Step 1 Array of responses during machining [11]

A=

x x

x x

Where u is the number of trails and v is the number of responses, A is an array of responses.
Trails

1
2
3
4
5
6
7
8
9

Open
Voltage
(A)
1
1
1
2
2
2
3
3
3

Servo
Voltage
(B)
1
2
3
1
2
3
1
2
3

Wire Feed
(C)
1
2
3
2
3
1
3
1
2

Material Removal
Rate (MRR)
(mm3/min)
78.9600
62.8800
55.1200
79.6000
65.9200
51.2000
75.6000
58.3200
49.2800

Machining Time
(MT) (min)

Gap Voltage
(GV) (Volts)

5.2160
6.4500
7.3170
5.1170
6.1670
7.9000
5.4000
6.9500
8.2170

30.3400
38.9600
48.5200
30.2000
37.8400
50.6900
30.5200
42.1200
51.2300

Step 2 Convert into S/N ratio

All responses were converted to signal to the noise ratio (S/N ratio) with the help of given formulas,
For Lower-is-better (LB),
v

1
(S/N)LB = -10log10[ v x
u=1
(1)
For Higher-is-better (HB),
v

1
1
(S/N)HB = -10log10[ v x ]
u=1

(2)

Where x is responses value, u and v are experimental condition.

Step 3 Normalization of S/N ratios values
S/N values are normalize using these mathematical formulas [10],
For Lower-is-better (LB),
Z=

Max ( x ) x
Max ( x ) Min ( x )
(3)

For Higher-is-better (HB),

Z=

x Min ( x )
Max ( x ) Min ( x )
(4)

Where Z is the normalised value for xuv responses, Min (xuv) is the smaller value of xuv and Max (xuv) is the
larger value of xuv, Z is a normalized array.

Z=

Trails

S/N (MRR)
(dB)
37.9481
35.9702
34.8262
38.0183
36.3803
34.1854
37.5704
35.3163
33.8534

1
2
3
4
5
6
7
8
9

S/N (MT)
(dB)
-14.3467
-16.1912
-17.2867
-14.1803
-15.8015
-17.9525
-14.6479
-16.8397
-18.2943

TOTAL
MEAN
VARIANCE
COVARIANCE
CORRELATION

X-X'

Y-Y'

Z-Z'

0.466

0.443
4

0.488
9
0.015
7

0.983
2

0.040
5

0.001
2

0.508
3

0.488
8

0.474
4

0.233
6

0.755
1

0.889
7

0.482
8

0.606
7

0.394
1

0.419
2

0.089
5

0.079
7

0.916
9

0.892
5

0.113
6

0.012
4

0.351
3

0.646
4

0.622

0.992
5

4.655
3
0.517
2

4.355
4
0.483
9

4.411
4
0.490
1

0.008
9
0.283
6

S/N (GV)
(dB)
-29.6403
-31.8124
-33.7184
-29.6001
-31.5590
-34.0984
-29.6917
-32.4898
-34.1905

0.9832
0.5083
0.2336
1.0000
0.6067
0.0797
0.8925
0.3513
0.0000

Z=

z z

z z

0.004
9

Normalized
(MRR)
0.9832
0.5083
0.2336
1.0000
0.6067
0.0797
0.8925
0.3513
0.0000

0.0405
0.4888
0.7551
0.0000
0.3941
0.9169
0.1136
0.6464
1.0000

Normalize
d (GV)
0.0012
0.4744
0.8897
0.0000
0.4192
1.0000
0.0124
0.6220
0.9925

0.0012
0.4744
0.8897
0.0000
0.4192
1.0000
0.0124
0.6220
0.9925

(X-X')(XX')

(Y-Y')(YY')

(Z-Z')(ZZ')

(X-X')(YY')

(Y-Y')(ZZ')

(X-X')(Z-Z')

