3. Experimentation

Two models (S1 and S2) were developed from 80 rows of experimental data performed of D2. Only training result of best performing model (S2) of 55 rows is Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique DOI: http://dx.doi.org/10.5772/intechopen.81816

Figure 4.

Predictions against observations of Ra for model-D2, S2, 7 N (training dataset).

Figure 5. Predictions against observations of Ra for model-D2, S2, 7 N (validation dataset).

presented here for achieving the aimed to optimization of influencing process parameters (Figures 4 and 5).

OPTIMIZATION OF PROCESS PARAMETER Ra: D2, 7 Neurons in hidden layer. The best model needs to be predicted among Model-S1 and S2, in D2 steel. Effect of individual input parameters will be observed on the Ra (Tables 3–6).

### 4. Optimization of process parameters

It is evident from Table 3, that each independent influencing input parameter has corresponding values of their square of residuals at each three levels. Two values at each level (2 3 = 6 rows) has been taken for each inputs, where lowest possible square of residuals are available, to draw the Figure 6(a–f).

#### 5. Result

controllable instructed programme in MATLAB 2010a. Steepest descent problem used for the training algorithm to train the multilayer network, where the values of gradient was smallest because of the small changes in weight and biases. p1, p2, p3, p4, p5 and p6 are the six input layer neurons and Oi is the single neurons in output layer, whereas I11-I17 and I21-I29 (7 neurons present in primary and 10 in secondary

Factors/three level (coding) 1 2 3 Gap voltage (Vg): (volt) 30 60 90 Flush rate (Fr): (L/min) 4 6 8 Pulse on time (Ton): (μS) 1.05 1.15 1.25 Pulse of time (Toff): (μS) 130 160 190 Wire feed rate (Wf):(m/min) 2 5 8 Wire tension (Wt): (g) 300 600 900

Two models (S1 and S2) were developed from 80 rows of experimental data performed of D2. Only training result of best performing model (S2) of 55 rows is

hidden layers) are the hidden layers (Figure 3).

3. Experimentation

Artificial neural network approach.

Figure 3.

114

Table 2.

Factors for screening test.

Applied Surface Science

Figure 6(a–f) shows the relations between individual influencing parameters (Vg, Fr, Ton, Toff, Wf and Wt) to their optimized response, surface roughness (Ra) with corresponding values of MRR. Table 5 also indicates that unique values of each influencing parameters (corresponding to its serial numbers of Table 5) gives optimum responses, which has been highlighted.

SN Gap voltage

117

Flush rate

Spark time

Spark time

Wire feed

Wire tension

Surface roughness

Surface roughness

Material removal

Square of

residuals

(MRR)

(Ra) Pred.

(Vg)

(V)

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

41

42

 90

8

1.05

160

2

600

1.0962

 90

8

1.05

130

2

900

1.1369

 60

6

1.05

160

2

900

1.3208

 60

6

1.05

130

2

600

1.4536

 90

8

1.25

160

5

900

1.1125

 90

8

1.25

130

5

600

1.1413

 90

8

1.05

160

2

600

1.1838

 90

8

1.05

130

2

900

1.2396

 90

6

1.25

160

2

600

1.0832

 90

6

1.25

130

2

900

1.1823

 90

6

1.05

160

5

900

1.2973

 90

6

1.05

130

5

600

1.3127

 60

8

1.25

160

2

900

1.5435

 60

8

1.25

130

2

600

1.5208

 60

8

1.05

160

5

600

1.3601

 60

8

1.05

130

5

900

1.4592

 60

6

1.15

160

5

600

1.4038

 90

8

1.15

190

5

600

1.1194

 90

8

1.15

160

5

900

1.2286

 90

4

1.25

190

5

600

1.3218

 90

4

1.25

160

5

900

1.3572

 (Lit./min)

 (μS)

 (μS)

 (m/min)

 (g)

(μm)

(μm) 1.3664 1.3425 1.2292 1.1062 1.4023

1.459 1.3441 1.5302 1.5535 1.3118 1.3023 1.1867 1.0812 1.2696 1.1739 1.1524 1.1364 1.4546 1.3474 1.1423 1.0905

