Documente Academic
Documente Profesional
Documente Cultură
A THESIS
Submitted by
M. USHA
in partial fulfillment for the award of the degree
of
DOCTOR OF PHILOSOPHY
DEPARTMENT OF MATHEMATICS
KALASALINGAM UNIVERSITY
(Kalasalingam Academy of Research and Education)
ANAND NAGAR
KRISHNANKOIL 626 126
MAY 2015
ii
iii
iv
v
vi
ABSTRACT
on the sample average and sample standard deviations of the quality characteristic.
When the distribution of the quality characteristic in the lot or process is known,
variables sampling plans that have specified risks of accepting and rejecting lots of
given quality may be designed. The primary advantage of the variables sampling
plan is that the same operating characteristic (OC) curve can be obtained with a
smaller sample size than would be required by an attributes sampling plans. Thus a
variables acceptance sampling plan would require less sampling. The
measurements data required by a variables sampling plan would probably cost
more per observation than the collection of attributes data. However, the reduction
in sample size obtained may more than offset this increased cost. When
destructive testing is employed, variables sampling is particularly useful in
reducing the costs of inspection. Another advantage is that measurements data
usually provide more information about the manufacturing process or lot than do
attributes data. Generally, numerical measurements of the quality characteristics
are more useful than simple classification of the item as conforming or non-
conforming. A final point to be emphasized is that when acceptable quality levels
are very small, the sample size required by attributes sampling plans are very
large. Under these circumstances, there may be significant advantages in switching
to variables measurements. Thus as many manufacturers begin to emphasize
allowable numbers of non-conforming parts per million (ppm), variables sampling
plan becomes very attractive.
Govindaraju and Balamurali (2000). So, this thesis is devoted to the optimal
designing of certain special purpose acceptance sampling plans for variables
inspection. These new plans will be very effective, efficient and attractive in terms
of reducing the cost of inspection. These plans will be particularly applied for
costly and destructive testing.
number and the constraints are related to lot acceptance probabilities at acceptable
quality level and limiting quality level under the operating characteristic curve.
ACKNOWLEDGEMENT
Thanks God, the merciful and the passionate, for providing me the
opportunity to step in the excellent world of science. To be able to step strong and
smooth in this way, I have also been supported by many people to whom I would
unstinted support and enormous encouragement to bring out this thesis. In his I
appreciate and thank, for his free availability and approach. I pray to God for his
well being.
facilities to carry out the research. I am extremely grateful to the Chairman and
Many friends have helped me stay sane through these difficult years. I
greatly value their friendship and I deeply appreciate their belief in me. I would
like to thank Ms. M. Jeyadurga , who as a good friend, is always willing to help
Most importantly, none of this would have been possible without the
my family.
the first place and supporting me spiritually throughout my life. I would also like
to thank my sister and brother. They were always supporting me and encouraging
M. Usha
xiii
TABLE OF CONTENTS
ABSRACT iv
LIST OF TABLES xvi
LIST OF FIGURES xix
LIST OF SYMBOLS AND ABBREVIATIONS xx
1. INTRODUCTION 1
Section 1 Basic Concepts of Quality Control 2
Section 2 Basic Concepts of Acceptance Sampling 6
Section 3 A Review of Variables Sampling Plans 20
Section 4 A Review on Certain Special Purpose Sampling 23
Plans by Attributes
Section 5 A Review on Special Purpose Sampling Plans 33
by Variables
LIST OF TABLES
LIST OF FIGURES
The following is the glossary of symbols and abbreviations used in this thesis.
N Lot Size
n Sample Size
p Lot or process quality or fraction non-conforming
Pa (p) Probability of acceptance as function of lot quality.
p1 Acceptable Quality Level (AQL)
p2 Limiting Quality Level (LQL)
Producers risk
Consumers risk
n1 First stage sample size
n2 Second stage sample size
d Number of non-conforming items
c Acceptance number in attributes single sampling plan
kN Acceptance criteria of normal inspection
kT Acceptance criteria of tightened inspection
Population standard deviation
S2 Sample variance
U Upper specification limit
L Lower specification limit
n Sample size for known sigma plan
k Acceptance criteria for known sigma plan
k Rejection criteria for known sigma plan
ns Sample size for unknown sigma plan
ks Acceptance criteria for unknown sigma plan
xxiv
INTRODUCTION
SECTION 1
BASIC CONCEPTS OF QUALITY CONTROL
1.1 QUALITY
In manufacturing sector, quality is a measure of excellence or a state
of being free from defects, deficiencies and significant variations. Quality of a
product is brought about by the strict and consistent adherence to measurable and
verifiable standards to achieve uniformity of output that satisfies specific consumer
or user requirements.
Incoming inspection
Product moving from one stage to other
In process
Machine start up
Process monitoring
Process adjustment
Final product
Field surveillance
SQC is systematic as compared to guess work of haphazard process
inspection. The mathematical and statistical approaches neutralize personal bias
and uncover poor judgment. The SQC consists of three general activities:
Systematic collection and graphic recording of accurate data.
Analyzing the data.
Practical engineering or management or management action, if the
information obtained indicates significant deviations from the specified
limits.
SECTION 2
BASIC CONCEPTS OF ACCEPTANCE SAMPLING
1.2.1 INTRODUCTION
Acceptance sampling is an important field of SQC that was
popularized by Dodge and Romig (1959) and originally applied by the U.S.
military to the testing of bullets during World War II. If 100 percent inspection
were executed in advance, no bullets would be left to shipment. If, on the other
hand, none were tested, malfunctions might occur in the field of battle, which may
result in potential disastrous result. Dodge proposed a middle way reasoning that
a sample should be selected randomly from a lot, and on the basis of sampling
information, a decision should be made regarding the disposition of the lot. In
general, the decision is either to accept or reject this lot. This process is called Lot
Acceptance Sampling or just acceptance sampling.
Single sampling plans and double sampling plans are the most basic
and widely applied testing plans when simple testing is needed. Multiple sampling
plans and sequential sampling plans provide marginally better disposition decision
at the expense of more complicated operating procedures. Other plans such as the
continuous sampling plan, bulk-sampling plan, and Tightened-Normal-Tightened
scheme etc., are well developed and frequently used in their respective working
condition.
4. Special purpose plans including chain sampling, skip-lot sampling and small
sample plans etc.
INSPECTION
NORMAL INSPECTION
Inspection that is used in accordance with an acceptance sampling
scheme when a process is considered to be operating at or slightly better than its
acceptance quality level.
