410
80.8061
0.1678
Dry

1.5000
1.5000
1.5000
1.5000

12.0000
12.0000
12.0000
12.0000

6.0589
9.4987
14.8064
11.9899

2027.4
1578.7
1904.1
1142.4

104.3667
51.64860
108.0094
53.7075

73.4938
92.6377
99.2323
114.1464

Minimization of effective cutting power


EN
75.0000
Wet
1.4462
Dry
75.0000
EN
75.0000
Wet
1.4410
Dry
75.0000

0.1000
0.1372
0.1000
0.1163

1.5000
1.5000
1.5000
1.5000

12.0000
12.0000
12.0000
12.0000

19.1845
14.1790
22.4745
14.7725

824.876
1022.90
953.4741
964.9281

146.1453
69.7186
184.9923
84.5455

102.5605
124.5878
113.1389
141.0985

Minimization of maximum cutting tool ank wear


EN
122.7485
Wet
1.4462
Dry
85.1292
EN
Wet
134.5424
1.4410
Dry
75.0000

0.2388
0.1802
0.1600
0.1732

1.5000
1.5000
1.5000
1.5000

12.0000
12.0000
12.0000
12.0000

12.2771
10.4841
15.1967
12.0794

1861.6
1240.8
1965.8
1136.9

124.9669
47.4439
98.4184
49.9869

88.1069
111.8589
105.7843
118.4307

Minimization of chip volume ratio


EN
Wet
1.4462
Dry
EN
Wet
1.4410
Dry

0.2500
0.2500
0.2432
0.2500

1.5000
1.5000
1.5000
1.5000

12.0000
12.0000
12.0000
12.0000

21.1169
32.1583
29.5881
40.2478

4151.2
4533.9
4299.4
5344.2

195.9552
210.0153
225.9687
259.3247

35.1433
19.6510
45.8620
8.9413

200.0000
200.0000
200.0000
200.0000

f (mm/rev)

Objective functions are assigned in bold fonts.

3.2.2. Decision variables


Process parameters considered in the optimization problem
were cutting speed, feed rate, depth of cut and length of cut. The
length of cut and depth of cuts were set constant at their maximum
values to account for the maximum production rates and ensure
an integer Lc /ap ratio due to the imposed practical considerations.
Therefore, the parameter bounds for the decision parameters were
set as shown below:
75 vc 200 m/min, 0.1 f 0.25 mm/rev, ap = 1.5 mm,
and Lc = 12 mm

3.2.3. Constraints
To ensure that the result of minimization of a performance will
not cause an excessive increase in the other three performances,
once a performance is selected as a primary objective the others are
selected as implicit constraints Cj . This is mathematically expressed
as below:
LBj Cj UBj

j=1:3

(9)

where LBj and UBj are lower and upper bounds of the implicit constraints which are selected based on practical considerations. In
this study, it would be appropriate when %Fr %Frmax . , Pe Pemax ,
VBmax VBmax(max) and R Rmax .
3.2.4. Optimization algorithm
For the purpose of nding minimum of constrained nonlinear multivariable functions described before, Cuckoo Search (CS)
algorithm was seen powerful enough to perform the task. The initializing optimization parameters for CS algorithms were: number
of nests (n) = 20 and probability (Pa ) = 0.25. The obtained optimization results showed that CS is highly reliable and converge
consistently to the optimum solution. In Table 5, the presented
results indicate a conicting inuence of the process parameter on
the performance characteristics in conjunction with the different
process conditions. As a summary, the following conclusion points
can be depicted from the table:

a. To optimize %Fr and VBmax in dry cutting, lower cutting speeds


should be considered.
b. Unsatisfactory chip forms are expected when the other performances are optimized. For example, when optimizing the
maximum tool ank wear, all performances show major
decreases, while chip volume ratios show a major increase and
the reverse is also true.
c. Dry cutting of DSS can outperform wet cutting of DSS in terms
of total progressive tool wear, when appropriate cutting parameters are selected.
d. Generally, better chip forms, lower effective cutting power consumption, percentage increase in radial cutting force and lower
wear rates are expected when varying depth of cuts at wet cutting than dry cutting is considered.
e. When varying the cutting speed, the effective cutting power
consumption, percentage increase in radial cutting force and
maximum tool ank wear show the identical behavior in terms
of increase or decrease.
3.3. Phase II: Economics of facing operations
The effective optimization of machining process affects dramatically the cost and production time of machined components as well
as the quality of the nal product. This section presents the details
of multi-objective optimization of multi-pass constant speed facing
operation. MADM optimization approach is proposed to determine
the optimal cutting parameter which simultaneously minimizes
multiple objectives. Since, the machining parameters that giving
maximum production rate would not be identical to those giving minimum production cost, therefore, machine utilization time,
main time and cutting tool related costs are considered as potential objectives, in order to optimize the economics of the cutting
process.
3.3.1. Calculation of machine utilization time
The machine utilization time tbB is dened as the sum of all nominal times that a machine requires to accomplish a specied job. It is
composed of the time required to produce m units and total set-up
time trB :
tbB = m te + trB

(10)

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

10

Production time or time per unit te can be expressed in terms of


the basic time tg and the idle time as follow:
te = tg + ti

(11)

The basic time tg is the sum of the main process time th and
auxiliary process time tn .
tg = th + tn

(12)

3.3.2. Production cost


A typical production cost for a workpiece produced by turning
operations is comprised of machine costs, labor costs and tool costs:
KF = KM + KL + KW

The machine hour-rate describes the costs to be calculated of a


machine tool per hour. The Machine hour-rate KM is calculated as
follow:

The main cutting time for constant cutting speed facing per
each cutting trial in Table 3 can be calculated using the following
expression:

KM =

th = tc + Cf

JAS =

(13)

1
JAS

(14)

where d is the diameter of the workpiece in mm, lr is the additional


traveling radial distance of the cutting tool in mm and n is the maximum rotational speed of the workpiece which set as 3500 rpm.
The constant Cf is function of cut lengths which are considered in
conjunction with the respective feed velocity.
The auxiliary process time tn is the time during which all indirect
processes arising during the machining operation (e.g. tightening,
measuring, adjusting, pro rata tool change and workpiece change)
are executed. The idle time ti takes all pauses into consideration
during which the machine tools are not in operation and the total
time required for all irregular events such as procuring necessary
resources. In this study, the following relation is considered valid
for calculating the idle time:
ti = 0.3(th + tn )