78

3.249E05

92 112 128 114

96

2.916E05

0.0007076

1E06

0.0005712

0.0001232

78 72 117 105

89 81

9.801E05

0.0009

4E06

1.936E05

2.5E05

8.1E07

139 202 168

0.0001

8.836E05

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique

0.000256

155 162

4E08

2.25E06

91 64

0.0001742

3.6E07

DOI: http://dx.doi.org/10.5772/intechopen.81816

(mg/min)

107

88

0.0004285

8.464E-05

 (μm)2

(Fr)

(TON)

(TOFF)

(Wf)

(Wt)

(Ra) Obs.

Applied Surface Science


#### Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique DOI: http://dx.doi.org/10.5772/intechopen.81816

SN Gap voltage

116

Flush rate

Spark time

Spark time

Wire feed

Wire tension

Surface roughness

Surface roughness

Material removal

Square of

residuals

(MRR)

(Ra) Pred.

(Vg)

(V)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

 90

4

1.15

190

8

900

1.1096

 90

4

1.15

160

8

600

1.1098

 30

8

1.25

190

5

900

1.6794

 30

8

1.25

160

5

600

2.1256

 30

8

1.15

190

8

600

1.363

 30

8

1.15

160

8

900

1.5124

 60

6

1.15

130

5

300

1.248

 60

6

1.05

160

2

300

1.1598

 60

6

1.05

130

2

600

1.3322

 60

4

1.15

160

2

300

1.3476

 60

4

1.15

130

2

600

1.273

 60

4

1.05

160

5

600

1.2076

 60

4

1.05

130

5

300

1.1772

 30

6

1.15

160

2

600

1.564

 30

6

1.15

130

2

300

1.676

 30

6

1.05

160

5

300

1.5278

 30

6

1.05

130

5

600

1.3836

 30

4

1.15

160

5

300

1.4658

 30

4

1.15

130

5

600

1.3884

 30

4

1.05

160

2

600

1.4452

 30

4

1.05

130

2

300

1.6858

 (Lit./min)

 (μS)

 (μS)

 (m/min)

 (g)

(μm)

(μm) 1.6863 1.4451 1.3713 1.4428 1.3788 1.5553 1.6756 1.4909 1.1754 1.2083 1.2663 1.3455 1.3277 1.1371 1.1945 1.5422 1.3482

2.128 1.6823 1.1096 1.0952

145 108 206 101

88 63

0.0002074

4E

08

8.41E

06

5.76E

06

0.000219

0.000888

115 118

0.0028623

0.0005153

92 133 95 125 110 97 95 104

88 136 116 110

2.025E

05

4.41E

06

4.489E

05

4.9E

07

3.24E

06

0.0053436

1.6E

07

0.0007562

2.304E

05

0.000529

0.0002924

1E

08

(mg/min)

102

2.5E

07

Applied Surface Science

 (μm)2

(Fr)

(TON)

(TOFF)

(Wf)

(Wt)

(Ra) Obs.


Table 3. D2, S1, 7N, training data (combined parameters).

SN

119

 Gap

Flush rate

Spark ON time

Spark OFF time

Wire feed

Wire

Surface roughness

Surface roughness (Ra)

(Residual)2

Material removal

predicted (MRR)

predicted.

(Ra) obs.

voltage

(Fr)

(TON)

(TOFF)

(Wf)

tension (Wt)

(Vg)

(V)

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

 90

 4

1.15

160

8

600

1.1098

 30

8

1.25

190

5

900

1.6794

 30

8

1.25

160

5

600

2.1256

 30

8

1.15

190

8

600

1.363

 30

8

1.15

160

8

900

1.5124

 60

 6

1.15

130

5

300

1.248

 60

 6

1.05

160

2

300

1.1598

 60

 6

1.05

130

2

600

1.3322

 60

 4

1.15

160

2

300

1.3476

 60

 4

1.15

130

2

600

1.273

 60

 4

1.05

160

5

600

1.2076

 60

 4

1.05

130

5

300

1.1772

 30

6

1.15

160

2

600

1.564

 30

6

1.15

130

2

300

1.676

 30

6

1.05

160

5

300

1.5278

 30

 30

 4 6

1.05

130

5

600

1.3836

1.15

160

5

300

1.4658

 30

 4

1.15

130

5

600

1.3884

 30

 4

1.05

160

2

600

1.4452

 30

 4

1.05

130

2

300

1.6858

 (Lit./min)

