A Note on
“Capability Assessment for
Processes with Multiple Characteristics: A
Generalization of the Popular Index
C
pk
”
W. L. Pearn, C. H. Wu*
,†and M. C. Tsai
The generalized yield indexCT
pkestablishes the relationship between the manufacturing specifications and the actual process
performance, which provides a lower bound on process yield for two-sided processes with multiple characteristics. The results attended are very practical for industrial application. In this article, we extended the results in cases with one-sided specification and multiple characteristics. The generalized index CT
PUwas considered, and the asymptotic distribution of the
natural estimator ^CPUT was developed. Then, we derived the lower confidence bounds as well as the critical values of index CT PU.
We not only provided some tables but also presented an application example. Copyright © 2012 John Wiley & Sons, Ltd. Keywords: critical values; lower confidence bounds; multiple characteristics; one-sided specification; process capability index
1. Introduction
P
rocess yield has been the most basic and common criterion used in the manufacturing industry as a base for measuring process performance. Recently, Pan and Lee2developed two novel indices to evaluate the performance of multivariate manufacturing process. The effects of the estimation of process capability index (PCI) on the nonconforming units in parts per million (NCPPM) estimates are analyzed by Ozkaya and Testik.3Later, Lin and Pearn4used the yield index Spkto deal with process selection problem.Itay et al.5investigated an advanced multistage sampling plan based on Cpkindex.
Afterward, Yum and Kim6provided a bibliography of PCIs for 2000–2009. Spiring7then presented several method of process capability using Mathematica 7 software. Next, an applicable methodology to achieve the robustness of the multivariate process capability vector was proposed by Awad and Kovach.8Lately, more dissertations about PCI were published such as Pearn et al.,9 Hsu et al.,10Yen and Pearn,11Pearn and Cheng,12and Goethals and Cho.13It can be observed that the new investigations of PCI mainly focus on processes with multivariate or multiple characteristics.
Pearn et al.14proposed a generalization of the popular index Cpkfor evaluating the yield of a gold bumping manufacturing process
with multiple characteristics. The CT
pk index is defined for a process with multiple characteristics and two-sided specifications.
However, the quality characteristics often have only one-sided specification. At this time, the overall capability index CT
PUproposed
by Wu and Pearn1is considered for this purpose. The index CT
PUis defined as follows: CT PU¼ 1 3Φ 1 Ym i¼1 Φ 3Cð PUiÞ ( ) (1)
where CPUi denotes the CPUvalue of the ith characteristic for i = 1, 2,. . ., m, and m is the number of characteristics. Φ() is the
cumulative distribution function of standard normal distribution. A one-to-one correspondence relationship between CT
PU and the
overall process NCPPM can be demonstrated as
NCPPM¼ 106 1 Φ 3CT PU
(2)
Department of Industrial Engineering and Management, National Chiao Tung University, Taiwan, ROC
*Correspondence to: C. H. Wu, Department of Industrial Engineering and Management, National Chiao Tung University, 1001 University Road, Hsinchu, Taiwan 300, ROC.
†E-mail: hexjacal.iem96g@nctu.edu.tw
(wileyonlinelibrary.com) DOI: 10.1002/qre.1295 Published online 8 February 2012 in Wiley Online Library
2.
Approximation distribution of the natural estimator
Consider the natural estimator of CT PUas ^CPUT¼ 1 3Φ 1 Ym i¼1 Φ 3^CPUi ( ) ¼13Φ1 Ym i¼1 Φ USL Xi Si ( ) ; i ¼ 1; . . . ; m (3) where Xi and Sidenote the sample mean and sample variance of ith characteristic. The exact distribution of ^CPUT is mathematically
intractable. By taking thefirst order of the Taylor expansion (see Appendix), the asymptotic distribution of ^CPUTis
^CPUT N CT PU; 1 9nf 3CT PU 2 Xm i¼1 a2i þ b2i ! (4) where ai¼ Ym j¼1;j6¼i Φ 3CPUj " # f 3CPUj and bi¼ 3CPUi ffiffiffi 2 p ai; i ¼ 1; . . . ; m
3.