(15)

Total setup time trB refers to the time required for machine setup tvM , tool change trW and nonproductive set-up activities trV . The
latter is often estimated as 30% of the machine set-up and tool
change time.
trB = tvM + trW + trV

(16)

For a batch of m workpieces per a machine, total tool change


time is dened as:
th
T

(17)

where tWZ is the time that passes till a single tool is changed, and
both the position correction and positioning for re-entry have taken
place. The tool life T is calculated based on tool wear criteria of
VBmax = 600 m. The developed RSM models of VBmax are utilized
to nd the maximum allowable length of cuts Lcmax when VBmax
is 600 m. Once the length of cuts are determined, they are substituted into Eq. 14 to nd the tool life for each cutting trial. The
overall working time ta per x number of machines is described as:
ta = te

m
x

(18)

Finally, the following relation is true for the machine utilization


time per workpiece or process:

tbB = m 1.3

(Lc /ap ) ((2lr + d) + (1000vc /n) )




4000
f vc

+ Cf + tn

+ tWZ

t
h
T

tl

+ kbW + kbZ + kbR

+ kbE

(21)

No. of operation hours


No. of working shifts
Week(w.)

No. of working weeks


Year

(22)

The yearly machine runtime JAS amounts for example to


16001800 h/a for single-shift operation. In the case of multi-shift
operation, the runtime is increased proportionately (e.g. two-shift
operation ca. 3200 h/a or three-shift operation ca. 4800 h/a). The
procurement cost kbB covers the purchasing, transportation and
installation costs. The time tl is dened as a timeframe at which the
machine is economically utilizable. The cost of maintenance and
repair services kbW can be expressed in terms of the percentage %p
of the procurement cost kbB :
kbW =

p
k
100 bB

(23)

Assumed interest rates can be set at the current value of the


machine also of a non-depreciated element as a calculated average
based on the procurement price at the full interest rate (%q):
kbZ = 0.5 %q kbB

(24)

To calculate the space cost kbR , planning estimations should consider the machine required area Q in m2 and the monthly rent A per
m2 :
kbR = 12 Q A

(25)

The operating costs kbE include the costs of operation energy,


illumination and coolant kk . The electricity costs can be estimated
based on effective cutting power PM , the standard cost of electricity
Ec in Euro/kWh and percentage %C2 of being in ON state:
kbE = PM Ec %C2 + kk

(26)

Labor costs KL are calculated as follows:


KL = Lm (1 + r)

(27)

where Lm is the gross hourly wage and r as the amount of nonwage costs of the operator. It is appropriate when a new factor that
summarizes the machine cost and the wage rate per hour in terms
of previously dened time scales is formulated as:
KML = KM +

trB + ta
KL
tbB

(28)

The costs of typical indexable carbide cutting tools are comprised of the cost of tool holders kWH , inserts kWP and spare parts
kET :
KW = kWH + kWP + kET

(29)

The total insert costs are dened as:


kWP =

+ trV + tvM

bB

(Lc /ap ) ((2lr + d) + (1000vc /n) )




4000
f vc

trW = m tWZ

k

Annual operating hours JAS of a machine is dened as:

Here, the cutting time tc is the time in which the tool is actually
cutting and determined by:
tc =

(20)

(19)

m th KWSP
0.8 T Zs

(30)

where KWSP is the cost of an insert, Zs is the number of usable cutting


edge per insert and 0.8 is a safety factor accounts for uncertainty

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

in tool life. The spare parts costs are often expressed in terms of
percentage of the tool holder and insert costs. The nal production
cost per unit is obtained by adding the terms in Eqs. (20)(35):

1
m th KWSP
KF =
t KM + (trB + ta ) KL + kWH +
+ kET
m bB
0.8 T Zs

(31)

Employing the previous relationships, one can dene the main


time-related costs as:
K1 =

KM +


1
x

KL

th

11

rating of alternative i, Ai , with respect to attribute j, Xj . Hence Ai ,


for i = 1, 2, 3, . . .,m is denoted by:
AF = (xi1 , xi2 , , xin )

(36)

and Xj for j = 1,2,3, . . .,n is denoted by:


Then the decision problem could be shown in matrix form:
Xj = (x1j , x2j , , xmj )

(37)

(32)

The xed or the workpiece-related costs as:


K2 =

1
m

KM +

1
KL
KL (trM + trV ) + KM +
x
x

(tn + tb + tvB ) (33)

and the tool related costs as:


K3 =

th
1
(KM + KL )(tW ) + KW
T
m

(34)

The production cost can also be determined using [46,51]:


KF = K1 + K2 + K3

(35)

3.3.3. Case study


To illustrate the application of the equations presented in previous two sections, productions of 12,000 identical components were
considered as a case study. The machining of each unit has to be
accomplished through removing a total cutting length of 12 mm
under constant cutting speed facing operation. Eqs. (10)(35) were
employed to dene each machining economics terms presented
in this study. The objective is to simultaneously minimize the
conicting objectives such as machine utilization time tbB , main
time-related costs K1 and the tool related costs K3 using a multiple attribute decision making (MADM) method called TOPSIS. In
conjunction with experimentations and utilizing the feature of simulation on the CNC-controller, the main cutting times of the trials
were accurately recorded. The data considered for the case study
are given below:
kbB = 100,000 D
m = 12,000
tn = 1.5 min
tvM = 30 min
tWZ = 0.5 min
tl = 2 years
d = 55 mm
lr = 6.875 mm
p = 10

Q = 40 m2
A = 30 D /m2 .Mo.
Ec = 0.130 D /kWh
kk = 0.90 D /h
Lm = 34 D /h
r = 2.5
kWH = 80D
Zs = 4
q = 10.5

KWSP = 6 D
x=1
PM = 25 kW
%C2 = 30
h/week = 40
week/annum = 40
working shifts/day = 1.9