0.217156

0.1966035
6

0.2390232
1

-0.2066244

0.21677826

-0.2278274

0.000079

0.000024

0.0002464
9

-0.000044

-0.000077

0.00013973

0.0804289
6

0.0735494
4

0.1596801
6

0.10837152

-0.11332656

0.2330958
4

0.2341592
1

0.2401980
1

0.23715939

-0.23662028

0.0080102
5

0.0080640
4

0.0050268
1

-0.0080371

0.00636682

-0.00634555

-0.1894375

0.2207867

-0.22308125

0.17689231

-0.17928081

0.02143375

-0.02188221

0.25928864

-0.25984128

-1.1475415

1.24700039

-1.26806561

-0.1434

0.1559

-0.1585

0.271
2

0.399
6

0.483
9
0.089
8

0.490
1
0.070
9

0.437
5

0.433

0.509
9

0.1914062
5

0.187489

0.2599980
1

0.375
3

0.370
3

0.477
7

0.1408500
9

0.1371220
9

0.2281972
9

0.162
5

0.131
9

0.0275228
1

0.0264062
5

0.0173976
1

0.516
1

0.502
4

0.2674958
4

0.2663592
1

0.2524057
6

1.1660450
4

1.1297768

1.4021733
5

0.1458

0.1412

0.1753

0.165
9
0.517
2

Normalized
(MT)
0.0405
0.4888
0.7551
0.0000
0.3941
0.9169
0.1136
0.6464
1.0000

0.0769123
2
0.2336269
2

0.1389735
9
0.0269587
5
0.2669269
2

-0.9997

0.9907

-0.9917

Principal component analysis

Principal component analysis (PCA) technique is a multivariate statistical approach developed by pearsion [18]
and further it is drive by Hotelling [8]. Initially this technique was applied to identify and qualify in social
science, where it is very difficult to directly measure the phenomenal changes. PCA is quit beneficial in
reducing data and interpretation of multiple objective sets of data. Now ever day PCA is widely using in
scientific fields. In this paper problem consists of three process parameters and is found with correlation
between them. To find the optimal process parameter from various process parameters with different technique
is quite difficult, but PCA make it easy to solve and overcome best parameters because of this quality PCA has
been used in countless statistical applications. PCA is capable to convert multiple correlated responses data into
several uncorrelated quality indices. Then it is converted into mathematical function which makes it easy to
calculate. Formula is made by gathering all or some quality indices. Finally rank can be used to find out the
optimal parameters or condition on which MRR is Maximum, at Minimum MT and GV [22].
The procedure steps to find out PCA are given as:
Step 4 calculate variance-covariance in array B through the normalized values

B=

N N

N N

]
Cov(x ( u ) , x (v))
Nu,v=

Var ( x ( u ))Var ( x ( v ) )
(5)

Where u=1, 2,3..v, and Cov(x(u), x(v)) is the covariance of x(u) and x(v).
Step 5 Calculate the Eigen values and Eigen vectors with the help of correlation coefficient,
(R-xIu)Vuv=0

(6)

Where R is correlation, is Eigen value, I is identity matrix and V is Eigen vectors.

Step 6 Find the principal components,
v

x V

PC=

u=1

(7)
Where i=1,2,3.n
Step 7 Calculate the Total Principal Component Index,
v

TPCI =

PC e (x)
x=1

(8)

eig( x )
Where, e(x) =

eig(x)
x=1

eig (x) = xth Eigen value

Step 8 Generate response table and choose the optimal processing parameters.
Variables
Open Voltage (OV)
Servo Voltage (SV)
Wire Feed (WF)

Level 1
m1
n1
o1

Level 2
m2
n2
o2

m1 =

Level 3
m3
n3
o3

(TPCI )+(TPCI )+(TPCI )

3
(9)

n1 =

(TPCI )+(TPCI )+(TPCI)

3
(10)

o1 =

(TPCI )+(TPCI )+(TPCI)

3
(11)

Step 9 Perform the statistical analysis of variance through ANOVA.

1.1. Selection of material
We selected En45A alloy steel as workpiece for our experimental work. It is widely used in most of the field for
manufacturing of oil and gases pipeline, springs, construction, automobiles, power plants, etc. It is widely used
due to its mechanical properties and hardness. Melting temperature of En45A alloy steel is 1425c and its
density is 8.08 gm/cm3. Chemical composition of En45A alloy steel material is as below shown in table 1 [14].
Element
s
%

Carbon (C)
0.66

Table 1. Chemical composition of En45A alloy steel

Silicon (Si)
Manganese
Chromium
Copper
(Mn)
(Cr)
(Cr)
1.62
0.89
0.23
0.20

Iron (Fe)
95.60

(b)

(a

Figure 2. Top view (a) and side view (b) of En45A Alloy steel material
1.2. Table of input parameters
S. No
1.
2.
3.

Variables
Open Voltage (OV)
Servo Voltage (SV)
Wire Feed (WF)

1.3. Taguchi L9 orthogonal array

Table 2. Variables with their levels

Symbols
Units
Level (L1)
A
Volts
80
B
Volts
30
C
m/min
5

Level (L2)
85
40
6

Level (L3)
90
50
7

This technique is found by Genichi Taguchi, a Japanese scientist which is based on orthogonal array. Nowever
day it is widely using day by day it makes trails compact and make less difficult to find out the optimal
parameters. This technique is totally depends on factors and number of trails (level). Here we consider Open
voltage (OV), Servo voltage (SV), and Wire feed (WF) as an input parameters and applied L 9 Taguchi
orthogonal array [19]. Taguchi L9 orthogonal array is shown in table 3.
Trails
1
2
3
4
5
6
7
8
9

Table 3. Taguchi L9 orthogonal array

Open Voltage (A)
Servo Voltage (B)
1
1
1
2
1
3
2
1
2
2
2
3
3
1
3
2
3
3

Wire Feed (C)

1
2
3
2
3
1
3
1
2

1.4. Experimentally measured responses

Trails
1
2
3
4
5
6
7
8
9

Table 4. Table of experimental responses

Material Removal Rate
Machining Time (MT)
3
(MRR) (mm /min)
(min)
78.9600
5.2160
62.8800
6.4500
55.1200
7.3170
79.6000
5.1170
65.9200
6.1670
51.2000
7.9000
75.6000
5.4000
58.3200
6.9500
49.2800
8.2170