 (μS)

(μS)

 (m/min)

 (g)

(μm)

(μm) 1.6863 1.4451 1.3713 1.4428 1.3788 1.5553 1.6756 1.4909 1.1754 1.2083 1.2663

1.3455 1.3277 1.1371 1.1945 1.5422 1.3482

2.128 1.6823 1.1096

4E08

79

8.41E06

97

5.76E06

189

0.000219

105

0.000888

136

0.0028623

112

0.0005153

117

2.025E05

111

4.41E06

123

 4.489E05

4.9E07

96

131

3.24E06

108

0.0053436

102

1.6E07

114

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique

0.0007562

116

2.304E05

115

0.000529

102

0.0002924

119

DOI: http://dx.doi.org/10.5772/intechopen.81816

1E08

95

2.5E07

105

(μm)2

(mg/min)


#### Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique DOI: http://dx.doi.org/10.5772/intechopen.81816

SN Gap voltage

118

Flush rate

Spark time

Spark time

Wire feed

Wire tension

Surface roughness

Surface roughness

Material removal

Square of

residuals

(MRR)

(Ra) Pred.

(Vg)

(V)

43

44

45

46

47

48

49

50

51

52

53

54

55

Table 3. D2, S1, 7N, training data (combined

 parameters).

 60

4

1.25

160

> Average

2

900

1.2036

 60

4

1.15

190

8

900

1.1871

 60

4

1.15

160

8

300

1.2136

 30

6

1.25

190

2

900

1.5609

 30

6

1.25

160

2

300

1.6368

 30

6

1.15

190

8

300

1.6128

 30

6

1.15

160

8

900

1.6402

 30

4

1.25

190

8

300

1.4658

 30

4

1.25

160

8

900

1.4935

 30

4

1.15

190

2

900

1.5782

 30

4

1.15

160

2

300

1.6813

 90

8

1.25

160

5

900

1.1723

 90

8

1.25

130

5

600

1.1551

 (Lit./min)

 (μS)

 (μS)

 (m/min)

 (g)

(μm)

(μm) 1.1551 1.1153 1.6628 1.5577 1.5283 1.4666 1.6368 1.6021 1.6354 1.5668 1.1945 1.1878 1.2035 1.3654

123 128 148 113.8

1E

08

4.9E

07

0.0003648

(mg/min)

99 74 112 108 163 155 121 132 103 108

3.481E

05

1.96E

06

0.0001145

1.156E

05

6.4E

07

0.001211

0.0004202

0.0003422

0.003249

0

Applied Surface Science

 (μm)2

(Fr)

(TON)

(TOFF)

(Wf)

(Wt)

(Ra) Obs.


SN

121

 Gap

Flush rate

Spark ON time

Spark OFF time

Wire feed

Wire

Surface roughness

Surface roughness (Ra)

(Residual)2

Material removal

predicted (MRR)

predicted.

(Ra) obs.

voltage

(Fr)

(TON)

(TOFF)

(Wf)

tension (Wt)

(Vg)

(V)

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

Table 4. Training data for model: S2, Ra, N7, D2 steel.