The problem can be dened in the context of multi-objective


optimization. A multiple attribute decision making method known
as Technique for Order Preference by Similarity to Ideal Solution
(TOPSIS) is proposed to convert the multi-objective optimization
of three objectives into a single objective optimization problem.
The technique is based on the concept that the chosen alternative
should have the shortest Euclidean distance from the ideal solution.
The ideal solution is a hypothetical solution for which all attribute
values correspond to the maximum attribute values in the database
comprising the satisfying solutions; the negative-ideal solution is
the hypothetical solution for which all attribute values correspond
to the minimum attribute values in the above-mentioned database.
TOPSIS, thus, gives a solution that is not only closest to the hypothetically best, but which is also farthest from the hypothetically
worst. The steps involved for calculating the TOPSIS values are as
follows:
Step 1. Determination of decision matrix: A decision matrix (D)
is a (m n) matrix whose element (xij ) indicates the performance

(38)
Rij =

xij

(39)

x2
i=1 ij

Step 2. Construct the normalized decision matrix, Rij . This can be


represented as:

vij = wj rij

(40)

Step 3. Construct the weighted normalized decision matrix. This


is obtained by the multiplication of each element of the column of
the matrix Rij with its associated weight wj . Hence, the elements
of the weighted normalized matrix vij are expressed as:
Step 4. Determine the ideal (best) and negative ideal (worst)
solutions. The ideal (best) and negative ideal (worst) solution can
be expressed as:
A = {maxi (vij )if j J; mini (vij ) if j J  , i = 1, 2, . . ., m}
Positive ideal solution =

v1 , v1 , , vj , , vn

A = {mini (vij ) if j J; maxi (vij ) if j J  , i = 1, 2, . . ., m}


Negative ideal solution =

v
, v
, , v
, , v
n
1
1
j

(41)

(42)

Step 5. Obtain the separation measures. The separation of each


alternative from the ideal one is given by Euclidean distance.
Therefore, the separation from positive ideal alternative is:



 n
Si = 
(vij v )2
i

(43)

j=1

similarly, the separation from the negative ideal alternative is:



 n
2
S = 
(vij v )
i

(44)

j=1

Step 6. The relative closeness of a particular alternative to the ideal


solution Ci can be evaluated as:
Ci =

Si

Si

+ Si

(45)

i = 1, 2, . . ., m; 0 Ci 1

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model
JMP-244; No. of Pages 17
12

ARTICLE IN PRESS
R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

Fig. 7. Fuzzy Inference System prediction of OSI.


Fig. 6. The overlapping circles model of sustainable development.

3.4. Phase III: Operational sustainability index (OSI)

simple, it is appealing because manageable number of variables,


and as relatively low uncertainty can be achieved by measurement of local manufacturing operations [57]. In this study, the
applications of sustainability principles in manufacturing processes
are presented using machining as an example. With the implementation of sustainability principles in machining technologies,
end-users have the potential to reduce the cost, enhance operational safety and reduce power consumption.
Operational sustainability in machining can be dened in terms
of the cost of machining, power consumption and chip volume ratio
which greatly affect the waste management. Based on these three
interacting and contradicting elements, a compromise solution has
to be introduced for a comprehensive evaluation of machining
operational sustainability. To obtain this sustainability measure,
fuzzy logic system (see Fig. 7.) is employed to combine the total
production cost per unit KF , effective cutting power Pe and chip
volume ratios R of each experimental trial into a single sustainability characteristics index called Operational Sustainability Index
(OSI). Eqs. (10)(34) and constants presented in the case study of
phase II were respectively employed to calculate the total production cost per unit KF . Matlab software was used to construct the
inference model of the OSI. The three performance values were
rst adjusted to a notionally common scale between null and one,
using simple normalization methods, so that the digit one represents the most desirable and null is the least desirable alternative.
Small (S), Medium (M) and Large (Lg) fuzzy sets are assigned to
the performances. The sustainability index has the following seven
levels: Very Low (VL), Low (L), Lower Medium (LM), Medium (M),
Upper Medium (UM), High (H), Very High (VH). Mamdani implication method is employed for the fuzzy inference reasoning. The
relationship between system input and output is expressed by an
If-Then type. Totally 27 fuzzy rules per material were formulated.
The predicted values of OSI are presented in Fig. 8 and the following conclusions are extracted:

Originally, sustainability relates to forestry. In a very broad and


fuzzy denition, forestry is called sustainable, if just as much timber is cut down as can replenish to maintain the basis of life for
future generations [54]. A more widely accepted general denition of sustainable development is provided by the United Nations
Brundtland Commission in 1987: development that meets the needs
of the present without compromising the ability of future generations
to meet their own needs [55]. Based on this view, the United Nations
2005 World Summit Outcome document refers to the independent and mutually reinforcing pillars of sustainable development
as economic development, social development and environmental
protection (see Fig. 6.) [56].
In the context of manufacturing, sustainability means the ability to produce specic product operations and the circulation
of resources at the rate of production. Although improving sustainability through manufacturing process optimization is far from

a. Generally, higher OSI values were noticed when cutting EN


1.4462 is performed.
b. The average wet cutting OSI is 10% higher than dry cutting due
to the fact that the production cost, the effective cutting power
and the chip volume ratios were lower in the wet cutting.
c. Lower cutting speed, intermediate feed rate and depth of cut
ranges, and higher cutting speed, intermediate feed rate and
lower depth of cut ranges tend to maximize OSI in dry and wet
cutting, respectively.
d. OSI deteriorates with increasing number cutting passes (Lc /ap ),
since all non-benecial performances are expected to increase
as Lc /ap increases.
e. For the equal cutting length scenario of facing EN 1.4462 and
EN 1.4410, the maximum OSIs are obtained when alternatives
number 8 are selected.

