Int. J. Production Economics 106 (2007) 506–512
A Bayesian approach based on multiple samples for measuring
process performance with incapability index
Gu-Hong Lin
Department of Industrial Engineering and Management, National Kaohsiung University of Applied Sciences, Kaohsiung, Taiwan, ROC Received 14 October 2004; accepted 29 June 2006
Available online 15 September 2006
Abstract
Process incapability index, which provides an uncontaminated separation between information concerning the process
accuracy and the process precision, has been proposed to the manufacturing industry for measuring process performance.
Investigations concerning the estimated incapability index have focused on single sample in existing quality and statistical
literatures. However, contributions based on multiple samples have been comparatively neglected. In this paper,
investigations based on multiple samples are considered for normally distributed processes. A Bayesian approach to obtain
an upper bound for the incapability index is proposed. A computational program using Maple software is proposed to
evaluate critical values required to ensure the posterior probability reaching a certain desirable level for the incapability
index. A practical example is also provided to illustrate how the proposed reliable Bayesian procedure may be applied in
process capability assessment.
r
2006 Elsevier B.V. All rights reserved.
Keywords: Bayesian; Incapability index; Multiple samples
1. Introduction
Process capability indices, whose purpose is to
provide numerical measures on whether or not a
manufacturing process is capable of reproducing
items satisfying the quality requirements preset by
the customers, the product designers, have received
substantial research attention in the quality control
and statistical literature. The three basic capability
indices C
p, C
aand C
pk, have been defined as (
Kane,
1986
;
Pearn et al., 1998
;
Lin, 2006a
):
C
p¼
USL LSL
6s
,
(1)
C
a¼
1
m m
d
,
(2)
C
pk¼
min
USL m
3s
;
m LSL
3s
,
(3)
where USL and LSL are the upper and lower
specification limits preset by the customers, the
product designers, m is the process mean, s is the
process standard deviation, m ¼ ðUSL þ LSLÞ=2
and d ¼ ðUSL LSLÞ=2 are the mid-point and half
length of the specification interval, respectively.
The index C
preflects only the magnitude of the
process variation relative to the specification
toler-ance and, therefore, is used to measure process
potential. The index C
ameasures the degree of
process centering (the ability to cluster around the
ARTICLE IN PRESS
www.elsevier.com/locate/ijpe
0925-5273/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpe.2006.06.012
Tel.: +886 7 3814526x7124; fax: +886 7 3923375. E-mail address:ghlin@cc.kuas.edu.tw.
exp
1
2
X
m i¼1X
n j¼1x
ijx
s
2
"
#)
d
1
2
X
m i¼1X
n j¼1x
ijx
s
2
"
#
¼
ffiffiffiffiffiffiffiffiffi
p
2mn
r
X
m i¼1X
n j¼1x
ijx
ffiffiffi
2
p
2
"
#
ðmn1Þ=2G
mn 1
2
.
Therefore, the posterior distribution of y ¼ ðm; sÞ
given x is
f ðyjxÞ ¼
hðy x
j Þ
k
¼
2 exp½ðs
2b
Þ
1s
mnGða
Þðb
Þ
affiffiffiffiffiffiffiffiffiffi
mn
2ps
2r
exp
mn
2
m x
s
2
"
#
,
where x ¼ fx
11; x
12; . . . ; x
mng, N
omoN, 0o
soN, a
¼ ðmn 1Þ=2, b
¼
2½ðmnÞS
2mn1¼
2=
S
mi¼1S
nj¼1x
ijx
2h
i
.
Appendix B
Derivation of expression (8): The posterior
probability
p
¼
PrfC
ppoC
0jxg
¼
Prf½ðm TÞ
2þ
s
2=D
2oC
0jxg
¼
Prfs
2þ ðm T Þ
2oðD
ffiffiffiffiffiffi
C
0p
Þ
2jxg
¼
Z
b 0Z
TþgðsÞ TgðsÞf ðm; s x
j Þ
dm ds
¼
Z
b 02 exp½ðs
2b
Þ
1s
mnGða
Þðb
Þ
affiffiffiffiffiffiffiffiffiffi
mn
2ps
2r
(
Z
T þgðsÞ T gðsÞexp
mn
2
m x
s
2
"
#
dm
)
ds
¼
Z
b 02 exp½ðs
2b
Þ
1s
mnGða
Þðb
Þ
aF
ffiffiffiffiffiffiffi
mn
p
½T x þ gðsÞ
s
F
ffiffiffiffiffiffiffi
mn
p
½T x gðsÞ
s
ds,
where b ¼ D
ffiffiffiffiffiffi
C
0p
, gðsÞ ¼
p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b
2s
2, a
¼ ðmn
1Þ=2, b
¼
2½ðmnÞS
2mn1, F is the cumulative
dis-tribution function of the standard normal
distribu-tion N(0, 1).