 60

 4

1.25

160 Average

2

900

1.2036

 60

 4

1.15

190

8

900

1.1871

 60

 4

1.15

160

8

300

1.2136

 30

6

1.25

190

2

900

1.5609

 30

6

1.25

160

2

300

1.6368

 30

6

1.15

190

8

300

1.6128

 30

 30

 4 6

1.15

160

8

900

1.6402

1.25

190

8

300

1.4658

 30

 4

1.25

160

8

900

1.4935

 30

 4

1.15

190

2

900

1.5782

 30

 4

1.15

160

2

300

1.6813

 90

8

1.25

160

5

900

1.1723

 90

8

1.25

130

5

600

1.1551

 90

8

1.05

160

2

600

1.0962

 90

8

1.05

130

2

900

1.1369

 (Lit./min)

 (μS)

(μS)

 (m/min)

 (g)

(μm)

(μm) 1.1423 1.0905

1.1551 1.1153 1.6628 1.5577 1.5283 1.4666 1.6368 1.6021 1.6354 1.5668 1.1945 1.1878 1.2035 1.3654

0.002642

114.8

1E08

144

4.9E07

125

0.0003648

123

3.481E05

109

1.96E06

112

0.0001145

141

1.156E05

115

6.4E07

163

0.001211

158

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique

0.0004202

100

0.0003422

117

0.003249

88

0

94

DOI: http://dx.doi.org/10.5772/intechopen.81816

 3.249E05

2.916E05

96

74

(μm)2

(mg/min)

Applied Surface Science


Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique DOI: http://dx.doi.org/10.5772/intechopen.81816

> Table 4. Trainingdata

 for model: S2, Ra, N7, D2 steel.

SN

120

 Gap

Flush rate

Spark ON time

Spark OFF time

Wire feed

Wire

Surface roughness

Surface roughness (Ra)

(Residual)2

Material removal predicted (MRR)

predicted.

(Ra) obs.

voltage

(Fr)

(TON)

(TOFF)

(Wf)

tension (Wt)

(Vg)

(V)

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

 60

 6

1.05

160

2

900

1.3208

 60

 6

1.05

130

2

600

1.4536

 90

8

1.25

160

5

900

1.1125

 90

8

1.25

130

5

600

1.1413

 90

8

1.05

160

2

600

1.1838

 90

8

1.05

130

2

900

1.2396

 90

6

1.25

160

2

600

1.0832

 90

6

1.25

130

2

900

1.1823

 90

6

1.05

160

5

900

1.2973

 90

 60

 8 6

1.05

130

5

600

1.3127

1.25

160

2

900

1.5435

 60

 8

1.25

130

2

600

1.5208

 60

 8

1.05

160

5

600

1.3601

 60

 8

1.05

130

5

900

1.4592

 60

 6

1.15

160

5

600

1.4038

 90

8

1.15

190

5

600

1.1194

 90

8

1.15

160

5

900

1.2286

 90

 4

1.25

190

5

600

1.3218

 90

 4

1.25

160

5

900

1.3572

 90

 4

1.15

190

8

900

1.1096

 (Lit./min)

 (μS)

(μS)

 (m/min)

 (g)

(μm)

(μm) 1.0952 1.3664

1.3425 1.2292 1.1062 1.4023

1.459 1.3441 1.5302 1.5535 1.3118 1.3023 1.1867 1.0812 1.2696 1.1739 1.1524 1.1364 1.4546 1.3474

0.0007076

111

1E

06

135

0.0005712

106

0.0001232

81

9.801E

05

86

0.0009

97

4E

06

111

1.936E

05

122

2.5E

05

93

8.1E

07

74

0.0001

163

8.836E

05

208

0.000256

143

4E

08

158

2.25E

06

153

0.0001742

69

3.6E

07

101

0.0004285

92

 8.464E

05

110

0.0002074

70

(μm)2

(mg/min)

Applied Surface Science


Again experiment has been conducted on D2 steel using WEDM by setting the individual optimum parametric combinations (Vg, Fr, Ton, Toff, Wf and Wt) as 90 (V), 8 (Lit./min), 1.05 (μS), 190 (μS), 2 (m/min) and 900 (g) respectively and found the values of Ra = 0.9638 (μm) at MRR = 105 (mg/min) (Table 7).

Summary of R<sup>2</sup> values of training validation and testing data: 7 N in 1st and 10 N in 2nd L, Ra.

It has been concluded that the best fitted model (S2) for material removal rate and surface roughness of D2 steel has been achieved by artificial neural network

6. Conclusion

123

Table 6.

