135

CHAPTER 7

OPTIMIZATION OF ELECTRIC DISCHARGE

MACHINING PARAMETERS ON AA7075-TIO2
COMPOSITE

7.1 INTRODUCTION

To conduct the electric discharge machining (EDM) studies, the

composite material is selected based on the mechanical properties of the
composites. From the characterization studies discussed in chapter 4, it is
understood that the addition of 10% wt. TiO2 reinforcement enhanced the
higher mechanical properties for the composites, fabricated by stir casting
process. Therefore, 10% wt. TiO2 reinforcement composite was chosen to
study the electric discharge machining behavior of the selected composite.

The EDM process was carried out by die sinking method. The
experimental setup and experimental procedures used for conducting the
machining process are discussed in chapter 3.7.1 & 3.7.2. The necessary
formula required for computing the material removal rate (MRR), surface
roughness (SR) and electrode wear ratio (EWR) have also been discussed in
chapter 3.7.4. To conduct the experiments, the Taguchi method was applied in
this study, since the Taguchi method uses a special design of orthogonal array
that helps to reduce the number of experiments. The following procedure is
adopted in the Taguchi method, to analyze and to find out the optimal
combinations of the electric discharge machining parameters.
 136

7.1.1 Identification of Process Parameters

The selection of an independent machining parameter for the

electric discharge machining process can be obtained from the literature
survey. From the literature survey it is understood that the following process
parameters affect the material removal rate (MRR), surface roughness (SR)
and electrode wear ratio (EWR). (Lin et al. 2002 and Anil Kumar et al. 2011)

Polarity

Pulse current

Pulse ON time

Pulse OFF time

Duty cycle

Electrode material

Dielectric medium.

In this study, the electrode material, pulse current (amps), pulse

ON time (µs) and pulse OFF time (µs) are selected as the predominant key
factors, which affect the MRR, SR and EWR during EDM process.

1.7.2 Deciding the Number of Levels

After determining the process parameter, the number of levels is

selected based on the trend in which process parameters influenced on the
output performances. Table 7.1 shows the selected machining parameters and
their level values for the electric discharge machining process.
 137

Table 7.1 Process Parameters with their Levels

Levels
Process
Symbol Units
parameters
1 2 3

A Electrode material - Cu Br EN-8

B Pulse current Amp 5 10 15

s

C Pulse ON time µs 300 600 900

D Pulse OFF time µs 30 60 90

7.1.3 Selection of Orthogonal Array

In this study, L27 orthogonal array is chosen to conduct the

experiments. L27 orthogonal array has 13 columns; the first column is
assigned to the electrode material (A), second column to the pulse current (B),
the fifth column to the pulse ON time and the eighth column to the pulse OFF
time (D). This orthogonal array can handle four-three level process
parameters, and has eight degrees of freedom. The output response
characteristics are the material removal rate (MRR), surface roughness (SR)
and electrode wear ratio (EWR). The L27 orthogonal array design and input
parameters are shown in Table 7.2.
 138

Table 7.2 Orthogonal Array Design for L27 Taguchi Design

Ex. Electrode Pulse Pulse ON Pulse OFF

A B C D
No material current (A) time (µs) time (µs)
1 1 1 1 1 Cu 5 300 30
2 1 1 2 2 Cu 5 600 60
3 1 1 3 3 Cu 5 900 90
4 1 2 1 2 Cu 10 300 60
5 1 2 2 3 Cu 10 600 90
6 1 2 3 1 Cu 10 900 30
7 1 3 1 3 Cu 15 300 90
8 1 3 2 1 Cu 15 600 30
9 1 3 3 2 Cu 15 900 60
10 2 1 1 1 Br 5 300 30
11 2 1 2 2 Br 5 600 60
12 2 1 3 3 Br 5 900 90
13 2 2 1 2 Br 10 300 60
14 2 2 2 3 Br 10 600 90
15 2 2 3 1 Br 10 900 30
16 2 3 1 3 Br 15 300 90
17 2 3 2 1 Br 15 600 30
18 2 3 3 2 Br 15 900 60
19 3 1 1 1 EN-8 5 300 30
20 3 1 2 2 EN-8 5 600 60
21 3 1 3 3 EN-8 5 900 90
22 3 2 1 2 EN-8 10 300 60
23 3 2 2 3 EN-8 10 600 90
24 3 2 3 1 EN-8 10 900 30
25 3 3 1 3 EN-8 15 300 90
26 3 3 2 1 EN-8 15 600 30
27 3 3 3 2 EN-8 15 900 60
 139

7.2 SINGLE RESPONSE OPTIMIZATION OF EDM

PARAMETERS USING TAGUCHI TECHNIQUE

Based on the L27 orthogonal array all the experiments are

conducted to collect the necessary data for analysis. The collected data is used
to compute the MRR, SR and EWR using the Equations.3.7 to 3.9. The
calculated output performance values are provided in Table 7.3. The objective
of the experimentation is to maximize the MRR, minimize the SR and EWR
as much as possible. In the Taguchi method, the S/N ratio is used to represent
a performance characteristic, and the highest value of the S/N ratio is
required. In this work, the larger-the-better characteristic is selected for MRR
and the smaller-the-better characteristic is required for other output responses
such as SR and EWR (Yan-Cherng Lin et al. 2006). Equations 5.1 and 5.2 are
used to calculate the corresponding S/N ratios for the output performance
characteristics, using Minitab software version 18. The calculated S/N ratios
values for all the output responses are given in Table 7.3.
 140