Step 7. Rank the preference order, so that the alternative that has
the shortest distance to the ideal solution is ranked rst [17,52,53].
The data given in Table 3 are represented as matrix D48 4 for
each material case. The matrix is not shown here as it is nothing
but the repetition of data given in Table 3, which is represented in
a matrix form. The attribute weights for machine utilization time
tbB , main time-related costs K1 and the tool related costs K3 were
0.45, 0.3, 0.25, respectively. The normalized decision matrix and
then the weighted normalized matrix are determined by using Eqs.
(39) and (40) respectively. The positive ideal solution (A*) and the
negative ideal solution (A ) could be found by Eqs. (41) and (42).
Eqs. (43) and (44) are used to determine the separation measures.
Finally, Eq. (45) is utilized to calculate the relative closeness to the
ideal solution Ci . The results of performing TOPSIS are summarized
in Table 6.
The rankings of alternatives by their corresponding TOPSIS
indices (see Fig. 5.) revealed the favourability of adapting wet
cutting for a simultaneous minimization approach. This is mainly
attributed to the fact that when the components are dry machined
the cost of increasingly tool wear rate and power consumption
overtakes the advantage of not employing cutting uids. It is also
seen that the order of ranking of both materials is almost identical.
Because of the higher effective power consumption and tool wears
rate at higher cutting speed and higher feed rate ranges and the
worst chip morphology at lower ranges, an intermediate range has
given the preference over both ranges. It is also noticed that the
lower cutting speed when dry cutting EN 1.4410 and the higher
cutting speed when wet cutting EN 1.4462 has given higher preference than the similar dry cutting EN 1.4462 and wet cutting EN
1.4410, respectively.

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

EN 1.4462

EN 1.4410

vc (m/min)

f (mm/r)

ap (mm)

Lc (mm)

Cutting cond.

tbB h/year

K1 D /unit

K3 D /unit

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48

200
75
75
200
200
200
75
137.5
75
200
75
200
200
200
75
75
75
200
200
200
75
137.5
137.5
75
75
137.5
200
106.25
137.5
200
75
200
200
75.0
137.5
200
75
168.75
75
200
200
137.5
200
75
75
75
75
200

0.25
0.25
0.25
0.25
0.25
0.10
0.17
0.17
0.25
0.10
0.10
0.25
0.10
0.10
0.25
0.10
0.25
0.25
0.25
0.25
0.10
0.17
0.17
0.25
0.10
0.10
0.10
0.21
0.10
0.25
0.17
0.17
0.17
0.10
0.25
0.10
0.10
0.14
0.17
0.10
0.25
0.25
0.10
0.25
0.10
0.25
0.10
0.17

1.5
1.5
1.5
0.5
0.5
0.5
1
1.5
1
1.5
1.5
0.5
1.5
0.5
1.5
0.5
0.5
1.5
1
1.5
1.5
1
1
0.5
0.5
1
1.5
0.75
1.5
1.5
1.5
1
1.5
0.5
1.5
0.5
1
0.75
0.5
0.5
0.5
0.5
1
0.5
0.5
0.5
1.5
0.5

12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12
12

Wet
Dry
Wet
Wet
Wet
Wet
Wet
Wet
Dry
Wet
Wet
Wet
Wet
Wet
Wet
Dry
Wet
Dry
Dry
Wet
Wet
Wet
Wet
Wet
Wet
Wet
Dry
Dry
Dry
Dry
Dry
Wet
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Dry
Wet
Wet
Dry
Dry
Dry

734.477
959.736
959.220
1261.236
1261.236
1969.404
1496.224
877.472
1230.424
943.050
1562.386
1261.236
943.050
1969.404
959.220
3910.637
1990.176
748.012
901.821
734.477
1562.386
1095.527
1095.527
1990.176
3842.968
1443.586
992.902
1379.183
1140.819
748.012
1155.752
956.205
819.295
3910.637
791.462
2082.064
2173.947
1383.132
2614.182
2082.064
1334.430
1497.952
1272.385
1990.176
3842.968
2026.145
1588.641
1558.531

2.543
4.880
4.910
7.629
7.629
14.652
10.138
4.205
7.320
4.884
10.802
7.629
4.884
14.652
4.910
32.207
14.730
2.527
3.791
2.543
10.802
6.308
6.308
14.730
32.405
9.692
4.854
8.578
6.422
2.527
6.717
4.916
3.257
32.207
3.155
14.562
16.104
8.394
20.151
14.562
7.582
9.464
7.281
14.730
32.405
14.640
10.736
9.772

1.581
1.043
1.035
1.334
1.334
0.587
0.688
0.884
1.270
0.752
0.600
1.334
0.752
0.587
1.035
1.882
0.850
1.862
2.266
1.581
0.600
0.788
0.788
0.850
0.459
0.572
1.800
1.581
1.397
1.862
1.050
0.958
1.726
1.882
1.370
2.957
1.435
2.075
1.625
2.957
2.871
2.094
2.240
0.850
0.459
1.605
1.151
2.678

0.845
0.885
0.886
0.791
0.791
0.609
0.749
0.918
0.804
0.914
0.729
0.791
0.914
0.609
0.886
0.389
0.604
0.813
0.764
0.845
0.729
0.869
0.869
0.604
0.370
0.766
0.800
0.748
0.814
0.813
0.838
0.894
0.825
0.389
0.869
0.566
0.558
0.715
0.472
0.566
0.675
0.693
0.723
0.604
0.370
0.592
0.714
0.653

tbB h/year

K1 D /unit

K3 D /unit

Ci

748.025
985.958
971.803
1288.984
1288.984
1985.940
1512.126
884.971
1252.431
952.169
1571.265
1288.984
952.169
1985.940
971.803
3922.912
2016.453
795.556
942.674
748.025
1571.265
1109.346
1109.346
2016.453
3860.786
1455.102
1031.209
1402.554
1167.960
795.556
1178.422
972.606
860.453
3922.912
826.793
2128.465
2188.291
1412.783
2631.993
2128.465
1383.321
1530.679
1305.274
2016.453
3860.786
2047.829
1608.526
1603.641

2.543
4.880
4.910
7.629
7.629
14.652
10.138
4.205
7.320
4.884
10.802
7.629
4.884
14.652
4.910
32.207
14.730
2.527
3.791
2.543
10.802
6.308
6.308
14.730
32.405
9.692
4.854
8.578
6.422
2.527
6.717
4.916
3.257
32.207
3.155
14.562
16.104
8.394
20.151
14.562
7.582
9.464
7.281
14.730
32.405
14.640
10.736
9.772