SN Gap voltage (Vg)

Correlation coefficient (R<sup>2</sup>

D2 S1

Table 5.

and secondary hidden layers.

Material Model R<sup>2</sup>

Training

S1 Validation

> S1, Testing Testing

S2 set, Training Training

S2 validation

S2, testing testing

value

Flush rate (Fr)

DOI: http://dx.doi.org/10.5772/intechopen.81816

Spark time (TON)

Spark time (TOFF)

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique

Wire feed (Wf)

7 30 6 1.15 130 2 300 1.6760 97 1.6E07 1 30 4 1.05 130 2 300 1.6858 102 2.5E07 2 30 4 1.05 160 2 600 1.4452 92 1E08 31 90 6 1.05 130 5 600 1.3127 78 8.1E07 54 60 4 1.15 190 8 900 1.1871 128 4.9E07 21 90 4 1.15 190 8 900 1.1096 63 0.0002074

D2, S2, 7N, training data (individual parameters corresponding to least square of residuals).

Equation of lines (correlation between obs. and pred. values of Ra)

(V) (Lit./min) (μS) (μS) (m/min) (g) (μm) (mg/min) (μm)2

Wire tension (Wt)

Wt

): Training data of D2 steel (best performing model S2) using 7 and 10 neurons in primary

Average predicted Ra (μm)

0.983 y = 1.005x - 0.010 1.3864 0.003401 0.2453 0.8129

0.991 y = 1.004x - 0.008 1.3654 0.002642 0.1934 0.3865

0.967 y = 1.067x - 0.090 1.3008 0.01077 0.8279

0.963 y = 0.879x + 0.154 1.4016 0.01914 1.3655

0.988 y = 0.984x + 0.028 1.3888 0.007015 0.5051

0.979 y = 1.006x - 0.006 1.4232 0.006565 0.4612

Root mean square error (μm)

Percentage RMSE (%)

Average % RMSE RMSE

0.7353

Surface roughness (Ra)

Material removal (MRR)

Square of residuals (Ra)

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique DOI: http://dx.doi.org/10.5772/intechopen.81816


Correlation coefficient (R<sup>2</sup> ): Training data of D2 steel (best performing model S2) using 7 and 10 neurons in primary and secondary hidden layers.

#### Table 5.

SN Gap voltage (Vg)

122

Vg

Applied Surface Science

Flush rate (Fr)

Fr

Spark time (TON)

Ton

Toff

Spark time (TOFF) Wire feed (Wf)

2 30 4 1.05 160 2 600 1.4452 92 1E08 7 30 6 1.15 130 2 300 1.676 97 1.6E07 27 60 8 1.05 130 5 900 1.4592 162 4E08 54 60 4 1.15 190 8 900 1.1871 128 4.9E07 20 90 4 1.15 160 8 600 1.1098 88 4E08 43 90 8 1.25 130 5 600 1.1551 99 0

1 30 4 1.05 130 2 300 1.6858 102 2.5E07 55 60 4 1.25 160 2 900 1.2036 148 1E08 7 30 6 1.15 130 2 300 1.676 97 1.6E07 31 90 6 1.05 130 5 600 1.3127 78 8.1E07 27 60 8 1.05 130 5 900 1.4592 162 4E08 43 90 8 1.25 130 5 600 1.1551 99 0

41 90 8 1.05 130 2 900 1.1369 96 2.916E05 42 90 8 1.05 160 2 600 1.0962 78 3.249E05 54 60 4 1.15 190 8 900 1.1871 128 4.9E07 20 90 4 1.15 160 8 600 1.1098 88 4E08 55 60 4 1.25 160 2 900 1.2036 148 1E08 43 90 8 1.25 130 5 600 1.1551 99 0