Table 7.3 Experimental Results with their S/N Ratios

Output responses Signal- to- Noise ratios

Ex.
No MRR
SR (µm) EWR (%) MRR (dB) SR (dB) EWR(dB)
(g/min)
1 0.0292 8.257 43.441 -30.6923 -18.3364 -32.7580
2 0.0462 6.608 37.634 -26.7072 -16.4014 -31.5116
3 0.0315 10.113 41.865 -30.0338 -20.0976 -32.4370
4 0.0443 7.652 38.622 -27.0719 -17.6755 -31.7367
5 0.1133 6.211 29.595 -18.9154 -15.8632 -29.4244
6 0.1271 8.412 28.821 -17.9171 -18.4980 -29.1942
7 0.0247 3.171 57.581 -32.1461 -10.0239 -35.2056
8 0.1541 5.482 26.083 -16.2439 -14.7788 -28.3272
9 0.3077 6.471 19.080 -10.2375 -16.2194 -25.6116
10 0.0772 3.651 43.248 -22.2477 -11.2482 -32.7193
11 0.1138 4.385 41.203 -18.8772 -12.8394 -32.2986
12 0.1564 7.184 38.384 -16.1153 -17.1273 -31.6830
13 0.2242 3.307 32.743 -12.9873 -10.3887 -30.3024
14 0.1784 5.106 34.494 -14.9721 -14.1616 -30.7549
15 0.3279 6.862 30.917 -9.6852 -16.7290 -29.8039
16 0.0337 3.286 50.947 -29.4474 -10.3334 -34.1424
17 0.3819 4.274 29.865 -8.3610 -12.6167 -29.5033
18 0.6402 3.265 18.661 -3.8737 -10.2777 -25.4187
19 0.0410 4.519 37.200 -27.7443 -13.1008 -31.4109
20 0.0576 5.697 31.393 -24.7916 -15.1129 -29.9367
21 0.0632 8.475 29.898 -23.9857 -18.5628 -29.5128
22 0.0844 3.765 26.049 -21.4732 -11.5153 -28.3158
23 0.1764 4.377 17.564 -15.0700 -12.8235 -24.8925
24 0.2935 7.835 16.564 -10.6478 -17.8808 -24.3833
25 0.0376 4.168 41.888 -28.4962 -12.3986 -32.4418
26 0.2962 3.986 15.565 -10.5683 -12.0107 -23.8430
27 0.4495 5.854 11.333 -6.9454 -15.3491 -21.0869
 141

In Taguchi analysis, it is clearly observed that the S/N ratio at each

level of the process parameter and how it is changed when settings of each
parameter are changed from level 1 to 3. From S/N ratio table, the strongest
influence of process parameters were determined by the difference between
the maximum and minimum value of the mean of S/N ratios and it is denoted
as delta ( ). The highest value of the delta ( ) indicates that the most
influencing factor on the output responses. The response table also includes
ranks based on delta value ( ); rank 1 is assigned to the process parameter
with highest delta value, rank 2 is assigned to second highest delta value and
so on. The main effect plots are used to determine the optimal level of process
parameters to obtain the objective of the output responses (Banh Tien Long
et al. 2016).

ANOVA was used to categorize the role of each input parameter

which statistically significant and also determine the percentage contribution
of that parameter on the output responses. ANOVA calculates the F-ratio,
which is the ratio between the regression mean square to the mean square
error. This ratio is used to measure the significance of the parameters with
respect to the variance of all the terms at the desired significance level, . If
the calculated value of the F-ratio is higher than the tabulated value of the F-
ratio, then the parameter is significant at a desired level. In general, when F
>4, the related design parameter has a significant effect on the output
responses. If the F-ratio value for a parameter is less than 4, it means this
parameter is not significant and can be neglected. The P-value of each
parameter is less than 0.05, which indicates that the parameter has a
statistically significant effect on the output responses (Aveek Mohanty et al.
2014). The ratio of individual sum of square of a particular input parameter to
the total sum of squares of all the input parameters gives the percentage
contribution (P %) of the input parameters on the output responses. The
models are then checked using a numerical method employing the coefficient
 142

of determination (R-sq), adjusted R-sq (R-sq adj). R-sq indicates how much
of the observed variability in the data is accounted for by the model, while R-
sq adj. modifies R-sq by taking into account the number of predictors in the
model.

7.2.1 Analysis of S/N Ratio for MRR

Table 7.4 shows the response table of S/N ratio and Table 7.5
provide the response table of means for MRR by using Taguchi analysis.
From the tables, it can be understood that the pulse ON time has highest delta
value thus rank 1is assigned which is the most influencing factor on MRR,
followed by pulse current, electrode material and pulse OFF time. It is
observed that, increase in pulse ON time and pulse current increases the
MRR. The increase in pulse ON time increases the development of spark
energy and thus the increase in MRR achieved. The main effect plots are used
to determine the optimal level of process parameters to obtain the maximum
MRR. The main effect plot of different machining parameters like electrode
material, pulse current, pulse ON time and pulse OFF time on MRR is shown
in Figure 7.1. In the plot, the x-axis indicates the value of process parameters
at three levels and y-axis indicates the S/N ratio of response values.
Horizontal line indicates the mean S/N ratios value of the response. In Figure
7.1, it can be clearly observed that the maximum MRR is attained by the
optimum level of parameter combinations are A2B3C3D2, which indicates that
the electrode material at level 2 (Br), pulse current at level 3 (15amps), pulse
ON time at level 3 (900 µs) and pulse OFF time at level 2 (60 µs).
 143

Table 7.4 Response Table for S/N Ratio of MRR

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)
1 -23.33 -24.58 -25.81 -17.12
2 -15.17 -16.53 -17.17 -17.00
3 -18.86 -16.26 -14.38 -23.24
Delta 8.16 8.32 11.43 6.25
Rank 3 2 1 4

Table 7.5 Response Table for Mean of MRR

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)
1 0.09757 0.06846 0.06626 0.19201
2 0.23708 0.17439 0.16866 0.21866
3 0.16660 0.25840 0.26633 0.09058
Delta 0.13951 0.18994 0.20008 0.12808
Rank 3 2 1 4

Figure 7.1 Main Effect Plot for MRR

 144

7.2.2 Analysis of Variance for MRR

ANOVA was carried out to determine the influence of EDM

parameters such as electrode material, pulse current, pulse ON time and pulse
OFF time on the MRR. Table 7.6 shows the results of ANOVA for MRR.
From the table, it can be clearly understood that the pulse ON time (P-value =
0.000) is the most predominant factor on the MRR followed by pulse current
(P-value = 0.000) and electrode material (P-value = 0.004). The frequency
(F-ratio) test was performed at 95% confidence interval (CI). The calculated
F-ratio values of the pulse ON time and the pulse current were greater than
tabulated F0.05 = 3.39, which also confirms the statistical physical influence on
the MRR. Figure 7.2 shows the percentage contribution of each input
parameter affect the MRR. The ratio of individual sum of square of a
particular input parameter to the total sum of squares of all the input
parameters gives the percentage contribution (P %) of the input parameters on
the MRR. In Figure 7.2, it can be observed that the pulse ON time is the most
significant parameter that contributed (P = 29.31%) followed by the pulse
current (P = 26.53%), the electrode material (P = 14.25%) and the pulse OFF
time (P = 13.37%). The R-sq and R-sq (adj) value is 86.47% and 76.13% of
MRR which indicates that the model is able to predict the response with high
accuracy. A similar result is also observed by earlier researcher Rengasamy et
al. (2016) when they conducted on EDM machining of Al 4032 reinforced
with ZrB2 and TiB2 in-situ composites.
 145

