Академический Документы
Профессиональный Документы
Культура Документы
Abstract
Grey relational analysis is useful for the multi-input, discrete data and uncertain experimental study. Developed in this paper is the appli-
cation of the grey relational analysis for optimizing the electropolishing of 316L stainless steel with multiple performance characteristics.
The processing parameters (temperature, current density, and electrolyte composition) are optimized with considerations of the multiple
performance characteristics (surface roughness and passivation strength). The conducted experiments approve the effectiveness of the grey
relational analysis.
© 2003 Elsevier B.V. All rights reserved.
0924-0136/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0924-0136(03)00747-7
256 P.S. Kao, H. Hocheng / Journal of Materials Processing Technology 140 (2003) 255–259
Table 3 Table 5
Experimental results of surface roughness ratio (SRR) and corresponding Results of grey relational generation
S/N ratio
No. SRR PSR
No. SRR (%) S/N (dB)
Ideal 1 1
1 50 6.02 1 0.052724 0.770382
2 5.25 25.597 2 0.959739 1
3 45 6.936 3 0.095163 0
4 4.75 26.466 4 1 1
5 41 7.744 5 0.132598 0.770382
6 56 5.036 6 0.007135 0
7 7 23.098 7 0.843958 1
8 43 7.33 8 0.113417 0.527864
9 57 4.882 9 0 0.770382
the better, and the nominal the better. To obtain the optimal 3.2. Grey relational generation
electropolishing performance, the minimum surface rough-
ness and the maximum passivation strength are desired. The In the grey relational analysis, when the range of the
first criterion selects the-smaller-the-better characteristic of sequence is large or the standard value is enormous, the
the surface roughness. The calculation of the S/N ratio ηij function of factors is neglected. However, if the factors
for the ith experiment at the jth test is as follows: goals and directions are different, the grey relational analy-
sis might also produce incorrect results. Therefore, one has
n
1 to preprocess the data which are related to a group of se-
ηij = −10 log yij2 (3) quences, which is called “grey theory relational generation”
n
j=1 [5,6].
A linear data preprocessing method for the S/N ratio is
where yij is the ith experiment at the jth test and n is the
(0) (0)
number of tests. xi (k) − mink xi (k)
xi∗ (k) = (0) (0)
(5)
The second criterion selects the-large-the-better charac- maxk xi (k) − mink xi (k)
teristic for the passivation strength. The equation of the S/N
ratio is as follows: where xi∗ (k) is the value after the grey relational generation;
(0) (0) (0)
mink xi (k) the smallest value of xi (k); maxk xi (k) the
n (0)
1 1 largest value of xi (k).
ηij = −10 log (4)
n y 2 Based on Tables 3 and 4, the experimental results of
j=1 ij
the grey relational generation using the experimental layout
(Table 2) are shown in Table 5.
Table 3 shows the results of the surface roughness and
signal-to-noise rate. Table 4 shows the results of the passi-
3.3. Grey relational grade
vation strength and signal-to-noise rate. To consider the two
different performance characteristics, the S/N ratios corre-
In the grey relational analysis, the grey relational grade
sponding to the surface roughness and passivation strength
is used to show the relationship among the series. Let (X,
are processed by the grey relational analysis.
Γ ) be a grey relational space, X stand for the collection of
the collection of grey relational factors, xi be the compared
series, and x0 be the reference series:
Table 4
Experimental results of passivation strength ratio (PSR) and corresponding xi = (xi (1), xi (2), . . . , xi (n)), xi ∈ X, i = 1, 2, . . . , m
S/N ratio
(6)
No. PSR (%) S/N (dB)
Table 6
grey relational grade
0.8
Grey relational grade and orders
0.6
No. Grey relational grade Order
0.4
1 0.4651 6
0.2
2 0.7638 2
3 0.4006 7 0
A1 A2 A3 B1 B2 B3 C1 C2 C3
4 0.7989 1
parameter level
5 0.5037 4
6 0.3576 9 Fig. 3. Influence of electropolishing parameters on multiple performance.
7 0.7122 3
8 0.4932 5
9 0.3853 8
4. Analysis of variance
∆max = max max∆0i (k) (10) The purpose of the analysis of variance (ANOVA) is to re-
i k
veal the process parameters that significantly affect the per-
∆min = min min∆0i (k) (11) formance characteristics. The traditional statistic technique
i k
can only obtain one parameter in a single sequence, one has
ζ ∈ [0, 1] (12) to do the analysis repeatedly to obtain other factors for the
The grey relational grade is obtained: experiment [7]. Using the experimental design module in the
n Statistica software, one obtains the results of two parameter
1 sets, as shown in Table 8.
γ(x0 , xi ) = γ(x0 (k), xi (k)) (13)
n In Table 8(a), the electrolyte is the only parameter show-
k=1
ing the P-value under 0.05, namely the electrolyte is the
Table 6 shows the experimental results of the grey relational only noticeable factor for the surface roughness. Table 8(b)
grade using the experimental layout (Table 2). shows that all the three factors influence the passivation
strength significantly based on the P-values. After identify-
3.4. Factor effects ing the significant parameters, one picks the middle level
for the-smaller-the-better surface roughness and considers
Since the experimental design is orthogonal, it is then the-larger-the-better passivation strength of the results. Con-
possible to separate the effects of each polishing parameter sequently, No. 2 and No. 4 in Table 2 are considered the
at different levels. For example, the mean of grey relational optimum in the experiment.
grade for the workpiece pulse-on time at level 1, 2 and 3
can be calculated by taking the average of the grey rela-
tional grade for the experiments 1–3, 4–6 and 7–9, respec- 5. Verification test
tively (Table 2). The mean of the grey relational grade for
each level of other machining parameters can be computed Once the optimal level of the process parameters is iden-
in the similar manner. The mean of the relational grade for tified, the following step is to verify the improvement of the
each level of the combining parameters is summarized in the performance characteristic using this optimal level. The es-
multi-response performance index table (Table 7). In addi- timated S/N ratio η̃ is calculated as
tion, the total mean of the grey relational grade of the eight
α
experiments is also calculated, as shown in Table 7. Fig. 3
shows the grey relational grade. The dash line shows the η̃ = ηm + (η̂i − ηm ) (14)
i=1
value of the total mean of grey relational grade. Basically,
the larger the grey relation grade is, the closer will be the where ηm is the total mean of the grey relational grade, η̂i the
product quality to the ideal value. Thus the larger grey rela- mean of the grey relational grade at the optimal level, and α
tional grade is desired. the number of the process parameter that significantly affect
the multiple performance characteristics. Based on Eq. (14),
Table 7 the estimated grey relational grade can be obtained.
Influence of electropolishing parameters on grey relational grade Table 9 shows the results of the confirmation experiment
Grey relational grade using the optimal electropolishing parameters. As shown in
Table 10, the surface roughness ratio is reduced from 50 to
Level 1 Level 2 Level 3 Max−min
4.75%, the passivation strength ratio is increased from 95 to
Temperature 0.543 0.553 0.530 0.023 100%. The comparison of the S/N ratios between the initial
Current density 0.659 0.587 0.381 0.278 and the optimal parameters is also shown in Table 10. It is
Electrolyte composition 0.439 0.649 0.539 0.21
clear that the surface roughness and passivation strength are
Mean value of the grey relational grade = 0.543
considerably improved through this study.
P.S. Kao, H. Hocheng / Journal of Materials Processing Technology 140 (2003) 255–259 259
Table 8
Results of ANOVA
SS d.f. MS F P