TIGHTENED INSPECTION
A feature of a sampling scheme using stricter acceptance criteria than
those used in normal inspection.
10
REDUCED INSPECTION
A feature of a sampling scheme permitting smaller sample sizes than
those used in normal inspection.
HYPERGEOMETRIC MODEL
This is an exact model for the case of non-conforming units under
Type A situations and is useful for isolated lots. In this model the probability mass
function is given by
m N m
x n x
P X x
N
n
where N is the population size
n is the size of the sample
k is the number of successes
BINOMIAL MODEL
This model is exact for the case of non-conforming units under type B
situations. This can also be used for type A situations for the case of non-
conforming units, whenever (n /N) 0.10. Under type B situation, for the case of
non conforming units, Poisson model can be used whenever n is large and p is
small such that np < 5.
n n x
P X x p x 1 p , x 0,1,2,3.........n
x
where N is the lot size
n is the sample size taken from a lot
p is the proportion defective in the sample
13
POISSON MODEL
The Poisson distribution can be applied to systems with a large number
of possible events, each of which is rare to occur. The probability mass function is
given by
e x
P X x , x 0 ,1, 2 , 3 .......
x!
is a positive real number, equal to the expected number of occurrences during
the given interval or average.
NORMAL DISTRIBUTION
No area of statistics seems to have escaped the impact of the normal
distribution. This is certainly true of acceptance sampling where it forms the basis
of a large number of variables acceptance sampling plans. It has pervaded other
areas of acceptance sampling as well. The normal distribution is completely
specified by two parameters and .
Hamaker (1960) also pointed out that these aims are partly conflicting
and all of them cannot be simultaneously realized.
16
Non Bayesian 1 2
Bayesian 3 4
According to Peach (1947), the following are some of the major types
of designing the plans, based on the OC curves, which are classified according to
types of risk protection.
1. The plan is specified by requiring the OC curve to pass through two fixed
points. In some cases, it may be possible to impose certain additional
conditions also.
17
The two points generally selected are (p1, 1-) and (p2, ) where,
2. The plan is specified by fixing one point only through which the OC curve is
required to pass and one or more conditions, not explicitly in terms of the
OC curves. Dodge and Romig (1959) LTPD Sampling plans are the
examples for this type of designing.
The main advantage of the variables sampling plan is that the same OC
curve can be obtained with a smaller sample size than would be required by an
attributes sampling plan. Thus, a variables acceptance sampling plan would require
less sampling. The measurements data required by a variables sampling plan
would probably cost more per observation than the collection of attributes data.
However, the reduction in sample size obtained may more than offset this
increased cost. When destructive testing is employed, variables sampling is
particularly useful in reducing the costs of inspection. Another advantage is that
measurements data usually provide more information about the manufacturing
process or lot than do attributes data. Generally, numerical measurements of
quality characteristics are more useful than simple classification of the item as
conforming or non-conforming. Another advantage of the variables sampling plan
is that when AQLs are very small, the sample size required by it is very less than
the attributes sampling plans. Under these circumstances, the variables sampling
plans have significant advantages. For compliance testing of a measurable
characteristic, a variable sampling plan may be preferred.
20
SECTION 3
A REVIEW OF VARIABLES SAMPLING PLANS
When the standard deviation of the lot quality is known, the criteria for
acceptance and the associated mathematical computations get simplified. When
products are manufactured by automatic machinery whose inherent variation is
known and tested, we have an example where the lot standard deviation is known.
When we assume the lot standard deviation as known and give it a particular value
it is assumed as constant. We assume that the directly measurable quality
characteristic X follows the normal law of pattern of variation in the lot these
assumptions must be examined and reviewed from time to time when variables
plans with known sigma are in use. The n units in the sample are measured and the
values x1 , x 2 ,....x n are obtained. The mean is calculated. If the individual product
21
quality has an upper specification limit U then acceptance criteria for the lot based
on the single sampling results would be if x k U , then accept the lot and if
x k U , then reject the lot. In the case of unknown sigma variables sampling
plan, sample standard deviation S is used instead of .
acceptance sampling plans. Bravo and Wetherill (1980) developed a method for
designing variables double sampling plans with OC curves matching with the OC
curves of the equivalent single sampling plans. Sommers (1981) developed tables
for selecting variables double sampling plans and matched variables single
sampling plan having two fixed points on the OC curve. Schilling (1982) has
written an exclusive book on acceptance sampling. Bruhn Suhr and Krumbholz
(1990) studied the variables single sampling with double specification limits for
normally distributed quality characteristics. Collani (1990) criticized the variables
sampling plans and argued that the acceptance sampling by variables is
inappropriate if one is interested in the fraction non-conforming in incoming
batches. But, Seidel (1997) has proved that sampling by variables is always
optimal. Baillie (1992) developed tables for variables double sampling plans when
the process standard deviation is unknown. Hamilton and Lesperance (1995)
described the operating characteristics of the variables single sampling plans
having double specification limits. Govindaraju and Kuralmani (1998) have
studied the nature of the OC curve of known sigma single sampling variables plan.
Jun et al. (2006) developed variables acceptance sampling plans for Weibull
distributed items under sudden death testing.
SECTION 4
A REVIEW ON CERTAIN SPECIAL PURPOSE SAMPLING PLANS BY
ATTRIBUTES
Some of the special purpose sampling plans are Chain sampling plan
(ChSP) of Dodge (1955), Repetitive group sampling (RGS) plan of Sherman
(1965), Multiple deferred/dependent state sampling plans of Wortham and Baker
(1976), Quick Switching System (QSS) of Romboski (1969), Tightened-Normal-
Tightened (TNT) sampling scheme developed by Calvin (1977), Skip-lot sampling
plan (SkSP) of Perry (1973) etc.
MEASURES OF QSS
The important measures of QSS that describe the operation of an
acceptance sampling plan for various fraction nonconforming are,
PT
Pa ( p )
1 PN PT
where PN Pr v k N is the probability of accepting a lot based on a
OPERATING PROCEDURE
The operating procedure of the attributes ChSP-1 is as follows.
Step1: For each lot, select a sample of n units and test each unit for conformance
to the specified requirements
28
Step 2: Accept the lot if the observed number of non-conforming units d is zero,
reject the lot if d 2.
Step 3: Accept the lot if d is one and if no defective units are found in the
immediately preceding i samples of size n.
Thus a ChSP 1 plan has two parameters namely n, the sample size for
each submitted lot and i, the number of previous samples on which the decision of
acceptance or rejection of the lot is based.