Table 7.6 ANOVA Table for MRR

Factors DF Seq. SS Adj. MS F-ratio P-value P (%)

Electrode material (A) 2 0.087588 0.043794 7.76 0.004 14.25
Pulse current (B) 2 0.163076 0.081538 14.45 0.000 26.53
Pulse ON time (C) 2 0.180173 0.090087 15.96 0.000 29.31
Pulse OFF time (D) 2 0.082208 0.041104 7.28 0.005 13.37
Residual error 18 0.101581 0.005643 - - 16.52
Total 26 0.614627 - - - 100
S = 0.0751226; R-sq = 83.47%; R-sq (adj) = 76.13%

Electrode
Residual Error Material (A)
17% 14%

Pulse
OFF
time
(D) Pulse Current (B)
13% 27%

Pulse ON time
(C)
29%

Figure 7.2 Percentage Contributions of the Parameters on MRR

 146

Figure 7.3 Interaction Plot for MRR

The interaction of electrode material, pulse current, pulse ON time

and pulse OFF time with MRR is shown in Figure 7.3. From the graph it is
observed that, while considering electrode material and pulse current increase
in pulse current increases the MRR for all the electrode materials such as Cu,
Br and EN-8. However, among the three different electrode materials tested,
Br electrode is contributed more for removing more materials from work
piece. The maximum MRR is observed for the Br electrode while using same
pulse current level. The reason is that, low electrical conductivity nature of
brass (159 W/mK), not absorbing the heat, and all the generated heats are
used for melting and vaporization of work materials. While interacting pulse
ON time with pulse current, the increase of both factors increases the MRR
and comparing with each other, pulse ON time plays a significant role to
affecting MRR. The reason is that, increase in pulse ON time, and maintain
the discharge of heat energy continuously and melting the material to the
larger extent. The interaction of pulse ON time with pulse OFF time is clearly
 147

displays that, pulse OFF time is not contributing for the MRR at any point.
While interacting electrode material with pulse ON time, the trend is observed
very similar to trend observed for electrode material and pulse current. The
interaction between electrode material and pulse current with pulse OFF time
shows the pulse OFF time is the insignificant factor for MRR.
 148

Figure 7.4 (a-c) Contour Plots for MRR, (a) Electrode Material vs.
Pulse Current, (b) Electrode Material vs. Pulse ON Time,
(c) Pulse Current vs. Pulse ON Time

Figures 7.4(a-c) shows the contour plots for the MRR with respect
to variable input process parameters. For Figure 7.4(a) & (b), the y-axis
indicates the electrode materials 1 represents Cu, 2 represents Br and 3
represents EN-8. In Figure 7.4(a) show the response of electrode material and
pulse current on MRR. The dark area indicates the maximum MRR with
respect to electrode material and pulse current. It is clearly shows that, Br
electrode gives the higher MRR for high pulse current. The Cu and EN-8
electrode gives moderate level MRR with increase in pulse current. Figure
7.4(b) represents the response of electrode material and pulse ON time on
MRR. From the Figure 7.4(b), the dark area shows the optimum level of
MRR. It is observed that, Br electrode produces maximum MRR for the
higher pulse ON time. Figure 7.4(c) shows the response of pulse current and
pulse ON time on MRR. It is noticed that, the higher values of pulse current
and pulse ON time are required to remove more materials in the work piece.
 149

7.2.3 Prediction of Optimum MRR

The optimum value of MRR ) is predicted at the selected

optimal setting of parameters. Referring to the graph (Figure 7.1), it could be
observed that the maximum MRR is attained for the combination of optimum
parameter is A2B3C3D2. The expected mean at the optimum settings could be
estimated from Equation (7.1)

= A +B +C +D T (7.1)

Corresponding S/N ratio could be predicted as Equation (7.2)

A B C D T (7.2)

Where,

TMRR - overall mean of MRR

TMRR - S/N ratio’s overall mean of MRR

A2, B3, C3 and D2 are the means of mean values of the MRR with
parameters at optimum levels. (Referring to the Table 7.5)

A2, B3, C3 and D2 are the S/N ratio mean values of the MRR
with parameters at optimum levels. (Referring to the Table 7.4)

By substitution in Equation (7.1)

= A +B +C +D T

= (0.23708 + 0.25840 + 0.26633 + 0.21866) (3 0.167081)

= 0.479227 g/min
 150

By substitution in Equation (7.2)

= A B C D T

= ( 15.17 16.26 14.38 17.00) (3 19.12)

= 5.45 dB

Hence the predicted optimum mean value for MRR was calculated
to be 0.479227 g/min and their corresponding S/N ratio value is – 5.45 dB.

The 95% Confidence interval (CI) for the expected results from the
confirmation experiments (CICE) and the mean of population (CIPOP) are
calculated by using the Equations (5.6) & (5.7) in chapter 5.

By substituting the following values:

27
eff 3
1 8

N = 27 x 1= 27 (treatments = 27, repetitions = 1)

fe = 18 (Refer Table 7.6)

F0.05, (1, 18) = 4.41 (Tabulated F-value)

Ve = 0.005643 (Refer Table 7.6)

Therefore, CICE = + 0.182156 and CIPOP = + 0.091078

The predicted confidence interval for confirmation experiment of

MRR is given by

(µ CI ) µ (µ + CI )

(0.479227 0.182156) µ (0.479227 + 0.182156)

(0.297071 g/min) µ (0.661383 g/min)

 151

The 95% confidence intervals for the mean of population of MRR

should be given by

(µ CI ) µ (µ + CI )

(0.479227 0.091078) µ (0.479227 + 0.091078)

(0.388149 g/min) µ (0.570305 g/min)

The mean value of MRR is obtained through confirmation

experiments must be within confidence interval ranges. Figure 7.5 shows the
normal probability plot of MRR and it is evident the all the residuals are
found to be normally distributed along the straight line at 95% confidence
level.

Figure 7.5 Probability Plot of MRR

 152

7.2.4 Confirmation Experiments for MRR

The confirmation experiments are used to verify the process

parameters and their levels chosen from an experiment. The predicted
optimum value of MRR has been evaluated by the mean value of three
confirmation experiments at optimal level of process parameters. The values
of MRR obtained through confirmation experiments are within 95% of CICE
and CIPOP of respective output response. The confirmation experiments and
predicted result of MRR is given in Table 7.7.