Estimation and testing on
C
T PUFrom Equation (4), it can be seen that ^CPUT is an asymptotic unbiased estimator of ^C
PUT. The determination of the lower confidence
bound on the actual process capability is essential for quality assurance. An approximate 100(1 a)% lower confidence bound for CT PUcan be expressed as CT PULB ^CPUT Za 1 9nf 3CT PU 2 Xm i¼1 a2 i þ b2i " #1=2 (5)
It is noted that aiand biparameters are unknown in Equation (5). To investigate the effects of CPUion C
T
PULB, the cases with processes
that have two independent characteristics are considered. Figure 1 displays the curves of CT
PULBfor various combinations of CPU1and
CPU2, with C
T
PU=1.0, 1.33, 1.5, 1.67. Then, we examined the results presented in Figure 1, which indicate that
(i) CT
PULBobtains its absolute maximum as CPU1 ¼ CPU2.
(ii) The minimum CT
PULB occurs when one of CPUi approaches infinity, that is, another CPUi equals C
T
PU. Under this condition, the
minimum CT
PULBis the most reliable lower confidence bound for a given CPUT .
By the discussion mentioned earlier, after doing some algebra, the asymptotic distribution of ^CPUT becomes
^CPUT N CT PU; 1 9nþ 1 2nC T PU2 (6) Therefore, an approximate 100(1 a)% lower confidence bound for CT
PUcan be expressed as CPUT LB¼2^C T PU ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 4z2 a 9nþ 2z2 a n ^C T PU2 2z4 a 9n2 q 2 z2 a=n (7) Table I tabulates the 95% lower confidence bound CT LB
PU for ^CPUT =1.0(0.1)2.0, n = 10(10)400. For the convenience of hypothesis
testing, we also provide the critical value. From the asymptotic distribution listed in Equation (6), the critical value to hypothesis testing H0: CPUT ⩽C versus Ha: CTPU> C is expressed as
c0¼ C þ za ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 9nþ C2 2n r : (8)
Table II performs the critical values for type I error a = 0.05 with C=1.0(0.1)2.0, n = 10(10)400. Next, an application example is presented.
4.
A case study
We applied the methodology to a set of real data (n = 100) presented in Wu and Pearn1for measuring manufacturing capability of a process making couplers and wavelength division multiplexers (WDM). Two quality characteristic including the polarization-dependent
loss and the insertion loss, which are critical infiber-optic transmission quality, are considered. Table III (cited from Wu and Pearn1) displays
the manufacturing capabilities and the corresponding NCPPM for coupler and WDM process using ^CPUT values and CT PULB.
In the coupler case, if the quality requirement is CT
PU⩾; 1:3, some statistical inferences can be made. First, because the 95% lower
confidence bound CT
PULB¼1.35588 > 1.3, we say that the process satisfies the requirement. Moreover, CTPULB¼1.35588 means that there
are no more than 23.7 NCPPM, or, from Table III, the critical value 1.460835 (n = 100, C = 1.3) is less than the observation value ^CPUT =1.5261.
The two results are agreed. In WDM case, ^CPUT = 0.7352< 1 is obviously inadequate and incapable for high-tech product manufacturing.
T PU C =1.0 T PU C =1.33 T PU C =1.5 T PU C =1.67 T LB PU C T LB PU C T LB PU C T LB PU C Figure 1. Curves of CT
PULBwitha = 0.05, n = 10(20)90 (bottom to top in plot) and CPUT =1.0, 1.33, 1.5, 1.67
Table I. 95% lower confidence bounds of CT