MEASURES OF ChSP
The important measures of ChSP that describe the operation of an
acceptance sampling plan are,
Pa p P0 , n P1, n P0 , n
i
ASN p n
AOQ p pPa ( p )
29
OPERATING PROCEDURE
The operating procedure of the attributes TNT scheme is as follows.
Step 1: Start with the tightened inspection level using the single sampling
attributes plan with sample size n1 and the acceptance number c. Accept
the lot if the number of non-conforming units, d c and reject the lot if
d > c. If t lots in a row are accepted under tightened inspection, then
switch to normal inspection.
Step 2: During the normal inspection, inspect the lots using the single sampling
attributes plan with a sample size n2 and the acceptance number c .
Accept the lot if d c and reject the lot if d > c. Switch to tightened
inspection after a rejection of lot if an additional lot is rejected in the next
s lots.
30
P2 P1
Pa ( p )
where
s t
2 P2 1 P1
s
, s 1 and t
, t s
(1 P2 )(1 P2 ) P1 (1 P1 )
n mn
ASN ( p ) ,m 1
s t
2P2 1P1
where s
, s 1 and t
, t s
(1P2)(1P2 ) P1 (1 P1)
AOQ p pPa ( p)
the quality of the lot is not improved by sorting etc. They have also discussed the
need for a provision for resampling of lots in case of zero acceptance sampling
plans. A resubmitted lot is defined in the ANSI/ASQC Standard A2-1987 (1987)
as the one which has been designated as not-acceptable and which is submitted
again for acceptance inspection after having been further tested, sorted,
reprocessed etc. If the lot is not accepted on original inspection, the producer may
test it and may also resubmit it without sorting or reprocessing it for resampling.
OPERATING PROCEDURE
The operating procedure of the attributes resampling scheme is as follows.
Step 1: Perform original inspection. i.e., apply a reference (single) sampling plan
(with a sample size n and acceptance number c).
Step 2: On non acceptance on the original inspection, apply the reference plan
m times and reject the lot if it is not accepted on (m-1)st resubmission.
32
1 1 Pa ( p )
m
where Pa ( p ) Pr v k
2. The Average Sample Number (ASN) is
ASN ( p ) n 1 Pa ( p )n 1 Pa ( p ) n .... 1 Pa ( p )
2 m 1
n
n 1 1 Pa ( p)
m
Pa ( p)
AOQ p pP a p
33
SECTION 5
A REVIEW ON SPECIAL PURPOSE SAMPLING PLANS BY VARIABLES
CHAPTER 2
OPTIMAL DESIGNING OF
VARIABLES QUICK SWITCHING SAMPLING SYSTEM
BY MINIMIZING THE AVERAGE SAMPLE NUMBER
2.1 INTRODUCTION
This chapter deals with optimal designing of variables quick switching
system (VQSS) where the quality characteristic under study follows normal
distribution and has upper specification limit or lower specification limit. The
known sigma as well as unknown sigma VQSS are designed by minimizing the
average sample number by formulating nonlinear programming problem where the
constraints are related to lot acceptance probabilities at AQL and LQL. Tables are
constructed for finding the optimal parameters of the known sigma as well as
unknown sigma VQSS. The results obtained are compared with that of the existing
plans and proved that the results obtained are optimal.
Step 1: Start with normal inspection. During normal inspection, take a random
n
1
X
n
X
i 1
i .
Step 2: Accept the lot if v k N and reject the lot if v k N . If a lot is rejected on
normal inspection, then switch to tightened inspection as in Step 3.
X , X U X , 1 n
The OC function of the VQSS, which gives the proportion of lots that
are expected to be accepted for given product quality, p under known sigma case
is given by
PT Pr(v kT )
Pa ( p ) (2.1)
1 PN PT 1 Pr(v k N ) Pr(v k T )
accepting a lot based on a single sampling plan with parameters (n, kT). Under
Type B situation (i.e. a series of lots of the same quality), forming lots of N items
from a process and then drawing random sample of size n from these lots is
equivalent to drawing random samples of size n directly from the process. Hence
the derivation of the OC function is straightforward.
U
p 1 1 (v ) ( v ) (2.2)
where ( y ) is given by
y
1 z2
( y) exp dz , (2.3)
2 2
( wT )
Pa ( p ) (2.4)
1 (w N ) ( wT )
where wT v k T n and wN v k N n
( wT 2 )
and (2.6)
1 ( w N 2 ) ( wT 2 )
Here wT1 is the value of wT at p=p1, wN1 is the value of wN at p=p1, wT2 is the value
of wT at p=p2 and wN2 is the value of wN at p=p2.
wT 2 (v 2 k T ) n and w N 2 (v 2 k N ) n (2.7)
where v1 is the value of v at AQL and v2 is the value of v at LQL. For given AQL
or LQL, the values of kN , kT and the sample size n are determined by using a
search procedure.
of ASN is meaningful under Type B sampling situations. It is also known that the
ASN of the known sigma VQSS is
n n
ASN ( p ) n (2.8)
The ASN given above can be used as an objective function to solve for
the parameters (n, kN, kT). Since there are several choices to obtain the objective
function, it is considered here to minimize ASN at AQL. If the objective is to
minimize the ASN at AQL, then the problem will be reduced to the following
nonlinear optimization problem.
Minimize ASN(p1)= n
Subject to
Pa ( p1 ) 1
Pa ( p 2 )
n 1, k N k T 0 (2.9)
where Pa ( p1 ) and Pa ( p 2 ) are the lot acceptance probabilities at AQL and LQL
respectively and are given in (2.5) and (2.6) respectively.
Step 1: Start with the normal inspection level using the variables single sampling
plan with a sample size nS and the acceptance criterion kNS. Accept the
U X
lot if v k NS and reject the lot if v k NS , where v ,
S
1 nS
(X i X )2
X
nS
X
i 1
i and S
nS 1
. If a lot is rejected under normal
Step 2: During the tightened inspection, inspect the lots using the variables
single sampling plan with a sample size nS and the acceptance criterion
kTS(>kNS). Accept the lot if v kTS and reject the lot if v kTS , where
U X 1 nS
(X i X )2
v
S
, X
nS
X
i 1
i and S
nS 1
. If a lot is accepted
in Step 1.
Thus, the unknown sigma VQSS has the parameters namely the sample
size nS, and the acceptable criterion kNS and kTS. If kNS=kTS, then the VQSS will be
reduced to the variables single sampling plan with unknown standard deviation.