1.866
1.595
1.300
1.919
1.919
0.936
1.023
1.042
1.733
0.945
0.787
1.919
0.945
0.936
1.300
2.141
1.405
2.862
3.125
1.866
0.787
1.079
1.079
1.405
0.835
0.814
2.606
2.072
1.968
2.862
1.527
1.304
2.592
2.141
2.114
3.933
1.737
2.699
2.000
3.933
3.900
2.783
2.932
1.405
0.835
2.062
1.570
3.627

0.886
0.882
0.904
0.792
0.792
0.607
0.748
0.941
0.810
0.924
0.728
0.792
0.924
0.607
0.904
0.385
0.602
0.799
0.771
0.886
0.728
0.873
0.873
0.602
0.373
0.765
0.799
0.758
0.818
0.799
0.839
0.904
0.817
0.385
0.860
0.569
0.561
0.728
0.472
0.569
0.681
0.701
0.737
0.602
0.373
0.595
0.716
0.658

ARTICLE IN PRESS

Component

Ci

G Model

JMP-244; No. of Pages 17

Independent variables

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx


13

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

Table 6
Results of multi-objective optimization using TOPSIS.

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

14

Fig. 8. Process parameters interaction effects on OSI.

3.4.1. Optimization of OSI using CSNNS


Due to the highly complex relation between cutting parameters
and OSI, the RSM was seen no longer efcient to accurately predict
the values of OSI. Serious divergences were noticed between experimental data and predicted values for several points. Therefore, a
multilayer perceptron (MLP) articial neural network (ANN) which
can describe the relationships with more precision was integrated
with Cuckoo Search meta-heuristic algorithm to perform the task of
modeling and optimization of OSIs. The neural networks have two
layers: one hidden layer and one output layer. The hidden layer
uses a sigmoid-type transference function:
f (x) =

1
1 + exp[(b +

wi xi )]

(46)

while the output layer uses a linear function:


Outputpredicted = b +

wi xi

(47)

where w an b are the weights and biases of the network respectively. To facilitate the neural network training process, all the
inputs were normalized using the following equation:
2(x xmin )
xi =
1
(xmax xmin )

(48)

This normalization maps all the inputs and OSI between 1 and
+1. The ANN architecture consists of 4 neurons in the input layer, 1
neuron in the output layer. The weights and biases of the network
are initialized to small random values to avoid immediate saturation in the respective functions. The network was trained by using
gradient descendent with momentum back propagation algorithm.
In this algorithm four parameters must be tuned: learning rate LR ,
momentum constant Mc , training epochs Ep and number of hidden
neurons Hn . For this tuning the Taguchi design L9 (34 ) was used to
nd the most convenient values for achieving no only lower root
mean square, but also good generalization capability, giving the
following values: LR = 0.0705, Mc = 0.5895, Ep = 5000 and Hn = 7. The
performance of the ANNs was statistically measured by the root
mean squared error (RMSE), the coefcient of determination (R2 )
and the absolute average deviation ( avg ) obtained as follows:


 n
1
RMSE = 
[(yactual )i (ypredicted ) ]2
i

(49)

i=1

R2 = 1

n
[(yactual )i (ypredicted )i ]2
i=1

n
2
i=1

[(yactual )i (ymean )i ]

(50)

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model

ARTICLE IN PRESS

JMP-244; No. of Pages 17

R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx


vc

15

f
ap

Neural Network
Model

Lc

OSI

Cuckoo Search
Algorithm

Independent Variables

Maximum
OSI
Fig. 9. Cuckoo Search Neural Network System.
Fig. 10. Maximization of OSI with CSNNS.


avg

1
=
n

 
n 

 (yactual )i (ypredicted )i 

 100


(yactual )i

(51)

i=1

where yactual represents the actual, ypredicted obtained by neural network and ymean the mean of actual OSI values respectively, and n
is the number of experimentation trials. It must be remarked that
number of hidden neurons guarantee that there are more training
samples than the total amount of free parameters, thus the training process is mathematically determined. The R-squared statistics
of the models were generally greater than 0.99, which indicate
that the models as tted explain over 99% of the variability in
OSI. The trained networks achieved RSME and  avg values below
target (0.0001) and (1%) respectively. It must be mentioned that
the relationships between variables are complex, which prove the
application of articial neural networks very advantageous.
Neural network models are then integrated with the cuckoo
search optimization algorithm, so that solutions which will provide
useful information to the user during the selection of machining
parameters are obtained. The architecture of the cuckoo search
neural network system (CSNNS) is shown in Fig. 9. CS outperforms
many existing algorithms such as genetic algorithm and particle
swarm optimization. This superiority can be attributed to the fact
that cuckoo search uses a combination of vectorized mutation,
crossover by permutation and Levy ights and selective elitism
among the best solutions. In addition, the not-so-good solutions
can be replaced systematically by new solutions, and new solutions
are often generated by preferring quality solutions in the solution
sets. Thus, the mechanism of the overall search moves is more subtle and balanced, compared with the simple mechanism used in
particle swarm optimization [58]. The number of host nests (or the
population size n) and the probabilities (Pa ) were tuned using trial
and error method. The population size of (25) and probability of
(0.25) were found sufcient in this case.
The selected decision variables were cutting parameters for each
process condition. They were dened for the ranges between the
minimum and maximum experimental levels presented in Table 2.
OSIs were selected as the objective functions to maximize:
OSIi =

i (vc , f, ap , Lc )

(52)

where i are the neural-network-based models. The considered


constraints are the percentage increase in radial cutting force (%Fr )
and the arithmetic average roughness (Ra ). The latter is dened as:
Ra =

0.032f 2
r

(53)

where Ra is in m and r is the tool nose radius in mm. In this


case, the optimum cutting settings should always lead to the results
of preferably smaller or equal to the respective values of 15% and
2 m.
CSNNS optimization of OSI has to yield minimum production
cost, minimum effective cutting power and the best chip morphology, while considering technological constraints. Fig. 10 shows the
performance of proposed CSNNS. Total computation time were less
than 3 min with an Intel Xeon CPU 3.47 GHz and 24 GB RAM
computer. Less than 2000 iterations were sufcient to reach to
the global optimums in each case. It is evident that the developed CSNNS is very efcient and highly reliable approach for the
selection of optimum control parameters.
The obtained optimization results are listed in Table 7. Generally, the following conclusion point can be depicted:

a. Higher optimum cutting speeds were more often to observe in


wet cutting process than in dry cutting process which promotes
higher production rates.
b. Based on the estimated optimum OSIs, the machinability of EN
1.4462 is higher than the machinability of EN 1.4410.
c. Wet cutting of EN 1.4462 and EN 1.4410 outperforms their
respective dry cutting in operational sustainability by 9.768% and
12.383%, respectively.
d. While wet cutting generally gives lower non-benecial performance values, this conclusion is hard to notice due to the different
adopted optimum cutting speeds and feed rates.
e. The magnitudes of surface roughness and chip volume ratio are
showing similar trends.
f. The coincidence between optimum cutting conditions of the rst
rank TOPSIS and those of the optimized OSI is observed.