Gap Voltage (GV)

(Volts)
30.3400
38.9600
48.5200
30.2000
37.8400
50.6900
30.5200
42.1200
51.2300

1.5. Responses converted into S/N ratio

Trails
1
2
3
4
5
6
7
8
9

Table 5. Sequences of S/N ratio

S/N (MRR) (dB)
S/N (MT) (dB)
37.9481
-14.3467
35.9702
-16.1912
34.8262
-17.2867
38.0183
-14.1803
36.3803
-15.8015
34.1854
-17.9525
37.5704
-14.6479
35.3163
-16.8397
33.8534
-18.2943

S/N (GV) (dB)

-29.6403
-31.8124
-33.7184
-29.6001
-31.5590
-34.0984
-29.6917
-32.4898
-34.1905

1.6. Normalization of S/N values

Trails
1
2
3
4
5
6
7
8

Table 6. Normalized S/N ratio values

Normalized (MRR)
Normalized (MT)
0.9832
0.0405
0.5083
0.4888
0.2336
0.7551
1.0000
0.0000
0.6067
0.3941
0.0797
0.9169
0.8925
0.1136
0.3513
0.6464

Normalized (GV)
0.0012
0.4744
0.8897
0.0000
0.4192
1.0000
0.0124
0.6220

0.0000

1.0000

0.9925

1.7. Analysis Eigen values and Eigen vectors

Table 7. Eigen values & Eigen vectors
PC1
PC2
-0.4780
0.373
0.4780
-0.441
0.476
0.816

Eigen Vectors

Eigen Values
Proportion
Cumulative

2.9882
0.9960
0.9960

PC3
0.726
0.687
-0.039

0.0116
0.0040
1.0000

0.0002
0.0000
1.0000

1.8. Calculate Principal Component Analysis

Table 8. Principal component for experimental trails
VPC1
VPC2
-0.2417
0.3499
0.3620
0.3611
0.1139
0.4801
0.6780
0.3730
0.6245
0.3946
0.5991
0.4414
-0.4430
0.2929
0.1044
0.3535
0.6497
0.3689

Trails
1
2
3
4
5
6
7
8
9

VPC3
0.7415
0.6863
0.6536
0.7260
0.6949
0.6488
0.7255
0.6748
0.6483

1.9. Calculate Total Principal Component Index

Trails
1
2
3
4
5
6
7
8
9

Table 9. Total principal component index with ranks

TPCI
Ranks
-0.2393
8
0.3620
5
0.1153
6
0.6768
1
0.6237
3
0.5985
4
-0.4401
9
0.1054
7
0.6486
2

2. Applying ANOVA method

ANOVA is conducted to identify the most significant factor which effects performance and result of ANOVA is
given in Table 10. It shows the individual affect of Open voltage, Servo voltage & Gap voltage. In our F-test, if
F value of any input parameter is high, it means that input parameter is affecting most [5].
Symbol
A
B
C
Error
SUM

DOF
2
2
2
2

Table 10. ANOVA table

Sum of square
Variance
0.586
0.293
0.348
0.174
0.383
0.192
0.047
0.023
1.364

F-values
12.466
7.401
8.151

Contribution (%)
42.96
25.50
28.09
3.45
100

3. Results and Discussion

When we conduct our operations on wire cut machine, we took Open voltage, Servo voltage & Wire feed as our
input parameters and material removal rate, machine timing, & gap voltage as our responses. After conducting
experiments principal component analysis method is used with the support of Taguchis orthogonal array and we

identify optimal process parameters (A2 B1 C2), means on 85 Open Voltage, 30 Servo Voltage and 6 Wire feed,
machining will be optimum it will give maximum MRR at minimum MT and GV which is in table 11.
S. No.

Open Voltage

85

Table 11. Optimal process parameters

Servo Voltage Wire Feed Material Removal
Rate
30
6
79.6000

Machining
Time
5.1170

Gap
Voltage
30.2000

Individual optimal process parameters are identified and shown in Table 12. According to this optimal process
parameters are (A3 B1 C2).
Inputs
Open Voltage (OV)
Servo Voltage (SV)
Gap Voltage (GP)
* Optimal values

Table 12. Main effect of factors on the TPCI

Level 1
Level 2
Level 3
Max-Min
0.0794
-0.0009
0.1549*
0.1558
0.6330*
0.3637
0.5625
0.2693
0.1046
0.4541*
0.0996
0.3545

Rank
3
1
2

4. Conclusions
In this paper Taguchi orthogonal array is coupled with Principal component analysis are used to improve
multiple responses like MRR (material removal rate), MT (machining time), & GV (gap voltage) of En45A
alloy steel for WEDM. As well as our result is compare and tested with the help of ANOVA. The conclusions of
this experimental optimization work as follows:
Experimental results were found with good agreement.
The optimal parameters combination was identified as A 2B1C2 i.e. Open voltage 85V, Servo Voltage 30V
and wire feed 6m/min.
Individual optimal parameter combination was identified as A3B1C2.
The result of ANOVA shows affect of Open voltage is more, than other input parameters.
ANOVA find out the 3.45% of error in experimental calculation.