2
variance k NS Var ( S ) (see Duncan (1986)).
nS
That is.,
2 2
2
X k NS S ~ N k NS , k NS
nS 2n S
Therefore, the probability of accepting a lot under normal inspection is given by
P X U k NS S p P X k NS S U p
U k NS nS
(v k NS ) 2
2
( / n ) 1 k NS k NS
1
S 2
2
nS
If we let w NS (v k NS ) 2
then the probability of acceptance under
k NS
1
2
tightened inspection is considered (w NS ) .
41
nS
Similarly if we let wTS (v k TS ) 2
then the probability of acceptance
k
1 TS
2
under tightened inspection is taken as ( wTS ) . Hence the lot acceptance
probability of the proposed system for sigma unknown case under two-points on
the OC curve is given by
( wT 1S )
Pa ( p1 ) (2.10)
1 ( w N 1S ) (wT 1S )
(wT 2 S )
and Pa ( p 2 ) (2.11)
1 ( w N 2 S ) ( wT 2 S )
nS nS
w N 1S (v1 k NS ) 2
, w (v k )
T 1S 1 TS 2
k NS k TS
1 1
2 2
nS nS
wN 2 S (v 2 k NS ) 2
and w (v k )
T 2 S 2 TS 2
k NS k TS
1 1
2 2
In this case, the optimization problem becomes,
Minimize ASN(p1) = nS
Subject to
Pa ( p1 ) 1
Pa ( p 2 )
n s 1, k NS k TS 0 (2.12)
42
2.9 EXAMPLES
For the above example, the operation of the VQSS is explained as follows.
Step 1: Take a random sample of size 12 from the submitted lot for inspection
U X , where 1 12
and compute v
X X i . Accept the lot if v 1.552
12 i 1
and reject the lot if v 1.552 . If a lot is rejected, then switch to tightened
inspection as in step 2.
Table 2.1 can also be used for the selection of the parameters of the
unknown VQSS for given values of AQL and LQL. Suppose that AQL=1%,
LQL=5%, =5% and =10%. From Table 2.1, the parameters of the VQSS can be
determined as nS = 24, kNS = 1.729 and kTS = 2.214.
In this section, we will discuss the advantages of the VQSS over attributes
QSS and variables single sampling plans. For the purpose of comparison, we will
consider the plans which have the same AQL and LQL.
44
2.11. COMPARISON
minimum ASN than the attributes single sampling plan (see Romboski (1969) and
Soundararajan and Arumainayagam (1992)). This should be valid for variables
sampling also. Hence the entire design of variables QSS provided in
Soundararajan and Palanivel (1997) for both known and unknown sigma seems
faulty or doubtful. Hence the parameters given in this chapter are more reliable
and optimum.
Table 2.2 shows the ASN values of the variables single sampling plan
and the variables double sampling plan along with the VQSS for some arbitrarily
selected combinations of AQL and LQL under known sigma case. Table 2.3
gives the ASN values of the above said plans when sigma is unknown. These
ASN values are calculated at the producers quality level for both known and
unknown sigma plans. The sample size of the variables single sampling plan and
the ASN of the variables double sampling plan can be found in Sommers (1981).
Tables 2.4 and 2.5 apparently show that the VQSS will have minimum
ASN when compared to the variables single and double sampling plans for both
known and unknown sigma cases. Similar reduction in ASN can be achieved for
any combination of AQL and LQL values. This implies that VQSS will give
desired protection with minimum inspection so that the cost of inspection will
greatly be reduced. Thus the VQSS provides better protection than the variables
single sampling plans and variables double sampling.
46
Table 2.1. Variables Quick Switching Sampling Systems Indexed by AQL and LQL
for =5% and =10% Involving Minimum ASN
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n kT kN nS kTS kNS
0.001 0.002 134 2.998 2.943 234 3.193 2.773
0.003 45 2.958 2.833 108 3.166 2.666
0.004 31 2.899 2.784 104 2.999 2.704
0.005 14 2.994 2.599 71 2.998 2.628
0.006 11 2.985 2.535 60 2.955 2.600
0.007 10 2.939 2.519 45 2.985 2.525
0.008 14 2.769 2.639 38 2.982 2.482
0.009 8 2.897 2.457 35 2.947 2.467
0.010 7 2.899 2.409 32 2.926 2.446
0.012 9 2.710 2.525 28 2.880 2.415
0.015 6 2.751 2.381 24 2.819 2.379
0.025 4 2.672 2.222 16 2.715 2.260
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n kT kN nS kTS kNS
0.0075 0.010 149 2.499 2.244 575 2.502 2.252
0.012 58 2.527 2.137 209 2.549 2.144
0.015 34 2.485 2.080 107 2.549 2.059
0.020 23 2.391 2.041 64 2.486 1.996
0.025 14 2.402 1.922 47 2.426 1.951
0.030 11 2.367 1.867 36 2.395 1.900
0.035 11 2.263 1.893 31 2.337 1.877
0.040 11 2.179 1.909 26 2.313 1.838
0.050 8 2.151 1.816 20 2.266 1.776
0.060 7 2.086 1.781 17 2.196 1.741
0.070 6 2.036 1.686 14 2.175 1.685
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n kT kN nS kTS kNS
0.015 0.090 7 1.888 1.508 16 1.956 1.511
0.100 5 1.856 1.376 14 1.927 1.477
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n kT kN nS kTS kNS
0.18 7 1.460 1.090 12 1.521 1.096
0.200 4 1.317 0.882 10 1.496 1.046
From Figure 2.1, it can be easily observed that, for good quality, i.e. for smaller
values of fraction nonconforming, the composite OC curve (OC curve of the
VQSS) coincides with the OC curve of the variables single sampling plan (10,
1.754). As quality deteriorates the OC curve of the composite OC curve moves
toward that for the single sampling plan (10, 2.179) and comes close to it beyond
the indifference quality level.