Table 7
Results of CSNNS.
Material

Process condition

vc (m/min)

f (mm/rev)

ap (mm)

Lc (mm)

EN
1.4462

Wet
Dry

156.283
91.793

0.1359
0.1848

1.500
1.500

12
12

EN
1.4410

Wet
Dry

146.30
82.3259

0.1286
0.2019

1.500
1.500

12
12

%Fr
6.6477
9.9772
14.878
13.204

Pe (W)

VBmax (m)

OSI

2218.1
1354.6

104.19
48.049

69.073
105.6

0.7105
0.6411

2066.7
1.4175

106.401
62.135

97.052
100.7

0.6969
0.6106

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model
JMP-244; No. of Pages 17
16

ARTICLE IN PRESS
R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

4. Conclusions
Machining of duplex stainless steel grades such as EN 1.4462
and EN 1.4410 DEDM has been systematically investigated under
a multi-pass constant cutting speed facing operation. In the rst
phase of the investigation, D-optimal experimental design is used
extensively to investigate the effect of process variables on performance characteristics such as percentage increase in radial cutting
forces, effective cutting power, maximum tool ank wear and chip
volume ratio. Based on RSM, effective empirical relationships to
predict performance characteristics at 95% condence level were
developed. ANOVA used to check the adequacy of the models. The
models were then analyzed using 3D surface graphs and used to
study the interaction effects of process parameters. At the end of
the rst phase, constrained cuckoo search algorithm is selected to
perform optimization of the performance characteristics thereby
dening the optimum process conditions. The following general
conclusions can be drawn from the rst phase of the investigation:

a. The values of obtained %Fr when EN 1.4410 steels are cut were
higher than EN 1.4462 steels. They have shown direct linear
correlations with maximum tool ank wear and proportional
dependency on Lc /ap ratio, and approached their minimum values as cutting speed and feed rate increased to certain specied
limit. However, when the materials are cut dry, lower cutting
speed and feed rates than wet cutting are preferred to minimize
%Fr .
b. Generally, the two-factor interaction models were seen accurate
enough to explain the dependency relation between effective cutting power and independent variables. Their minimum
consumption were seen when the feed rate is in the range
0.1250.175 mm/rev of dry cutting and 0.10.15 mm/rev of wet
cutting and the rest of the remained independent variables are
kept at their lowest levels.
c. Wet cutting at cutting parameter ranges of vc 100160 m/min,
f 0.150.25 mm/rev, ap 0.751 mm for machining EN
1.4462
and
vc 120160 m/min, f 0.150.20 mm/rev,
ap 1.251.5 mm for machining EN 1.4462 were seen optimum in minimizing the tool ank wear. Dry machining
is possible when cutting parameters are set appropriately. The settings include; machining of EN 1.4462 at
vc 75110 m/min,
f 0.150.2 mm/rev,
ap 11.5 mm
and maximum Lc /ap 6, and machining of EN 1.4410 at
vc 7590 m/min, f 0.1250.175 mm/rev, ap 0.81.35 mm
and maximum Lc /ap 4.
d. Ribbon and snarled chip forms were common when the DSSs are
machined at feed rate and depth of cuts lower than 0.15 mm/rev
and 1 mm, respectively. In wet cutting, friendlier-to-machine
chip forms were produced and the R values of cutting EN 1.4410
were generally seen higher than of cutting EN 1.4462. Under the
same machining conditions which might produce continuous
chips, different chip forms were also possible, especially when
Lc /ap 6.
In the second phase of the study, the measurements were used
to develop a comprehensive machining economics model. A case
study of producing 12000 units of per each experimental run is
considered and the corresponding machining costs and production rates were determined. Summarizing the conclusion points
depicted at this stage:

a. The optimization conict between machining economics


attributes, such as simultaneous minimization of machine utilization time tbB , main time-related costs K1 and the tool