Table 7.7 Confirmation Experiment Result and Predicted Value of MRR

Optimal setting Predicted Confirmation

Output
of process optimum experimental Error (%)
response
parameters value result

MRR
A2B3C3D2 0.479227 0.574404 16.5
(g/min)

7.2.5 Analysis of S/N Ratio for SR

Table 7.8 shows the response table of S/N ratio and Table 7.9
provide the response table for means of SR by using Taguchi analysis. From
the tables, it can be understood that the pulse ON time has highest delta value
thus rank 1is assigned which is the most influencing factor on SR, followed
by electrode material, pulse current and pulse OFF time. . It is clearly noticed
that, pulse ON time increases, the delivered spark energy increases causing
the craters are produced on machined work piece surface which results in
poor surface finish. The main effect plots are used to determine the optimal
level of process parameters to obtain the minimum SR. The main effect plot
of different machining parameters like electrode material, pulse current, pulse
ON time and pulse OFF time on SR is shown in Figure 7.6. In the plot, the x-
 153

axis indicates the value of process parameters at three levels and y-axis
indicates the S/N ratio of response value. Horizontal line indicates the mean
S/N ratios value of the SR. From the Figure 7.6, it can be clearly observed
that the minimum SR is achieved by the optimum level of parameter
combinations are A2B3C1D2, which indicates that the electrode material at
level 2 (Br), pulse current at level 3 (15amps), pulse ON time at level 1 (300
µs) and pulse OFF time at level 2 (60 µs).

Table 7.8 Response Table for S/N Ratio of SR

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)

1 -16.43 -15.87 -12.78 -15.02

2 -12.86 -15.06 -14.07 -13.98

3 -14.31 -12.67 -16.75 -14.60

Delta 3.57 3.20 3.97 1.05

Rank 2 3 1 4

Table 7.9 Response Table for Mean of SR

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)

1 6.931 6.543 4.642 5.920

2 4.591 5.947 5.125 5.223

3 5.408 4.440 7.163 5.788

Delta 2.340 2.104 2.522 0.697

Rank 2 3 1 4
 154

Figure 7.6 Main effect Plot for SR

7.2.6 Analysis of variance for SR

ANOVA was carried out to determine the influence of EDM

parameters such as electrode material, pulse current, pulse ON time and pulse
OFF time on the SR. Table 7.10 shows the results of ANOVA for SR. From
the table, it can be clearly understood that the pulse ON time (P-value =
0.000) is the most predominant factor on the SR followed by electrode
material (P-value = 0.000) and pulse current (P-value = 0.001). The frequency
(F-ratio) test was performed at 95% confidence interval (CI). The calculated
F-ratio values of the pulse ON time and the electrode material were greater
than tabulated F0.05 = 3.39, which also confirms the statistical physical
influence on the SR. Figure 7.7 shows the percentage contribution of each
input parameter affect the SR. The ratio of individual sum of square of a
particular input parameter to the total sum of squares of all the input
parameters gives the percentage contribution (P %) of the input parameters on
the SR. From the figure, it can be observed that the pulse ON time is the most
 155

significant parameter that contributed (P = 32.56%) followed by the electrode

material (25.62%), pulse current (P = 21.36%). The pulse OFF time is less
significant parameter with contribution of (P = 2.49%). This is in good
agreement with the previous research of Saravanan P Sivam et al. (2012). The
R-sq and R-sq (adj) value is 82.05% and 74.07% of SR which indicates that
the model is able to predict the response with high accuracy.

Table 7.10 ANOVA Table for SR

Factors DF Seq. SS Adj. MS F-ratio P-value P (%)

Electrode material (A) 2 25.379 12.689 12.85 0.000 25.62
Pulse current (B) 2 21.160 10.580 10.72 0.001 21.36
Pulse ON time (C) 2 32.242 16.121 16.33 0.000 32.56
Pulse OFF time (D) 2 2.469 1.234 1.25 0.310 2.49
Residual error 18 17.773 0.987 - - 17.94
Total 26 99.021 - - - 100
S = 0.993662; R-sq = 82.05%; R=sq (adj) = 74.07%

Residual Error Electrode

Pulse OFF time Material (A)
18%
(D) 26%
2%

Pulse ON time Pulse

(C) Current (B)
33% 21%

Figure 7.7 Percentage Contributions of the Parameters on SR

 156

Figure 7.8 Interaction Plot for SR

The interaction of electrode material, pulse current, pulse ON time

and pulse OFF time with SR is shown in Figure 7.8. From the graph it is
observed that, while considering the interaction of electrode material and
pulse current confirms, increase in pulse current affects the SR for all the
electrode materials such as Cu, Br and EN-8. However, among the three
different electrode materials tested, better surface finish was observed for Br
electrode. The minimum SR is observed for the Br electrode while using
higher pulse current level. While interacting pulse current and pulse ON time,
the increase in both the parameters increase the SR. The interaction of
electrode material with pulse ON time and pulse OFF time shows that, Br is
the most suitable electrode material for EDM of AA7075-10wt. % TiO2
composite. Interacting of pulse OFF time with all other factors shows that,
pulse OFF time is the insignificant factor for SR.
 157

Figure 7.9 (a-c) Contour Plots for SR, (a) Electrode Material vs. Pulse
Current, (b) Electrode Material vs. Pulse ON Time, (c) Pulse
Current vs. Pulse ON Time
 158

Figures 7.9(a-c) shows the contour plots for the SR with respect to
various input parameters. For Figure 7.9(a) & (b), the y-axis indicates the
electrode materials 1 represents Cu, 2 represents Br and 3 represents EN-8.
Figure 7.9(a) presented the response on SR with respect to electrode material
vs. pulse current. The minimum SR achieved for Br electrode with higher
pulse current. Cu electrode produces greater SR for low pulse current. Figure
7.9(b) depicted the response of electrode material vs. pulse ON time on SR. It
is noticed that, Br electrode gives minimum SR when the pulse ON time is
low. Whereas, EN-8 electrode gives moderate level of SR. Figure 7.9(c)
shows the response of pulse current vs. pulse ON time on SR. It is observed
from the graph, the low level of pulse ON time and high level of pulse current
is preferred to good surface finish.

7.2.7 Prediction of Optimum SR

The optimum value of SR ) is predicted at the selected

optimal setting of parameters. Referring to the graph (Figure 7.6), it could be
observed that the minimum SR is obtained for the combination of optimum
parameter is A2B3C1D2. The expected mean at the optimum settings could be
estimated from Equation (7.3).