0.9
0.6
VQSS
0.5
0.2
0.1
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Fraction Nonconforming, p
Table 2.2: ASN Values of the Known Sigma Variables Single Sampling Plan,
Variables Double Sampling Plan and VQSS
ASN
p1 p2 Variables Variables Variables* Variables **
SSP DSP QSS QSS
Table 2.3: ASN Values of the Unknown Sigma Variables Single Sampling
Plan, Variables Double Sampling Plan and VQSS
ASN
Variables Variables Variables* Variables **
SSP DSP QSS QSS
p1 p2
0.001 0.002 1032 829.1 1032 234
0.001 0.003 381 302.4 412 108
0.005 0.010 547 437.1 941 123
0.005 0.012 327 263.0 823 89
0.03 0.04 1333 1138.7 945 269
0.03 0.06 197 316.5 357 89
0.04 0.07 258 417.6 263 104
0.04 0.08 159 125.8 201 47
0.05 0.07 660 535.4 768 132
0.05 0.08 319 258.0 572 73
* ASN given in Soundararajan and Palanivel (1997)
** ASN given in Table 2.1
52
*Parameters of **Parameters of
p1 p2 VQSS VQSS
n kN kT n kN kT
0.001 0.002 370 2.93 3.10 134 2.943 2.998
0.001 0.003 95 2.87 3.00 45 2.833 2.958
0.005 0.010 399 2.40 2.60 31 2.183 2.683
0.005 0.012 126 2.39 2.50 23 2.146 2.641
0.03 0.04 1856 1.73 2.00 116 1.679 1.934
0.03 0.05 515 1.73 1.90 33 1.493 1.998
0.04 0.06 965 1.62 1.78 58 1.476 1.801
0.04 0.08 121 1.55 1.68 21 1.311 1.796
0.05 0.06 1311 1.55 1.66 280 1.521 1.666
0.05 0.07 853 1.53 1.66 66 1.373 1.723
*Parameters of **Parameters of
VQSS VQSS
p1 p2 nS kNS kTS nS kNS kTS
0.001 0.002 2053 2.93 3.10 234 2.773 3.193
0.001 0.003 504 2.87 3.00 108 2.666 3.166
0.005 0.010 1647 2.40 2.60 123 2.205 2.705
0.005 0.012 502 2.39 2.50 89 2.169 2.664
0.03 0.04 5100 1.73 2.00 269 1.665 1.965
0.03 0.05 1365 1.73 1.90 89 1.535 1.999
0.04 0.06 2362 1.62 1.78 104 1.427 1.892
0.04 0.08 279 1.55 1.68 47 1.339 1.819
0.05 0.06 3001 1.55 1.66 462 1.488 1.708
0.05 0.07 1939 1.53 1.66 132 1.366 1.761
CHAPTER 3
3.1 INTRODUCTION
The following assumptions should be valid for the application of the VQSS.
(i) Production is in a steady state, so that results of past, present and future lots
are broadly indicative of a continuing process.
(ii) Lots are submitted for inspection serially either in the order of production or
in the order of being submitted for inspection.
54
Step 1: Start with normal inspection. During normal inspection, take a random
U X , where 1 n
v2
X
n
X
i 1
i
Step 2: Accept the lot if v1 k1N and v2 k2 N . Reject the lot if v1 k1N and/or
v2 k2 N . If a lot is rejected on normal inspection, then switch to tightened
X , X
... X n and compute v1
X L and v U X , where
1 2 2
n
1
X
n
X
i 1
i .
55
Step 4: Accept the lot if v1 k1T and v2 k2T . Reject the lot if v1 k1T and/or
v2 k2T (k1T k1N and k2T k 2 N ) . If a lot is accepted on tightened
sampling plan with double specification limits. Also when the true mean is located
at the middle of double specification limits, that is, ( L U ) / 2 , and if
k1N k 2 N and k1T k2T , then it is called as the symmetric fraction
P(v1 k1T , v 2 k 2T )
Pa ( p) (3.2)
1 P(v1 k1N , v 2 k 2 N ) P(v1 k1T , v2 k 2T )
56
sampling plan with parameters n , k1T , k 2T . Under type B situation (i.e. a series
of lots of the same quality), forming lots of N items from a process and then
drawing random sample of size n from these lots is equivalent to drawing random
samples of size n directly from the process. Hence the derivation of the OC
function is straightforward.
p P{ X L} P{ X U } (3.3)
L
z1 p / 2 z p / 2 (3.6)
where z is the standard normal variate corresponding to the tail probability of .
PT Pv1 k T , v 2 kT p = P L k T X U kT p
PT 2 (( z p / 2 k T ) n ) 1 (3.7)
where k N z p /2 .
2[ (( z p / 2 kT ) n )] 1
Pa ( p) (3.9)
2 (( z p / 2 k T ) n ) 2 (( z p / 2 k N ) n ) 1
Pa ( p ) 2 (( z p / 2 k ) n ) 1 (3.10)
where n is the sample size of the variables single sampling plan and k is the
acceptance criterion.
U
Then, z p (3.12)
U
L
and z1 pL z pL (3.13)
PT Pv1 k1T , v 2 k 2T p= P L k1T X U k 2T p
PT (( z pU k 2T ) n ) ((k1T z pL ) n ) (3.14)
PN 1 (( z pU k 2 N ) n ) ((k1N z pL ) n ) (3.15)
The above function reduces to the OC function of the single sampling plan with
double specifications when k1T k1N k1 (say) and k 2T k 2 N k 2 (say). In this
case, (3.16) reduces to
Pa ( p ) (( z pU k 2 ) n ) ((k1 z pL ) n ) (3.17)
where n is the sample size of the variables single sampling plan and k1 and k 2 are
the acceptance criterion under double specification limits.
Step 1: Start with normal inspection. During normal inspection, take a random
U X , 1 n 1 n 2
v2
S
where X
n i 1
X i and S
n 1 i 1
Xi X .
60
Step 2: Accept the lot if v1 k1NS and v2 k 2 NS . Reject the lot if v1 k1NS and/or
v2 k 2 NS . If a lot is rejected on normal inspection, then switch to tightened
1 n 1 n 2
X
n i 1
X i and S
n 1 i 1
Xi X .
Step 4: Accept the lot if v1 k1TS and v2 k 2TS . Reject the lot if v1 k1TS and/or
v2 k 2TS ( k1TS k1NS and k 2TS k 2 NS ). If a lot is rejected on tightened
Thus, the proposed sampling system under unknown standard deviation is also
characterized by five parameters n , k1NS , k 2 NS , k1TS and k 2TS .
Here we consider only the symmetric case for the purpose of explaining
the designing methodology. That is, it will be assumed that k1NS k 2 NS k NS and
k1TS k 2TS kTS .
Then, the probability of accepting a lot based on the tightened inspection becomes
PT PL k TS S X U kTS S p
2
2
k 2
X kS ~ N k , (3.19)
n 2n
2n
PT 2 ( z p / 2 kTS ) 1 (3.20)
2 kTS
2
2n
PN 2 2 ( z p / 2 k NS ) (3.21)
2 k NS
2
than 1-, whereas the probability should be less than when the quality is at the
specified LQL.
If the AQL and LQL are designated as p1 and p 2 , respectively, then the
probability of acceptance should satisfy the following conditions.