related costs K3 could be effectively solved employing a multiple


attribute decision making approach called TOPSIS.
b. The alternatives were ranked based on their computed relative
closeness to the ideal solution C* . The ranking of the alternatives
has revealed that the intermediate range of cutting speed, feed
rates at wet cutting produce the optimum choice to minimize
the considered attributes simultaneously.
c. The order of ranking of both materials is almost identical. However, under the same cutting condition, lower cutting speed
when dry cutting EN 1.4410 and higher cutting speed when wet
cutting EN 1.4462 has shown higher preference than similar dry
cutting EN 1.4462 and wet cutting EN 1.4410, respectively.
In the third phase of the study, the computed performances in
the rst and second phases were utilized to derive a new index
of measuring machining sustainability called operational sustainability index (OSI). Based on the Mamdani implication method for
the fuzzy inference reasoning, normalized production cost per unit
to consider the economics of the machining process, normalized
effective cutting power to assess the energy demand of the machining process and normalized chip volume ratio to consider the chip
morphology were successfully employed to dene the OSI. The
optimal machining parameters were tabulated and many conclusion points were extracted:
a. To accurately model and constrainedly optimize the highly nonlinear OSIs, neural network models could be integrated with
cuckoo search algorithm and form cuckoo search neural network
system (CSNNS). Results have indicated the efciency of the proposed approach for solving the optimization problem effectively.
b. Generally, higher OSI values were noticed when wet cutting of
DSSs are performed. For instance, in this study, wet cutting of EN
1.4462 and EN 1.4410 outperforms their respective dry cutting
in operational sustainability by 9.768% and 12.383% respectively.
c. Based on the predicted OSIs values, the machinability of EN
1.4462 is higher than the machinability of EN 1.4410.
d. Lower cutting speed, intermediate feed rate depth of cut ranges,
and higher cutting speed, intermediate feed rate and lower depth
of cut ranges tend to maximize OSI in dry and wet cutting respectively.
e. Results also showed coincidence between optimum cutting conditions in second and third phases.
Acknowledgments
The authors would like to acknowledge the signicant contribution of the workshop master Rolf Bauer for his important
experimental contributions as well as the many valuable suggestions made by Dr. Michael Schaal at the Institute for Machine Tools,
Stuttgart University.
References
[1] Armas IA, Moreuil SD. Duplex stainless steels. UK/USA: ISTE Ltd./John Wiley &
Sons Inc.; 2009. pp. 1.
[2] McGuire M. Stainless steels for design engineers. OH, USA: ASM International;
2008. p. 94.
[3] Chien WT, Tsai CS. The investigation on the prediction of tool wear and the
determination of optimum cutting conditions in machining 17-4PH stainless
steel. J Mater Process Technol 2003;140:3405.
[4] Korkut I, Kasap M, Ciftci I, Seker U. Determination of optimum cutting parameters during machining of AISI 304 austenitic stainless steel. Mater Des
2004;25(4):3035.
[5] Endrino J, Fox GR, Gey C. Hard AlTiN, AlCrN PVD coatings for machining of
austenitic stainless steel. Surf Coat Technol 2006;200(24):68405.
[6] Maranho C, Davim JP. Finite element modelling of machining of AISI 316 steel:
numerical simulation and experimental validation. Simul Model Pract Theory
2012;18(2):13956.

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

G Model
JMP-244; No. of Pages 17

ARTICLE IN PRESS
R.D. Koyee et al. / Journal of Manufacturing Processes xxx (2014) xxxxxx

[7] Sahoo AK, Sahoo B. Experimental investigations on machinability aspects in


nish hard turning of AISI 4340 steel using uncoated and multilayer coated
carbide inserts. Measurement 2012;45(8):215365.
[8] Ahmadi E, Homami RM, Rahmati S. Experimental investigation and mathematical modeling of the composite ceramic cutting tools with alumina base in the
machining process of PH-hardened austenitic-ferritic (Duplex) stainless steel.
Int J Adv Des Manuf Technol 2012;5(2):1725.
[9] Kurama E, Ozcelik B, Bayramoglu M, Demirbas E, Simsek BT. Optimization of
cutting uids and cutting parameters during end milling by using D-optimal
design of experiments. J Clean Prod 2013;42:15966.
[10] Iqbal A, Ning H, Khan I, Liang L, Dar NU. Modeling the effects of cutting parameters in MQL-employed nish hard-milling process using D-optimal method. J
Mater Process Technol 2008;199(13):37990.
[11] Arokiadass R, Palaniradja K, Alagumoorthi N. Prediction and optimization of
end milling process parameters of cast aluminium based MMC. Trans Nonferrous Met Soc China 2012;22(7):156874.
[12] Kao P, Hocheng H. Optimization of electrochemical polishing of stainless steel
by grey relational analysis. J Mater Process Technol 2003;140(1):2559.
[13] Walia R, Shan H, Kumar P. Multi-response optimization of CFAAFM process through Taguchi method and utility concept. Mater Manuf Process
2006;21(8):90714.
[14] Chang CK, Lu HS. Design optimization of cutting parameters for side milling
operations with multiple performance characteristics. Int J Adv Manuf Technol
2007;32(12):1826.
[15] Gaitonde V, Karnik S, Achyutha B, Siddeswarappa B. Taguchi optimization
in drilling of AISI 316L stainless steel to minimize burr size using multiperformance objective based on membership function. J Mater Process Technol
2008;202(13):3749.
[16] Gaitonde V, Karnik S, Davim JP. Multiperformance optimization in turning of
free-machining steel using taguchi method and utility concept. J Mater Eng
Perform 2009;18(3):2316.
[17] Singh A, Datta S, Mahapatra SS. Application of TOPSIS in the Taguchi method for
optimal machining parameter selection. J Manuf Sci Prod 2011;11(13):4960.
[18] Kaladhar M, Subbaiah KV, Rao CS, Rao KN. Application of taguchi approach and
utility concept in solving the multi-objective problem when turning AISI 202
austenitic stainless steel. J Eng Sci Technol Rev 2011;4(1):5561.
[19] Vijayakumar K, Prabhaharan G, Asokan P, Saravanan R. Optimization of multipass turning operations using ant colony system. Int J Mach Tools Manuf
2003;43(15):16339.
[20] Samanta S, Chakraborty S. Parametric optimization of some non-traditional
machining processes using articial bee colony algorithm. Eng Appl Artif Intell
2011;24(6):94657.
[21] Zarei O, Fesanghary M, Farshi B, Saffar RJ, Razfar MR. Optimization of multipass face-milling via harmony search algorithm. J Mater Process Technol
2009;209(5):238692.
[22] DAddona DM, Teti R. Genetic algorithm-based optimization of cutting parameters in turning processes. Procedia CIRP 2013;7:3238.
[23] Yildiz AR. Hybrid Taguchi-differential evolution algorithm for optimization of
multi-pass turning operations. Appl Soft Comput 2013;13(3):14339.
[24] Rao RV, Kalyankar VD. Multi-pass turning process parameter optimization using teachinglearning-based optimization algorithm. Sci Iran
2013;20(3):96774.
[25] Tzeng Y-C, Chen F-C. Multi-objective optimisation of high-speed electrical discharge machining process using a Taguchi fuzzy-based approach. Mater Des
2007;28(4):115968.
[26] Gupta A, Singh H, Aggarwal A. Taguchi-fuzzy multi output optimization
(MOO) in high speed CNC turning of AISI P-20 tool steel. Expert Syst Appl
2011;38(6):68228.
[27] Yang YS, Huang W. A grey-fuzzy Taguchi approach for optimizing multiobjective properties of zirconium-containing diamond-like carbon coatings.
Expert Syst Appl 2012;39(1):74350.
[28] Chien WT, Chou CY. The predictive model for machinability of 304 stainless
steel. J Mater Process Technol 2001;118:4427.
[29] Karpat Y, zel T. Multi-objective optimization for turning processes using
neural network modeling and dynamic-neighborhood particle swarm optimization. Int J Adv Manuf Technol 2007;35(34):23447.
[30] Karnik SR, Gaitonde VN, Davim JP. A comparative study of the ANN and RSM
modeling approaches for predicting burr size in drilling. Int J Adv Manuf Technol 2008;38(910):86883.