PUfor ^CTPU=1.0(0.1)2.0, n = 10(10)400 n ^CT PU 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 10 0.6920 0.7684 0.8443 0.9198 0.9950 1.0700 1.1447 1.2192 1.2936 1.3679 1.4420 20 0.7661 0.8477 0.9291 1.0101 1.0909 1.1716 1.2521 1.3324 1.4127 1.4929 1.5729 30 0.8024 0.8867 0.9708 1.0547 1.1384 1.2219 1.3053 1.3886 1.4718 1.5550 1.6380 40 0.8252 0.9113 0.9972 1.0828 1.1683 1.2537 1.3390 1.4241 1.5092 1.5943 1.6792 50 0.8414 0.9287 1.0158 1.1027 1.1895 1.2762 1.3628 1.4493 1.5358 1.6221 1.7085 60 0.8536 0.9419 1.0299 1.1178 1.2056 1.2933 1.3809 1.4684 1.5559 1.6433 1.7307 70 0.8633 0.9523 1.0411 1.1298 1.2184 1.3069 1.3953 1.4836 1.5719 1.6601 1.7483 80 0.8712 0.9608 1.0503 1.1396 1.2288 1.3180 1.4070 1.4960 1.5850 1.6739 1.7627 90 0.8778 0.9680 1.0580 1.1479 1.2376 1.3273 1.4169 1.5065 1.5960 1.6854 1.7749 100 0.8835 0.9741 1.0646 1.1549 1.2451 1.3353 1.4254 1.5154 1.6054 1.6953 1.7852 120 0.8928 0.9841 1.0753 1.1664 1.2574 1.3483 1.4392 1.5300 1.6208 1.7115 1.8022 150 0.9032 0.9954 1.0874 1.1794 1.2712 1.3630 1.4548 1.5465 1.6381 1.7297 1.8213 180 0.9111 1.0039 1.0965 1.1891 1.2816 1.3741 1.4665 1.5588 1.6511 1.7434 1.8357 200 0.9153 1.0084 1.1015 1.1944 1.2873 1.3801 1.4728 1.5655 1.6582 1.7509 1.8435 250 0.9237 1.0175 1.1112 1.2048 1.2984 1.3919 1.4854 1.5788 1.6722 1.7656 1.8589 300 0.9300 1.0243 1.1185 1.2126 1.3067 1.4008 1.4948 1.5887 1.6827 1.7766 1.8705 350 0.9349 1.0296 1.1242 1.2188 1.3133 1.4077 1.5021 1.5965 1.6909 1.7852 1.8795 400 0.9389 1.0339 1.1289 1.2237 1.3186 1.4134 1.5081 1.6028 1.6975 1.7922 1.8869
161
References
1. Wu CW, Pearn WL. Measuring manufacturing capability for couplers and wavelength division multiplexers. The International Journal of Advanced Manufacturing Technology 2005; 25:533–541. DOI: 10.1007/s00170-003-1793-9
2. Pan JN, Lee CY. New Capability Indices for Evaluating the Performance of Multivariate Manufacturing Process. Quality and Reliability Engineering International 2010; 26:3–15. DOI: 10.1002/qre.1024
3. Ozkaya BY, Testik MC. On the Expected Parts Per Million Nonconforming Levels Obtained from Estimated Process Capability Indices. Quality and Reliability Engineering International 2010; 26:817–829. DOI: 10.1002/qre.1156
4. Lin CJ, Pearn WL. Process Selection for Higher Production Yield Based on Capability Index Spk. Quality and Reliability Engineering International 2010;
26:247–258. DOI: 10.1002/qre.1051
5. Itay N, Yisrael P, Edna S. Developing a Sampling Plan Based on Cpk-Unknown Variance. Quality and Reliability Engineering International 2011;
27:3–14. DOI: 10.1002/qre.1094
6. Yum BJ, Kim KW. A Bibliography of the Literature on Process Capability Indices : 2000–2009. Quality and Reliability Engineering International 2011; 27:251–268. DOI: 10.1002/qre.1115
7. Spiring F. Exploring Process Capability with Mathematica. Quality and Reliability Engineering International 2011; 27:369–387. DOI: 10.1002/qre.1112 8. Awad MI, Kovach JV. Multiresponse Optimization using Multivariate Process Capability Index. Quality and Reliability Engineering International 2011;
27:465–477. DOI: 10.1002/qre.1141
9. Pearn WL, Kang HY, Lee AHI, Liao MY. Photolithography Control in Wafer Fabrication Based on Process Capability Indices with Multiple Characteristics. IEEE Transactions on Semiconductor Manufacturing 2009; 22(3):351–356. DOI: 10.1109/TSM.2009.2024851
10. Hsu YC, Pearn WL, Chuang YF. Sample size determination for production yield estimation with multiple independent process characteristics. European Journal of Operational Research 2009; 196:968–978. DOI: 10.1016/j.ejor.2008.04.029
11. Yen CH, Pearn WL. Select better suppliers based on manufacturing precision for processes with multivariate data. International Journal of Production Research 2009; 47(11):2961–2974. DOI: 10.1080/00207540701796985
12. Pearn WL, Cheng YC. Measuring production yield for processes with multiple characteristics. International Journal of Production Research 2010; 48(15):4519–4536. DOI: 10.1080/00207540701796985
13. Goethals PL, Cho BR. The Development of a Target-Focused Process Capability Index with Multiple Characteristics. Quality and Reliability Engineering International 2011; 27:297–311. DOI: 10.1002/qre.1120
14. Pearn WL, Shiau JJH, Tai YT, Li MY. Capability Assessment for Processes with Multiple Characteristics: A Generalization of the Popular Index Cpk.