=A +B +C +D T (7.3)

Corresponding S/N ratio could be predicted as Equation (7.4)

A B C D T (7.4)

where,

TSR - overall mean of SR

TSR - S/N ratio’s overall mean of SR

 159

A2, B3, C1 and D2 are the means of mean values of the SR with
parameters at optimum levels. (Referring to the Table 7.9)

A2, B3, C1 and D2 are the S/N ratio mean values of the SR
with parameters at optimum levels. (Referring to the Table 7.8)

By substitution in Equation (7.3)

=A +B +C +D T

= (4.591 + 4.440 + 4.642 + 5.223) (3 5.643)

= 1.967 m

By substitution in Equation (7.4)

= A B C D T

= ( 12.86 12.67 12.78 13.98) (3 14.532)

= 8.694 dB

Hence the predicted optimum mean value for SR was calculated to

be 1.967 µm and their corresponding S/N ratio value is – 8.694 dB.

The 95% Confidence interval (CI) for the expected results from the
confirmation experiments (CICE) and the mean of population (CIPOP) are
calculated by using the Equations (5.6) & (5.7) in chapter 5.

By substituting the following values:

27
eff 3
1 8

N = 27 x 1= 27 (treatments = 27, repetitions = 1)

fe = 18 (Refer Table 7.10)

 160

F0.05, (1, 18) = 4.41 (Tabulated F-value)

Ve = 0.987 (Refer Table 7.10)

Therefore, CICE = + 2.409 and CIPOP = + 1.204

The predicted confidence interval for confirmation experiment of

SR is given by

(µ CI ) µ (µ + CI )

(1.967 2.409) µ (1.967 + 2.409)

(0.442 m) µ (4.376 m)

The 95% confidence intervals for the mean of population of SR

should be given by

(µ CI ) µ (µ + CI )

(1.967 1.204) µ (1.967 + 1.204)

(0.763 m) µ (3.171 m)

The mean value of SR is obtained through confirmation

experiments must be within confidence interval ranges. Figure 7.10 shows
the normal probability plot of SR and it is evident the all the residuals are
found to be normally distributed along the straight line at 95% confidence
level.
 161

Figure 7.10 Probability Plot of SR

7.2.8 Confirmation Experiments for SR

The confirmation experiments are used to verify the process

parameters and their levels chosen from an experiment. The predicted
optimum value of SR has been evaluated by the mean value of three
confirmation experiments at optimal levels of process parameters. The values
of SR obtained through confirmation experiments are within 95% of CICE and
CIPOP of respective output response. The confirmation experiments and
predicted result of SR is given in Table 7.11.

Table 7.11 Confirmation Experiment Result and Predicted Value of SR

Optimal setting Predicted Confirmation

Output
of process optimum experimental Error (%)
response
parameters value result

SR (µm) A2B3C1D2 1.967 2.865 31.3

 162

7.2.9 Analysis of S/N Ratio for EWR

Table 7.12 shows the response table of S/N ratio and Table 7.13
provide the response table for means of EWR by using Taguchi analysis.
From the tables, it can be understood that the pulse ON time has highest delta
value thus rank 1is assigned which is the most influencing factor on EWR,
followed by electrode material and pulse current. It is noticed that, increasing
the pulse current and pulse ON time the EWR increases. Pulse ON time
increases, the high spark energy transferred on machined surface which
results in more heat is produced between the electrode and work piece causing
increased in electrode wear ratio. The main effect plots are used to determine
the optimal level of process parameters to obtain the minimum EWR. The
main effect plot of different machining parameters like electrode material,
pulse current, pulse ON time and pulse OFF time on EWR is shown in Figure
7.11. In the plot, the x-axis indicates the value of process parameters at three
levels and y-axis indicates the response value. Horizontal line indicates the
mean S/N ratios value of the EWR. From the figure, it can be clearly observed
that the minimum EWR is attained by the optimum level of parameter
combinations are A3B3C3D2, which indicates that the electrode material at
level 3 (EN-8), pulse current at level 3 (15amps), pulse ON time at level 3
(900 µs) and pulse OFF time at level 2 (60 µs).

Table 7.12 Response Table for S/N Ratio of EWR

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)
1 -30.69 -31.59 -32.11 -29.10
2 -30.74 -28.76 -28.94 -28.47
3 -27.31 -28.40 -27.68 -31.17
Delta 3.42 3.19 4.43 2.70
Rank 2 3 1 4
 163

Table 7.13 Response Table for Mean of EWR

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)

1 35.86 38.25 41.30 30.19

2 35.61 30.11 29.27 28.52

3 25.27 28.37 26.17 38.02

Delta 10.59 9.88 15.13 9.50

Rank 2 3 1 4

Figure 7.11 Main Effect Plot for EWR

7.2.10 Analysis of Variance for EWR

ANOVA was carried out to determine the influence of EDM

parameters such as electrode material, pulse current, pulse ON time and pulse
OFF time on the EWR. Table 7.14 shows the results of ANOVA for EWR.
 164

From the table, it can be clearly observed that the pulse ON time (P-value =
0.000) is the most significant factor on the EWR followed by electrode
material (P-value = 0.001), pulse current (P-value = 0.002) and pulse OFF
time (P = 0.003). The frequency (F-ratio) test was performed at 95%
confidence interval (CI). The calculated F-ratio values of the pulse ON time
and the pulse current were greater than tabulated F0.05 = 3.39, which also
confirms the statistical physical influence on the EWR. Figure 7.10 shows the
percentage contribution of each input parameter affect the EWR. The ratio of
individual sum of square of a particular input parameter to the total sum of
squares of all the input parameters gives the percentage contribution (P %) of
the input parameters on the EWR. In Figure 7.12, it can be noticed that the
pulse ON time is the most significant parameter that contributed (P = 35.02%)
followed by the electrode material (P = 19.99%), the pulse current (P =
15.23%) and the pulse OFF time (P = 14.10%). These results are very similar
to that of Narender sing et al. (2004) who reported for the Al-10wt.% SiC
metal matrix composite during EDM machining. The R-sq and R-sq (adj)
value is 84.36% and 77.40% of EWR which indicates that the model is able to
predict the response with high accuracy.