Pa ( p1 ) 1
Pa ( p2 ) (3.22)
There may exist multiple solutions to satisfy the above two inequalities,
so the objective function of minimizing the ASN will be considered. The ASN for
the VQSS at the quality level of p can be determined as
ASN ( p ) n (3.23)
Minimize ASN ( p )
Subject to
Pa ( p1 ) 1
Pa ( p2 )
n 1 , 0 k2 N k2T min( z pU , z pU )
1 2
used. The SQP has been implemented in Matlab Software using the routine
fmincon.
Four tables are developed and provided for the selection of optimal
parameters of the VQSS. Table 3.1 and Table 3.2 give the optimal parameters of
the known sigma VQSS for symmetric and asymmetric fraction non-conforming
respectively, whereas Table 3.3 and Table 3.4 provide the optimal parameters of
the VQSS for symmetric and asymmetric fraction non-conforming respectively
for the specified values of AQL and LQL when (1 ) =0.95 and =0.1.
Table 3.2 provides the design parameters of the known sigma VQSS
(asymmetric case) for the specified values of AQL and LQL when =0.05 and
=0.1. Here, p L and pU were assumed to be 1/4 and 3/4, respectively, of AQL
or LQL.
p1 =0.01 and p 2 =0.05 at =0.05 and at =0.1 respectively. Then, Table 3.2 gives
and k 2T 1.802 .
Step 1: Start with normal inspection. During normal inspection, take a random
Step 2: Accept the lot if v1 1.759 and v2 1.443 . Reject the lot if v1 1.759 and/or
v2 1.443 . If a lot is rejected on normal inspection, then switch to tightened
inspection as in Step 3.
Step 3: During tightened inspection, take a random sample of size 4, and compute
X L and v U X , 1 4
v1
2
where X Xi .
4 i 1
Step 4: Accept the lot if v1 2.168 and v2 1.802 . Reject the lot if v1 2.168
and/or v2 1.802 . If a lot is rejected on tightened inspection, then
immediately switch to normal inspection as in Step 1.
In a similar way, the unknown sigma VQSS can be developed and for
the easy selection of the optimal parameters Table 3.3 and Table 3.4 are also
constructed. Table 3.3 shows the design parameters of the unknown sigma VQSS
(symmetric case) and Table 3.4 gives the optimal parameters of unknown sigma
65
VQSS (asymmetric case) for the specified values of AQL and LQL for =0.05
and =0.1.
Wu and Pearn (2008) stated the example as Liquid crystals have been
used for display applications with various configurations. Most of the produced
displays recently involve the use of either twisted nematic (TN), or super twisted
nematic (STN) liquid crystals. The technology of the STN display was introduced
recently to improve the performance of LCD as an alternative to the TFT. A
larger twist angle can lead to a significantly larger electro-optical distortion. This
leads to a substantial improvement in the contract and viewing angles over TN
displays. An increasing number of personal computers are now network-ready
and multimedia-capable and are equipped with CD-ROM drives. Due to advances
in telecommunications technology, simple monochromatic displays are no longer
in popular demand. The next generation of telecommunication products will
require displays with rich, graphic quality images and personal interfaces.
Therefore, future display s must be clearer and sharper to meet these demands.
Until this point, STN-LCD has been used mainly to display still images, and
because of the slow response time needed to process still images, STN-LCD has
not been able to reproduce animated images at an adequate contrast level. Thus,
with the growing popularity of multimedia applications, there is a need for PCs
equipped with color STN-LCD capable of processing animated pictures instead of
still images. The space between the glass substrate is filled with liquid crystal
66
material and the thickness of the liquid crystal is kept uniform with glass fibers or
plastic balls as spacers. Thus, the STN-LCD is sensitive to the thickness of the
glass substrates.
To illustrate how the proposed VQSS can be established and applied to the
actual data collected from the factories, we present a case study on STN-LCD
manufacturing process as proposed by Wu and Pearn (2008). The STN-LCD is
popularly used in making the PDA (personal digital assistant), Notebook personal
computer, Word Processor, and other Peripherals. The factory manufactures
various types of the LCD. For a particular model of the STN-LCD investigated,
the upper specification limit (U) of a glass substrates thickness is 0.77 mm, the
lower specification limit (L) of a glass substrates thickness is 0.63 mm. If the
product characteristic does not fall within the specification limits (L, U), the
lifetime or reliability of the STN-LCD will be discounted. In the contract, the
AQL and LQL are set to 0.05 and 0.1 with =5% and =10% respectively.
Therefore, the problem for quality practitioners is to determine the optimal
parameters of the proposed sampling system that provide the desired levels of
protection for both the producer and the consumer. Suppose that the quality
characteristic of interest has symmetric fraction non-conforming for specified
AQL and LQL conditions and the variance of the process is unknown. Based on
the proposed procedure, we can obtain the optimal parameters from Table 3.3 as
n =33, k TS 1.632 and k NS 1.329 . In this example, the proposed VQSS with
double specification limits can be implemented as follows.
67
Step 1: Start with normal inspection. From each submitted lot, take a random
sample of size 33. Suppose that the data may be as follows.
1 nS (X i X )2
For this data, calculate X X i 0.681424 and S
n i 1 n 1
0.041264 .
( X L) 0.051424
Also calculate v1 = 1.24622
S 0.041264
(U X ) 0.088576
and v2 = 2.14657
S 0.041264
Step 3: During tightened inspection, take a random sample of size 33, from the
next consecutive lot. In this case, the data may be as follows.
0.695 0.764 0.786 0.699 0.757 0.732 0.657
0.730 0.764 0.695 0.744 0.765 0.671 0.653
0.718 0.693 0.742 0.735 0.679 0.753 0.739
0.655 0.623 0.751 0.699 0.639 0.691
0.642 0.675 0.739 0.748 0.666 0.742
68
1 nS (X i X )2
For this data, calculate X X i 0.68203 and S 0.043865 .
n i 1 n 1
( X L) 0.05203
Also calculate v1 = 1.18614
S 0.043865
(U X ) 0.08797
and v2 = 2.00547
S 0.043865
3.9 COMPARISONS
For the purpose of comparing the proposed system with the other
sampling plans, we provide Table 3.5, which shows the parametric values of
known and unknown sigma variables single sampling plans with double
specification limits for symmetric fraction non-conforming. It can be observed that
the proposed VQSS results a smaller sample size than the sample size of single
sampling plan for both known sigma and unknown cases. The reduction ratios are
much higher for the sigma unknown case than for the sigma known case. It is also
to be pointed out that the sample size for the VQSS with double specification
limits is smaller than the VQSS with single specification limit (refer Balamurali
and Usha (2012 a)) for any specified combinations of AQL and LQL.
