Table 1 (a) Structural design of ANN applied to construct model (b) Central composite design of four independent parameters with experimental and predicted values of percentage yield.

No of neurons in input layer

4

No of neurons in hidden layer

6

No of neurons in output layers

1

Training error

0.020893

Test error

0.001040

Hidden activation

Exponential

Output activation

Tanh

No of neurons in input layer

4

1

35

6.4

6

10

27.30

25.92

26.99

2

45

6.4

4

20

26.60

27.24

27.24

3

45

6.4

5

20

29.20

28.06

28.73

4

45

6.4

6

10

25.90

26.47

25.10

5

35

6.1

6

20

23.40

24.03

23.52

6

35

6.8

6

20

27.30

26.10

27.01

7

25

6.4

4

10

27.20

27.12

27.21

8

25

6.4

5

10

27.60

28.18

27.19

9

25

6.8

5

20

27.70

28.23

27.78

10

45

6.4

6

30

30.00

29.01

28.35

11

25

6.8

4

20

27.70

27.41

27.15

12

35

6.4

6

30

25.30

27.05

26.15

13

25

6.4

6

20

30.10

30.44

30.27

14

25

6.4

6

20

30.23

30.44

30.15

15

25

6.1

4

20

24.54

24.51

24.37

16

25

6.4

5

20

30.10

29.45

29.95

17

25

6.1

6

20

27.30

26.51

26.94

18

25

6.1

6

10

24.80

24.97

24.73

19

25

6.4

6

20

30.20

30.44

30.15

20

45

6.8

6

20

26.20

27.03

26.32

21

35

6.4

4

20

26.40

25.91

26.95

22

25

6.4

6

20

30.10

30.44

30.70

23

25

6.8

6

10

28.50

28.62

28.63

24

25

6.4

6

30

29.70

29.06

29.74

25

35

6.4

5

20

26.10

26.77

26.99

26

25

6.8

6

30

27.20

27.18

27.24

27

25

6.4

6

30

29.40

29.06

29.73

28

25

6.4

4

30

27.30

27.51

27.10

29

45

6.4

6

20

28.90

28.98

28.45