Table 7.14 ANOVA Table for EWR

Factors DF Seq. SS Adj. MS F-ratio P-value P (%)

Electrode material (A) 2 656.73 328.36 11.50 0.001 19.99
Pulse current (B) 2 500.54 250.27 8.77 0.002 15.23
Pulse ON time (C) 2 1150.38 575.19 20.15 0.000 35.02
Pulse OFF time (D) 2 463.20 231.60 8.11 0.003 14.10
Residual error 18 513.85 28.55 - - 15.64
Total 26 3284.70 - - - 100
S = 5.34298; R-sq = 84.36% ; R=sq (adj) = 77.40%
 165

Residual
Error Electrode
16% Material (A)
20%
Pulse OFF time
(D)
14% Pulse
Current
(B)
15%

Pulse ON time
(C)
35%

Figure 7.12 Percentage Contributions of the Parameters on EWR

Figure 7.13 Interaction Plot for EWR

 166

The interaction of electrode material, pulse current, pulse ON time

and pulse OFF time with EWR is shown in Figure 7.13. From the plot it is
observed that, while interacting electrode material with pulse current, pulse
ON time and pulse OFF time, EN-8 is the best electrode material for the
moderate level of pulse current, high pulse ON time, and low pulse OFF time.
The low EWR is observed for the above mentioned level of parameters. The
interaction of pulse current and pulse ON time shows that, lower pulse current
and high pulse ON time is the ideal condition for decreasing EWR. The
interaction of pulse ON time with pulse OFF time is clearly displays that,
pulse OFF time is not contributing for the EWR at any point. The interaction
between electrode material and pulse current with pulse OFF time shows the
pulse OFF time is the insignificant factor for affecting EWR during EDM of
AA7075-10wt. % TiO2 composite.
 167

Figure 7.14 Contour Plots for EWR, (a) Electrode Material vs. Pulse
Current, (b) Electrode Material vs. Pulse ON Time, (c) Pulse
Current vs. Pulse ON Time
 168

The contour plots for the EWR is presented in Figure 7.14(a-c). In

Figure 7.14(a) shows the response of electrode material and pulse current on
EWR. From the graph, minimum EWR obtained at EN-8 electrode with
moderate level of pulse current. Whereas both Cu and Br electrodes have
maximum EWR with respect to pulse current. Figure 7.14(b) shows the
response of electrode material and pulse ON time on EWR. It is noted that,
increase in pulse ON time the EWR decreases. Thermal conductivity of EN-8
electrode is low, and it resulted in less EWR. Due to these reasons, the
accumulation of heat in the electrode is less. Figure 7.14(c) presented the
response on EWR with respect to pulse current vs. pulse ON time. It is
observed from the graph, the EWR decreases for the increase in pulse current
and pulse ON time. While considering the low pulse ON time and high pulse
current the EWR is high.

7.2.11 Prediction of Optimum EWR

The optimum value of EWR ) is predicted at the selected

optimal setting of parameters. Referring to the graph (Figure 7.11), it could be
observed that the minimum EWR is achieved for the combination of optimum
parameter is A3B3C3D2. The expected mean at the optimum settings could be
estimated from Equation (7.5)

= A +B +C +D T (7.5)

Corresponding S/N ratio could be predicted as Equation (7.6)

A B C D T (7.6)

Where,

TEWR - overall mean of EWR

TEWR - S/N ratio’s overall mean of EWR

 169

A3, B3, C3 and D2 are the means of mean values of the EWR with
parameters at optimum levels. (Referring to the Table 7.13)

A3, B3, C3 and D2 are the S/N ratio mean values of the EWR
with parameters at optimum levels. (Referring to the Table 7.12)

By substitution in Equation (7.5)

= A +B +C +D T

= (25.27 + 28.37 + 26.17 + 28.52) (3 32.245)

= 11.595 %

By substitution in Equation (7.6)

= A B C D T

= ( 27.31 28.40 27.68 28.47) (3 29.57)

= 23.15 dB

Hence the predicted optimum mean value for EWR was calculated
to be 11.595 % and their corresponding S/N ratio value is – 23.15 dB.

The 95% Confidence interval (CI) for the expected results from the
confirmation experiments (CICE) and the mean of population (CIPOP) are
calculated by using the Equations (5.6) & (5.7) in chapter 5.

By substituting the following values:

27
eff 3
1 8

N = 27 x 1= 27 (treatments = 27, repetitions = 1)

 170

fe = 18 (Refer Table 7.14)

F0.05, (1, 18) = 4.41 (Tabulated F-value)

Ve = 28.55 (Refer Table 7.14)

Therefore, CICE = + 12.956 and CIPOP = + 6.478

The predicted confidence interval for confirmation experiment of

EWR is given by

(µ CI ) µ (µ + CI )

(11.595 12.956) µ (11.595 + 12.956)

(1.361 %) µ (24.551 %)

The 95% confidence intervals for the mean of population of EWR

should be given by

(µ CI ) µ (µ + CI )

(11.595 6.478) µ (11.595 + 6.478)

(5.117 %) µ (18.073 %)

The mean value of EWR is obtained through confirmation

experiments must be within confidence interval ranges. Figure 7.15 shows
the normal probability plot of EWR and it is evident the all the residuals are
found to be normally distributed along the straight line at 95% confidence
level.
 171

Figure 7.15 Probability Plot of EWR

7.2.12 Confirmation Experiments for EWR

The confirmation experiments are used to verify the process

parameters and their levels chosen from an experiment. The predicted
optimum value of EWR has been evaluated by the mean value of three
confirmation experiments at optimal levels of process parameters. The values
of EWR obtained through confirmation experiments are within 95% of CICE
and CIPOP of respective output response. The confirmation experiments and
predicted result of EWR is given in Table 7.15.

Table 7.15 Confirmation Experiment Result and Predicted Value of

EWR

Output Optimal setting of Predicted Confirmation

response process parameters optimum value experimental result

EWR (%) A3B3C3D2 11.595 16.373

 172

7.3 MULTI-OBJECTIVE OPTIMIZATION OF EDM

PARAMETERS USING GREY RELATIONAL ANALYSIS

Grey relational analysis is effectively used to solve the complicated

problems with interrelationships between the multi performance
characteristics. During the analysis, the optimization of complicated multi-
performance characteristics can be converted into optimization of a single
grey relational grade. In the present study, grey relational analysis was
applied to identify the optimum combination of the input machining
parameters such as electrode material, pulse current, pulse ON time and pulse
OFF time on multi response characteristics with an objective to maximize the
material removal rate (MRR), minimize the surface roughness (SR) and
electrode wear ratio (EWR) and also to find the effect of each parameter on
the performance characteristics at a same time. Grey relational analysis was
carried out by the following steps:

Step 1: Calculate the S/N ratio values depending upon the objective of
performance characteristics. In EDM process, the higher MRR, the
lower SR and EWR are selected as the best machining
performance characteristics. Hence, MRR is selected as a larger-
the - better S/N ratio and it was calculated by using Equation (5.1).
The SR and EWR are selected as a smaller - the - better S/N ratio it
was calculated by using Equation (5.2). The computed S/N ratios
for each performance characteristics are shown in Table 7.16.