17

[31] Paro J, Hnninen H, Kauppinen V. Tool wear and machinability of HIPed P/M
and conventional cast duplex stainless steels. Wear 2001;249(34):27984.
[32] Bordinassi E, Stipkovic M, Batalha G, Delijaicov S, de Lima NB. Supercial
integrity analysis in a super duplex stainless steel after turning. J Achiev Mater
Manuf Eng 2006;18(12):3358.
[33] Krlczyk G, Legutko S, Gajek M. Predicting the surface roughness in the dry
machining of duplex stainless steel (DSS). Metalurgija 2013;52(2):25962.
[34] Krlczyk G, Gajek M, Legutko S. Effect of the cutting parameters on tool life in
duplex stainless steel turning process. Tehnicki vjesnik 2013;20(4):58792.
[35] Krlczyk G, Gajek M, Legutko S. Predicting the tool life in the dry machining of
duplex stainless steel. Maint Reliab 2013;15(1):625.
[36] Nomani J, Pramanik A, Hilditch T, Littlefair G. Machinability study of rst generation duplex (2205), second generation duplex (2507) and austenite stainless
steel during drilling process. Wear 2013;304(12):208.
[37] de Oliveira Jr CA, Diniz AE, Bertazzoli R. Correlating tool wear, surface roughness
and corrosion resistance in the turning process of super duplex stainless steel.
J Braz Soc Mech Sci Eng 2013, http://dx.doi.org/10.1007/s11665-013-0832-4.
[38] Selvaraj DP, Chandramohanb P, Mohanraj M. Optimization of surface roughness, cutting force and tool wear of nitrogen alloyed duplex stainless steel in a
dry turning process using Taguchi method. Measurement 2014;49:20515.
[39] Krlczyk G, Nieslony P, Legutko S. Microhardness and surface integrity in turning process of duplex stainless steel (DSS) for different cutting conditions. J
Mater Eng Perform 2014;23(3):85966.
[40] Marksberry PW, Jawahir IS. A comprehensive tool-wear/tool-life performance
model in the evaluation of NDM (near dry machining) for sustainable manufacturing. Int J Mach Tool Manuf 2008;48(78):87886.
[41] Jayal AD, Badurdeen F, Dillon OW, Jawahir IS. Sustainable manufacturing: modeling and optimization challenges at the product, process and system levels.
CIRP J Manuf Sci Technol 2010;2(3):14452.
[42] Heisel U, Stehle T. Energy efciency in metal cutting production. Int SciConvenient Conf Mod Eng Sci Edu 2011;5(1):38598.
[43] Hana I, Khamlichi A, Cabrera FM, Almansa E, Jabbouri A. Optimization of cutting conditions for sustainable machining of PEEK-CF30 using TiN tools. J Clean
Prod 2012;33:19.
[44] Heisel U, Braun S. Prognose des prozessabhngigen Energieverbrauchs
Simulations- and Prognosemodell des prozessabhngigen Energieverbrauchs von Werkzeugm-aschinen Teil 1. Werkstattstechnik Online
2013;103(1):1621.
[45] Anderson-Cook CM, Borror CM, Montgomery DC. Response surface design evaluation and comparison. J Stat Plan Inference 2009;139:62941.
[46] Klocke F. Manufacturing processes 1, cutting. Heidelberg, Germany: SpringerVerlag; 2011. p. 33951.
[47] Myers RH, Montgomery DC, Anderson-Cook CM. Response surface methodology process and product and product optimization using designed
experiments. New Jersey, USA: John Wiley & Sons; 2009.
[48] Yang XS. Nature-inspired metaheuristic algorithms. UK: Luniver Press; 2010.
p. 6583.
[49] Rajabioun R. Cuckoo optimization algorithm. Appl Soft Comput
2011;11(8):550818.
[50] Tschtsch H. Applied machining technology. Heidelberg, Germany: Springer;
2008. p. 702.
[51] Paucksch E, Holsten S, Lin M, Tikal F. Zerspantechnik. Wiesbaden, Germany:
Vieweg Teubner Verlag; 2008. p. 936.
[52] Lai YJ, Hwang CL. Fuzzy multiple objective decision making methods and applications. Berlin: Springer-Verlag; 1994. p. 119.
[53] Chittaranjan D, Srinivas C. Evaluation of metal strip-layout selectionusing AHP
and TOPSIS method. Adv Mater Manuf Charact 2013;3(1):4259.
[54] Kandisch D, Hermann P, Meier C. Sustainable process management-status Quo
and perspectives. In: Advanced Manufacturing and Sustainable Logistics: 8th
International Heinz Nixdorf Symposium, 44. 2010. p. 94107.
[55] UN World Commission on Environment and Development. Our Common
Future. Oxford: Oxford University Press; 1987.
[56] Chick A, Micklethwaite P. Design for sustainable change: how design and
designers can drive the sustainability agenda. China: AVA Publishing; 2011.
[57] Nahkala S. Aligning product design methods and tools for sustainability. In:
Re-engineering Manufacturing for Sustainability: Proceedings of the 20th CIRP
International Conference on Life Cycle Engineering. 2013. p. 539.
[58] Yang XS, Deb S. Multiobjective cuckoo search for design optimization. Comput
Oper Res 2013;40(6):161624.

Please cite this article in press as: Koyee RD, et al. Modeling and optimization of turning duplex stainless steels. J Manuf Process (2014),
http://dx.doi.org/10.1016/j.jmapro.2014.05.004

S-ar putea să vă placă și