Let
y
¼
b
0s,
b
0¼
2=½S
mi¼1S
nj¼1ðx
ijT Þ
2,
g
¼
1 þ ðd
Þ
2,
d
¼
x T
=s
mn,
b
1ðy
Þ ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2ðd
Þ
2=½y
ð1þðd
Þ
2Þ
q
, b
2ðy
Þ¼
p
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
mnf½1=ðt
y
Þ
1
g,
t
¼
mn ~
C
pp
=ð2C
0Þ, then the posterior probability
based on C
ppwith multiple samples can be
expressed as
p
¼
Z
1=t 0F b
1ðy
Þ þ
b
2ðy
Þ
F b
1ðy
Þ
b
2ðy
Þ
ðg
Þ
aðy
Þ
aþ1Gða
Þ
exp
1
g
y
dy
.
Appendix C
Digits: ¼ 6: p
0: ¼ 0.95: c
0: ¼ 1: m: ¼ 25: n: ¼ 5:
delta: ¼ 0.4137:
t: ¼ Cpp-(m*n*Cpp)/(2*c
0):
b1: ¼ y-(2*delta^2/(y*(1+delta^2)))^0.5:
b2: ¼ (Cpp,
y)-(m*n*(2*c0/(m*n*Cpp*y)1))
^0.5:
c1: ¼ (Cpp, y)
-b1(y)+b2(Cpp, y):
c2: ¼ (Cpp, y)
-b1(y)b2(Cpp, y):
f1: ¼ (Cpp, y)
-(1+erf(c1(Cpp, y)*2^(0.5)))/
2(1+erf(c2(Cpp, y)*2^(0.5)))/2:
f2: ¼ (Cpp,
y)
-exp(1/(y*(1+delta^2)))*
(1+delta^2)^((m*n1)/2)*y^((m*n+1)/2)/
GAMMA((m*n1)/2):
f: ¼ (Cpp, y)
-f1(Cpp, y)*f2(Cpp, y):
pstar: ¼ Cpp-int(f(Cpp, y), y ¼ 0..(2*c
0)/(m*n*
Cpp)):
cofpstar: ¼ proc(r::numeric)
local Cpp;
Cpp: ¼ r;
if
evalf(pstar(Cpp))
o ¼ p
0and
evalf(pstar
(Cpp0.0001))4p
0then Cpp
elif evalf(pstar(Cpp))
op
0and evalf(pstar(Cpp
0.0001))
op
0then Cpp: ¼ Cpp0.0001: cofpstar(Cpp)
else Cpp: ¼ Cpp+0.0001: cofpstar(Cpp)
end if
end proc:
evalf(cofpstar(0.7938));
References
Boyles, R.A., 1991. The Taguchi capability index. Journal of Quality Technology 23, 17–26.
Chan, L.K., Cheng, S.W., Spiring, F.A., 1988. A new measure of process capability: Cpm. Journal of Quality Technology 20,
162–175.
ARTICLE IN PRESS
Chen, K.L., Chen, K.S., Li, R.K., 2005. Suppliers capability and price analysis chart. International Journal of Production Economics 98, 315–327.
Chou, Y.M., Owen, D.B., 1989. On the distributions of the estimated process capability indices. Communication in Statistics—Theory and Methods 18, 4549–4560.
Greenwich, M., Jahr-Schaffrath, B.L., 1995. A process incap-ability index. International Journal of Quality & Reliincap-ability Management 12, 58–71.
Johnson, T., 1992. The relationship of Cpmto squared error loss.
Journal of Quality Technology 24, 215–221.
Kane, V.E., 1986. Process capability indices. Journal of Quality Technology 18, 41–52.
Kotz, S., Pearn, W.L., Johnson, N.L., 1993. Some process capability indices are more reliable than one might think. Applied Statistics 42, 55–62.
Lin, G.H., 2002. Process incapability index for contaminated normal processes. Advances and Applications in Statistics 2, 119–130.
Lin, G.H., 2004. Upper limits of the estimated incapability index: A practical application on the reliability assessment of the
printed circuit boards. International Journal of Advanced Manufacturing Technology 24, 841–846.
Lin, G.H., 2005. Process reliability assessment with a bayesian approach. International Journal of Advanced Manufacturing Technology 25, 392–395.
Lin, G.H., 2006a. A random interval estimation on the estimated process accuracy index. International Journal of Advanced Manufacturing Technology 27, 969–974.
Lin, G.H., 2006b. Assessing process performance with incap-ability index based on multiple samples. International Journal of Advanced Manufacturing Technology 27, 715–719. Pearn, W.L., Lin, G.H., 2001. On the reliability of the estimated
process incapability index. Quality & Reliability Engineering International 17, 279–290.
Pearn, W.L., Lin, G.H., 2002. Estimated incapability index: reliability and decision making with sample information. Quality & Reliability Engineering International 18, 141–147. Pearn, W.L., Lin, G.H., Chen, K.S., 1998. Distributional and inferential properties of the process accuracy and process precision indices. Communications in Statistics—Theory and Methods 27, 985–1000.
ARTICLE IN PRESS
G.-H. Lin / Int. J. Production Economics 106 (2007) 506–512 512