Step 2: Normalize the S/N ratio values by using Equations (5.9) and
(5.10). The obtained response values were normalized in the
ranging from 0 to 1. The normalized S/N ratio for each
performance characteristics are given in Table 7.16.
 173

Table 7.16 Calculated S/N ratios and Normalized S/N Ratios

Ex. S/N ratios (dB) Normalized S/N ratios (dB)

No MRR SR EWR MRR SR EWR
1 -30.6923 -18.3364 -32.7580 0.007311 0.267358 0.305743
2 -26.7072 -16.4014 -31.5116 0.034931 0.504898 0.431305
3 -30.0338 -20.0976 -32.4370 0.011048 0 0.339820
4 -27.0719 -17.6755 -31.7367 0.031844 0.354509 0.409942
5 -18.9154 -15.8632 -29.4244 0.143948 0.562086 0.605129
6 -17.9171 -18.4980 -29.1942 0.166369 0.245030 0.621865
7 -32.1461 -10.0239 -35.2056 0 1 0
8 -16.2439 -14.7788 -28.3272 0.210236 0.667099 0.681067
9 -10.2375 -16.2194 -25.6116 0.459789 0.524633 0.832490
10 -22.2477 -11.2482 -32.7193 0.085297 0.930856 0.309916
11 -18.8772 -12.8394 -32.2986 0.144760 0.825122 0.354134
12 -16.1153 -17.1273 -31.6830 0.213972 0.421925 0.415088
13 -12.9873 -10.3887 -30.3024 0.324127 0.980409 0.537061
14 -14.9721 -14.1616 -30.7549 0.249716 0.721262 0.499200
15 -9.6852 -16.7290 -29.8039 0.492608 0.468309 0.576544
16 -29.4474 -10.3334 -34.1424 0.014622 0.983434 0.143444
17 -8.3610 -12.6167 -29.5033 0.580341 0.841112 0.599291
18 -3.8737 -10.2777 -25.4187 1 0.986459 0.841550
19 -27.7443 -13.1008 -31.4109 0.026483 0.805820 0.440689
20 -24.7916 -15.1129 -29.9367 0.053452 0.636128 0.566252
21 -23.9857 -18.5628 -29.5128 0.062551 0.235955 0.598577
22 -21.4732 -11.5153 -28.3158 0.096994 0.914434 0.681802
23 -15.0700 -12.8235 -24.8925 0.246466 0.826275 0.865270
24 -10.6478 -17.8808 -24.3833 0.436718 0.328148 0.886892
25 -28.4962 -12.3986 -32.4418 0.020959 0.856381 0.339323
26 -10.5683 -12.0107 -23.8430 0.441105 0.882599 0.908493
27 -6.9454 -15.3491 -21.0869 0.690171 0.613512 1
 174

Step 3: Calculate Grey Relational Co-efficient from the normalized S/N

ratio values by using Equation (5.11). The value for max is

taken as 0.5 in Equation (5.11). Since all the process parameter are
of equal weighting. The grey relational co-efficient for each
performance characteristics are given in Table 7.17.

Step 4: Next, the Grey Relational Grade (GRG) was obtained by averaging
the grey relational co-efficient for each performance characteristics
and the Equation (5.12) were used. The calculated grey relational
co-efficient and grey relational grades with order in rank are
provided in Table 7.17. Finally, the GRG are considered for
optimizing the multi response parameter design problem. The
largest value of GRG represents the optimal combination of
machining parameters for the desired performance characteristics.
 175

Table 7.17 Grey Relational Co-efficient and Grade with Rank

Ex. Grey relational coefficients Grey relational

Rank
No MRR SR EWR grade
1 0.334966 0.405633 0.418670 0.386423 26
2 0.341281 0.502461 0.467860 0.437201 23
3 0.335807 0.333333 0.430968 0.366703 27
4 0.340563 0.436494 0.458691 0.411916 25
5 0.368717 0.533098 0.558740 0.486852 19
6 0.374916 0.398416 0.569388 0.447574 21
7 0.333333 1 0.333333 0.555556 11
8 0.387668 0.600311 0.610551 0.532843 13
9 0.480672 0.512627 0.749053 0.580784 8
10 0.353431 0.878512 0.420138 0.550694 12
11 0.368938 0.740875 0.436351 0.515388 14
12 0.388794 0.463789 0.460867 0.437817 22
13 0.425216 0.962296 0.519244 0.635585 5
14 0.399909 0.642064 0.499600 0.513858 16
15 0.496331 0.484641 0.541444 0.507472 18
16 0.336615 0.967931 0.368580 0.557709 10
17 0.543680 0.758854 0.555118 0.619217 6
18 1 0.973633 0.759359 0.910997 1
19 0.339324 0.720274 0.472005 0.510534 17
20 0.345651 0.578789 0.535476 0.486639 20
21 0.347838 0.395556 0.554679 0.432691 24
22 0.356378 0.853875 0.611099 0.607117 7
23 0.398872 0.742142 0.787736 0.642917 4
24 0.470242 0.426675 0.815518 0.570812 9
25 0.338057 0.776858 0.430783 0.515232 15
26 0.472190 0.809846 0.845299 0.709112 3
27 0.617414 0.564023 1 0.727146 2
 176

Figure 7.16 Rank Plot for GRG

Figure 7.16 shows the GRG for all the experiments conducted
during machining of composite. From the figure, it is proved that experiment
number 18 has a higher grey relational grade (0.910997), which identifies the
better combination of optimal level process parameters (electrode material -
Br, pulse current – 15amps, Pulse ON time - 900µs and pulse OFF time -
60µs) for the best multiple performance characteristics with an objective to
maximize the MRR, minimize the SR and EWR for the EDM process of
AA7075-10wt. % TiO2 composite.

7.3.1 Analysis of Means for GRG

Table 7.18 shows the response table for means of grey relational
grade. From the tables, it can be understood that the pulse current has highest
delta value thus rank 1is assigned which is the most significant factor on multi
performance characteristics, followed by electrode material and pulse OFF
time. The main effect plots are used to determine the optimal level of process
parameters to obtain the maximum GRG. The main effect plot of different
 177

machining parameters like electrode material, pulse current, pulse ON time

and pulse OFF time on GRG is shown in Figure 7.17. In the plot, the x-axis
indicates the value of process parameters at three levels and y-axis indicates
the means of grey relational grade value. Horizontal line indicates the means
of mean value of the GRG. According to the figure, it can be clearly observed
that the maximum GRG is obtained by the optimum level of parameter
combinations are A2B3C3D2, which indicates that the electrode material at
level 2 (Br), pulse current at level 3 (15amps), pulse ON time at level 3 (900
µs) and pulse OFF time at level 2 (60 µs).