7 30 6 1.15 130 2 300 1.676 97 1.6E07 27 60 8 1.05 130 5 900 1.4592 162 4E08 55 60 4 1.25 160 2 900 1.2036 148 1E08 36 90 8 1.05 160 2 600 1.1838 81 9.801E05 21 90 4 1.15 190 8 900 1.1096 63 0.0002074 54 60 4 1.15 190 8 900 1.1871 128 4.9E07 Wf 34 90 6 1.25 160 2 600 1.0832 105 4E06 2 30 4 1.05 160 2 600 1.4452 92 1E08 43 90 8 1.25 130 5 600 1.1551 99 0 31 90 6 1.05 130 5 600 1.3127 78 8.1E07 48 30 4 1.25 190 8 300 1.4658 155 6.4E07 16 30 8 1.15 160 8 900 1.5124 145 0.000888

(V) (Lit./min) (μS) (μS) (m/min) (g) (μm) (mg/min) (μm)2

Wire tension (Wt)

Surface roughness (Ra)

Material removal (MRR)

Square of residuals (Ra)

D2, S2, 7N, training data (individual parameters corresponding to least square of residuals).


#### Table 6.

Summary of R<sup>2</sup> values of training validation and testing data: 7 N in 1st and 10 N in 2nd L, Ra.

Again experiment has been conducted on D2 steel using WEDM by setting the individual optimum parametric combinations (Vg, Fr, Ton, Toff, Wf and Wt) as 90 (V), 8 (Lit./min), 1.05 (μS), 190 (μS), 2 (m/min) and 900 (g) respectively and found the values of Ra = 0.9638 (μm) at MRR = 105 (mg/min) (Table 7).

#### 6. Conclusion

It has been concluded that the best fitted model (S2) for material removal rate and surface roughness of D2 steel has been achieved by artificial neural network

Author details

SN Gap voltage (Vg)

Table 7.

Flush rate (Fr)

(V) (Lit./ min) Spark time (TON)

DOI: http://dx.doi.org/10.5772/intechopen.81816

Best parametric combination with their possible responses.

Spark time (TOFF)

(μS) (μS) (m/

Wire feed (Wf)

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique

min)

Wire tension (Wt)

20 90 4 1.15 160 8 600 1.1098 1.1096 4.00E08 79 43 90 8 1.25 130 5 600 1.1551 1.1551 0 94 42 90 8 1.05 160 2 600 1.0962 1.0905 3.25E05 74 54 60 4 1.15 190 8 900 1.1871 1.1878 4.90E07 125 34 90 6 1.25 160 2 600 1.0832 1.0812 4.00E06 111 54 60 4 1.15 190 8 900 1.1871 1.1878 4.90E07 125

Surface roughness (Ra) obs.

Surface roughness (Ra) predicted

(g) (μm) (μm) (μm)<sup>2</sup> (mg/min)

(Zero residual)<sup>2</sup>

Material removal predicted (MRR)

Umesh K. Vates<sup>1</sup>

3 UTeM, Melaka, Malaysia

India

125

\*, N.K. Singh<sup>2</sup>

2 Indian Institute of Technology (ISM) Dhanbad, India

\*Address all correspondence to: u.k.vates@gmail.com

provided the original work is properly cited.

, B.P. Sharma<sup>1</sup> and S. Sivarao<sup>3</sup>

1 Department of Mechanical Engineering, ASET, Amity University, Uttar Pradesh,

© 2019 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium,

Figure 6. (a–f) 3D scattered plot between Ra vs. MRR vs. individual independent parameter.

using WEDM. From best modeled training data, optimum parametric combinations (Vg, Fr, Ton, Toff, Wf and Wt) observed as 90 V, 8 Lit./min, 1.05 μS, 190 μS, 2 m/min and 900 g respectively and found the values of Ra = 0.9638 μm at MRR = 105 mg/min, whereas the average Ra = 1.3654 μm at MRR = 114.8 mg/min. It has been concluded that ANN modeling technique is best fitted for surface roughness prediction and able to successfully minimize (SR) is 29.41% with 8.53% decreases the MRR from its average values on D2 steel using BPANN under WEDM. Such combinations may be applied for industrial application, where it is needed.

Optimization of Surface Roughness of D2 Steels in WEDM using ANN Technique DOI: http://dx.doi.org/10.5772/intechopen.81816


#### Table 7.

Best parametric combination with their possible responses.