pointed out that the normal distribution can be justified due to the central limit
theorem as long as the statistics related to X is used. Obviously, there are some
situations in that the normal distribution is not suitable. If the distribution of
quality characteristic is known to follow any distribution other than the normal
distribution, sometimes we can utilize the analytical solutions to design the plans
but sometimes we cannot. It may depend on the statistics to be used. In such cases,
use of appropriate distribution is advisable. Hence, we can say that the use of
normal distribution is always an approximation only. Montgomery (1985)
investigated the effect of non-normality in the variables sampling plans. Some of
the authors have studied the effect of non-normality in variables sampling plans
and developed appropriate variables sampling plans depends upon the distribution
of the quality characteristic. For further details, readers are advised to refer
Srivastava (1961), Zimmer and Burr (1963), Das and Mitra (1964), Singh (1966),
Owen (1969), Takagi (1972), Kocherlakota and Balakrishnan (1986), Lam (1994),
Sahli et al. (1997), Suresh and Ramanathan (1997) and Chen and Lam (1999), Das
et al. (2002)).
3.11 CONCLUSION
In this chapter, we have developed a sampling system which can be
applied when the quality characteristic of interest has two specification limits
namely the lower and upper specification limits. Whenever the quality
characteristic involves double specification limits, separate sampling plan/system
should be developed in a different manner compared to the single specification
limit sampling plans. In this chapter, procedures and methodologies of determining
the optimal parameters of the VQSS have been developed for the inspection of
measurable characteristics having double specification limits. We have constructed
tables for both symmetric and the asymmetric fraction nonconforming cases for
both the known sigma VQSS and the unknown VQSS. We have also made a
70
comparison of the proposed system with variables single sampling plan having
double specification limits. It has been proved that the sample size required for the
proposed system is lesser than the sample size of the variables single sampling
plan with double specification limits.
71
Optimal Parameters
p1 p2 n kT kN
0.001 0.002 42 3.082 2.972
0.003 17 2.948 2.789
0.004 10 2.863 2.648
0.006 6 2.719 2.471
0.008 4 2.617 2.283
0.010 3 2.538 2.117
0.015 2 2.391 1.853
0.020 2 2.278 1.868
Optimal Parameters
p1 p2 n kT kN
0.03 0.150 3 1.401 1.001
0.300 1 0.991 0.121
0.100 16 1.632 1.452
0.150 6 1.408 1.128
Optimal Parameters
p1 p2 n k1T k2T k1N k2N
0.001 0.002 42 3.272 2.968 3.163 2.859
0.003 16 3.143 2.842 2.967 2.678
0.004 10 3.048 2.753 2.828 2.548
0.006 6 2.912 2.608 2.642 2.363
0.008 4 2.818 2.513 2.449 2.182
0.010 3 2.742 2.429 2.281 2.018
0.015 2 2.589 2.294 2.013 1.757
0.020 2 2.468 2.172 2.029 1.779
Optimal Parameters
p1 p2 n k1T k2T k1N k2N
0.03 0.150 3 1.668 1.252 1.218 0.868
0.300 1 1.259 0.839 0.259 0.053
Optimal Parameters
p1 p2 n kTS kNS
0.001 0.002 200 3.084 2.859
0.003 72 2.952 2.627
0.004 42 2.848 2.473
0.006 23 2.712 2.289
0.008 15 2.608 2.052
0.010 12 2.528 2.029
0.015 8 2.383 1.903
0.020 6 2.264 1.772
Optimal Parameters
p1 p2 n kTS kNS
0.03 0.060 46 1.857 1.578
0.090 15 1.669 1.269
0.120 8 1.523 1.018
0.150 5 1.404 0.690
0.300 2 0.959 0.528
Optimal Parameters
p1 p2 n k1TS k2TS k1NS k2NS
0.001 0.002 228 3.262 2.969 3.139 2.878
0.003 80 3.129 2.848 2.937 2.698
0.004 44 3.028 2.758 2.793 2.564
0.006 24 2.892 2.628 2.572 2.369
0.008 15 2.793 2.528 2.442 2.242
0.010 12 2.712 2.449 2.318 2.129
0.015 8 2.549 2.229 2.073 1.903
0.020 6 2.438 2.187 1.909 1.753
Optimal Parameters
p1 p2 n k1TS k2TS k1NS k2NS
0.03 0.120 8 1.749 1.388 1.378 1.119
0.150 5 1.618 1.268 1.182 0.948
0.300 2 1.179 0.859 0.752 0.532
0.100 35 1.849 1.473 1.679 1.329
0.150 12 1.659 1.268 1.348 1.049
Table 3.5. Average Sample Number of Variables Single Sampling Plans and
VQSS with Double Specification Limits (Symmetric Fraction Non-
conforming)
0.05 0.100 34 16 79 33
0.150 13 6 25 10
0.200 8 4 13 6
0.250 6 1 9 4
0.500 3 1 3 2
81
CHAPTER 4
4.1 INTRODUCTION
(ii) Lots are submitted for inspection serially either in the order of
production or in the order of being submitted for inspection.
(iii) Inspection is by measurements, with quality is defined as the
fraction nonconforming, p.
(iv) The distribution of the quality characteristic must be known and
follows normal distribution.
In addition, the usual conditions for the application of variables single sampling
plan with known or unknown standard deviation should also be valid (see for
further details Schilling (1982), Grant and Leavenworth (1996) and Montgomery
(2005)).
Suppose that the quality characteristic of interest has the upper specification limit
U and follows a normal distribution with known standard deviation .
Step 1: Start with the tightened inspection level using the single sampling
variables plan with a sample size n1 and the acceptance criterion k.
n1
1
X1
n1
X
i 1
i .
Step 2: If t lots in a row are accepted under tightened inspection, then switch to
normal inspection.
Step 3: During the normal inspection, inspect the lots using the single sampling
variables plan with a sample size n2 (<n1) and the acceptance criterion k.