Table 7.18 Mean Table for GRG

Electrode Pulse Pulse ON Pulse OFF

Level
material, (A) current, (B) time, (C) time, (D)
1 0.4673 0.4582 0.5256 0.5372
2 0.5832 0.5360 0.5493 0.5903
3 0.5780 0.6343 0.5536 0.5010
Delta 0.1159 0.1761 0.0279 0.0893
Rank 2 1 4 3
Total mean grey relational grade = 0.542844

Figure 7.17 Main effect plot for GRG

 178

7.3.2 Analysis of Variance for GRG

ANOVA was used to determine the influence of EDM parameters

like electrode material, pulse current, pulse ON time and pulse OFF time on
the multi performance characteristics such as MRR, SR and EWR. Table 7.19
shows the results of ANOVA for GRG. From the table, it can be clearly
observed that the pulse current (P-value = 0.000) is the most significant factor
on the GRG followed by electrode material (P-value = 0.005). The frequency
(F-ratio) test was performed at 95% confidence interval (CI). The calculated
F-ratio values of the pulse current and electrode material were greater than
tabulated F0.05 = 3.39, which also confirms the statistical physical influence on
the multi performance characteristics. Figure 7.18 shows the percentage
contribution of each input parameter affect the GRG. The ratio of individual
sum of square of a particular input parameter to the total sum of squares of all
the input parameters gives the percentage contribution (P %) of the input
parameters on the GRG. From the figure, it can be noticed that the pulse
current is the most significant parameter that contributed (P = 39.41%)
followed by the electrode material (P = 21.69%), the pulse OFF time (P =
10.20%). These results are well agreed with the findings of Gopalakannan et
al. (2013) who reported for the Al6063-10% SiC-3%Gr hybrid metal matrix
composite during EDM machining. The R-sq and R-sq (adj) value is 72.47%
and 60.23% of GRG which indicates that the model is able to predict the
response with high accuracy.

Table 7.19 ANOVA table for GRG

Factors DF Seq. SS Adj. MS F-ratio P-value P (%)

Electrode material (A) 2 0.077129 0.038565 7.09 0.005 21.69
Pulse current (B) 2 0.140111 0.070056 12.88 0.000 39.41
Pulse ON time (C) 2 0.004075 0.002038 0.37 0.693 1.146
Pulse OFF time (D) 2 0.036294 0.018147 3.34 0.058 10.20
Residual error 18 0.097867 0.005437 - - 27.53
Total 26 0.355476 - - - 100
S = 0.0737362 R-sq = 72.47%; R=sq (adj) = 60.23%
 179

Electrode
Material (A)
Residual Error 22%
28%

Pulse OFF time Pulse

(D) Current (B)
10% 39%
Pulse ON time
(C)
1%

Figure 7.18 Percentage Contributions of the Parameters on GRG

Figure 7.19 Interaction Plot for GRG

 180

The interaction of electrode material, pulse current, pulse ON time

and pulse OFF time with GRG is shown in Figure 7.19. From the plot it is
observed that, while considering pulse current and electrode material the
increase in pulse current, increase the GRG, which means that, the improved
responses can be obtained for irrespective of all the electrode materials. While
considering the electrode material better response in observed for brass
electrode material. While interaction of pulse ON time with electrode
materials, the better responses (higher GRG) are observed for the moderate
level of pulse ON time while using both copper and EN-8 electrodes.
However better responses are observed for EN-8 electrode when compare
with copper electrodes. While consider brass electrode, it behaves differently
with other electrodes. The high value of pulse ON time (900µs) is required to
achieve better machining responses for brass electrode. While interacting the
pulse OFF time with electrode materials, the increase in pulse OFF time
increases the GRG when using the copper electrode. But for the brass and
EN-8 electrodes, the increase in pulse OFF time decreases the machining
characteristics. Better responses were observed for the pulse OFF time (60µs)
while using both brass & EN-8 electrodes. Overall, the brass electrode
provides better machining performance during EDM of AA7075- 10wt. %
TiO2 stir cast composite.

7.3.3 Prediction of Optimum GRG

The optimum value of GRG ) is predicted at the selected

optimal setting of parameters. Referring to the graph (Figure 7.17), it could be
observed that the maximum GRG is attained for the combination of optimum
parameter is A2B3C3D2. The expected mean at the optimum settings could be
estimated from Equation (7.7)

=A +B +C +D T (7.7)
 181

Where, TGRG - overall mean of GRG

A2, B3 C3 and D3 are the means of mean values of the GRG with
parameters at optimum levels. (Referring to the Table 7.18)

By substitution in Equation (7.7)

= A +B +C +D T

= (0.5832 + 0.6343 + 0.5536 + 0.5903) (3 0.542844)

= 0.732868

Hence the predicted optimum mean value for GRG was calculated
to be 0.732868.

The 95% Confidence interval (CI) for the expected results from the
confirmation experiments (CICE) and the mean of population (CIPOP) are
calculated by using the Equations (5.6) & (5.7) in chapter 5.

By substituting the following values:

27
eff 3
1 8

N = 27 x 1= 27 (treatments = 27, repetitions = 1)

fe = 18 (Refer Table 7.19)

F0.05, (1, 18) = 4.41 (Tabulated F-value)

Ve = 0.005437 (Refer Table 7.19)

Therefore, CICE = + 0.17881 and CIPOP = + 0.08941

 182

The predicted confidence interval for confirmation experiment of

GRG is given by

(µ CI ) µ (µ + CI )

(0.732868 0.17881) µ (0.732868 + 0.17881)

(0.554058) µ (0.911678)

The 95% confidence intervals for the mean of population of GRG

should be given by

(µ CI ) µ (µ + CI )

(0.732868 0.08941) µ (0.732868 + 0.08941)

(0.643458) µ (0.822278)

The mean value of GRG is obtained through confirmation

experiments must be within confidence interval ranges. Figure 7.20 shows
the normal probability plot of GRG and it is evident the all the residuals are
found to be normally distributed along the straight line at 95% confidence
level.
 183

Figure 7.20 Probability Plot of GRG

7.3.4 Confirmation Experiments for GRG

The confirmation experiments are used to verify the process

parameters and their levels chosen from an experiment. The predicted
optimum value of GRG has been evaluated by the mean value of three
confirmation experiments at optimal levels of process parameters. The values
of GRG obtained through confirmation experiments are within 95% of CICE
and CIPOP of respective output response. The confirmation experiments and
predicted result of GRG is given in Table 7.20.

Table 7.20 Confirmation Experiment Result and Predicted Value of GRG

Output Optimal setting of Predicted Confirmation

response process parameters optimum value experimental result

Grey
Relational A2B3C3D2 0.732868 0.910997
Grade