Quality and Reliability Engineering International 2011; 27(8):1119–1129. DOI: 10.1002/qre.1200
Table II. Critical values c0for C =1.0(0.1)2.0, n = 10(10)400
n C 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 10 1.4066 1.5401 1.6741 1.8086 1.9433 2.0783 2.2134 2.3488 2.4843 2.6200 2.7557 20 1.2875 1.4112 1.5353 1.6596 1.7841 1.9089 2.0338 2.1588 2.2839 2.4091 2.5344 30 1.2347 1.3541 1.4737 1.5936 1.7136 1.8338 1.9541 2.0746 2.1951 2.3156 2.4363 40 1.2033 1.3200 1.4370 1.5543 1.6716 1.7891 1.9067 2.0244 2.1421 2.2600 2.3778 50 1.1818 1.2968 1.4120 1.5274 1.6429 1.7586 1.8743 1.9901 2.1060 2.2219 2.3379 60 1.1660 1.2796 1.3935 1.5076 1.6218 1.7360 1.8504 1.9648 2.0793 2.1939 2.3085 70 1.1536 1.2663 1.3792 1.4922 1.6053 1.7185 1.8318 1.9452 2.0586 2.1721 2.2856 80 1.1437 1.2556 1.3676 1.4798 1.5920 1.7044 1.8169 1.9294 2.0419 2.1545 2.2672 90 1.1355 1.2467 1.3580 1.4695 1.5811 1.6927 1.8044 1.9162 2.0281 2.1400 2.2519 100 1.1285 1.2391 1.3499 1.4608 1.5718 1.6828 1.7940 1.9051 2.0164 2.1276 2.2389 120 1.1173 1.2270 1.3368 1.4468 1.5568 1.6669 1.7771 1.8873 1.9975 2.1078 2.2181 150 1.1049 1.2136 1.3224 1.4313 1.5402 1.6493 1.7584 1.8675 1.9767 2.0859 2.1951 180 1.0958 1.2037 1.3117 1.4198 1.5280 1.6363 1.7446 1.8529 1.9613 2.0697 2.1781 200 1.0909 1.1984 1.3060 1.4137 1.5214 1.6293 1.7371 1.8450 1.9530 2.0609 2.1689 250 1.0813 1.1880 1.2948 1.4017 1.5086 1.6156 1.7226 1.8297 1.9368 2.0440 2.1511 300 1.0742 1.1803 1.2865 1.3928 1.4991 1.6055 1.7120 1.8184 1.9249 2.0314 2.1379 350 1.0687 1.1744 1.2801 1.3859 1.4918 1.5977 1.7036 1.8096 1.9156 2.0217 2.1277
Table III. Calculations for process capability of the coupler and WDMS
Characteristic ^CPUT NCPPM ^C
PUT LB NCPPM
Coupler 1.5261 2.3439 1.3588 22.86916
WDM 0.7352 13706.01 0.6425 26958.67
Authors' biographies
Wen-Lea Pearn received the Ph.D. degree in operations research from the University of Maryland, College Park .He is a Professor of Operations Research and Quality Assurance at the National Chiao-Tung University (NCTU), Hsinchu, Taiwan. He was with Bell Labora-tories, Murray Hill, NJ, as a Quality Research Scientist before joining the NCTU, and others. His current research interests include pro-cess capability, network optimization, and production management. Dr. Pearn’s publications have appeared in the Journal of the Royal Statistical Society, Series C, Journal of Quality Technology, European Journal of Operational Research, Journal of the Operational Research Society, Operations Research Letters, Omega, Networks, and the International Journal Productions Research.
Chia-Huang Wu received his MS degree in Applied Mathematics from National Chung-Hsing University. Currently, he is a PhD candidate at the Department of Industrial Engineering and Management, National Chiao Tung University, Taiwan, ROC.
Meng-Chun Tsai received her MS degree in Department of Industrial Engineering and Management, National Chiao Tung University, Taiwan, ROC.