83
n2
1
X2
n 2
X
i 1
i .
s
2 P2
where s
, s 1
(1 P2 )(1 P2 )
t
1 P1
t
, ts
P1 (1 P1 )
s
, s 1
(1 ( w2 ))(1 ( w2 ) )
85
1 ( w1 )
t
, ts
( w1 )t (1 ( w1 ))
( w21 ) ( w11 )
1 (4.5)
(w22 ) (w12 )
and (4.6)
Here w11 is the value of w1 at p=p1, w21 is the value of w2 at p=p1, w12 is the value
of w1 at p=p2 and w22 is the value of w2 at p=p2. That is,
n mn
ASN ( p) , m>1 (4.7)
s
2 P2
where s
, s 1
(1 P2 )(1 P2 )
t
1 P1
t
, ts
P1 (1 P1 )
where P1 and P2 are the probability of acceptance of tightened and normal plans
respectively. The ASN given above can be used as an objective function to solve
for the parameters (n1, n2, k). Since there are several choices to obtain the
objective function, it is considered here to minimize ASN at AQL. If the objective
is to minimize the ASN at AQL, then the problem will be reduced to the following
nonlinear optimization problem.
Minimize ASN(p1)
Subject to
Pa ( p1 ) 1
Pa ( p 2 )
n 1, m 1, k 0 (4.8)
where Pa ( p1 ) and Pa ( p 2 ) are the lot acceptance probabilities at AQL and LQL
respectively and are given in (4.5) and (4.6) respectively and ASN(p1) is the ASN
at AQL.
87
Step 1: Start with the tightened inspection level using the single sampling
variables plan with a sample size n1 and the acceptance criterion kS.
1 n1 S
(X i X1)2
X1
n1S
X i and S
i 1 n1S 1
.
Step 2: If t lots in a row are accepted under tightened inspection, then switch to
normal inspection.
Step 3: During the normal inspection, inspect the lots using the single sampling
variables plan with a sample size n2 (<n1) and the acceptance criterion kS.
1 n2 S
(X i X 2 )2
X2
n2S
X i and S
i 1 n2S 1
.
Thus, the unknown sigma variables TNT scheme has the parameters
namely the sample sizes n1S, n2S and the acceptable criterion kS. If n1S=n2S, then the
variables TNT scheme reduced to the variables single sampling plan with
88
Matlab software using the routine fmincon. By solving the nonlinear problem
mentioned above, the parameters (n, m and k) for known sigma plan are
determined and tabulated in Table 4.1. The parameters (nS, m and kS) for unknown
sigma plan are also determined by using the approximation given in (4.9) and
(4.10) and also presented in Table 4.1.
4.7 EXAMPLES
4.7.1 SELECTION OF KNOWN SIGMA TNT SCHEME INDEXED BY
AQL AND LQL
For the above example, the operation of the variables TNT scheme is as follows.
Step 1: Start with tightened inspection. Take a random sample of size 25 and
U X , where 1 25
Compute v
X X i . Accept the lot if v 1.598 and
25 i 1
reject the lot if v 1.598 . If t = 5 consecutive lots are accepted with the
same sample size and acceptance criterion, then switch to normal
inspection as in step 2.
90
Step 2: During normal inspection, select a random sample of size 5 and calculate
U X , where X 1 5
v
5
X
i 1
i . Accept the lot if v 1.598 and reject the
Table 4.1 can also be used for the selection of the parameters of the
unknown variables TNT scheme for given values of AQL and LQL. Suppose
that AQL=1%, LQL=7%, = 5% and =10%. From Table 4.1, the parameters of
the unknown sigma variables TNT scheme can be determined as nS = 14, kS =
1.694 and m = 8. The normal plan sample size of unknown sigma variables TNT
scheme is nS = 14 and the tightened plan sample size is obtained as m x nS =
8x14=112. The acceptance criterion for both normal and tightened plans is
kS=1.694.
For the same AQL and LQL, we can determine the parameters of the variables
single, double sampling plans (from Sommers (1981)) and variables TNT scheme
(from Table 4.1) respectively as follows.
4.8 COMPARISONS
4.8.1 COMPARISON THORUGH OC CURVES
figure, it can be easily observed that, for good quality, i.e. for smaller values of
fraction nonconforming, the composite OC curve (OC curve of the variables TNT
scheme) coincides with the OC curve of the variables single sampling plan (12,
1.857). As quality deteriorates the OC curve of the composite OC curve moves
toward that for the single sampling plan (63, 1.857) and comes close to it beyond
the indifference quality level.
These tables apparently show that the variables TNT scheme provided
in this chapter will have minimum ASN when compared to the variables single,
double sampling plans and variables TNT scheme provided by Senthilkumar and
Muthuraj (2010) for both known and unknown sigma cases. Similar reduction in
ASN can be achieved for any combination of AQL and LQL values. This implies
that variables TNT scheme developed in this chapter will give desired protection
with minimum inspection so that the cost of inspection will greatly be reduced.
93
Thus the variables TNT scheme provides better protection than the variables single
sampling plans and variables double sampling plans. It is also to be noted that
Senthilkumar and Muthuraj (2010) have given several tables for finding the
parameters of variables TNT scheme for each value of m. But table developed in
this chapter will overcome this shortcoming and only one table given in this
chapter is enough for any combinations of AQL and LQL.
4.9 CONCLUSIONS
In this chapter, we have considered the designing of variables TNT sampling
scheme involving minimum average sample number for both known and unknown
standard deviation cases. In general, variables sampling plans require a smaller
sample size than do attributes sampling plans. This is also valid for the proposed
TNT sampling scheme. It has also been shown that the variables TNT scheme
provided in this chapter has smaller ASN than the ASN of the existing variables
single and double sampling plans. The variables TNT sampling scheme proposed
in this chapter also ensure the protection for the consumers in their point of view.
This variables TNT scheme will be effective and useful for compliance testing.
Further, tables provided in this chapter are compact and easy to apply for the
selection of parameters of variables TNT scheme for specified combinations of
AQL and LQL along with the producers and consumers risks.
94
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n m k ASN nS m kS ASNS
0.001 0.002 97 7.75 2.925 123.244 512 7.75 2.926 650.977
0.003 36 9.0 2.819 47.538 179 9.0 2.704 236.957
0.004 22 9.25 2.742 29.121 105 9.25 2.749 139.123
0.005 16 9.0 2.683 21.100 74 9.0 2.692 97.566
0.006 13 8.0 2.638 16.605 58 8.0 2.650 74.894
0.007 12 5.25 2.619 14.008 53 5.25 2.632 62.515
0.008 9 9.5 2.548 12.071 38 9.5 2.565 51.793
0.009 8 9.25 2.515 10.647 33 9.25 2.535 44.852
0.010 8 6.0 2.512 9.558 33 6.0 2.532 40.190
0.005 0.007 294 10.0 2.481 400.85 1199 10.0 2.482 1635.09
0.008 153 8.75 2.444 200.36 610 8.75 2.445 797.959
0.009 94 9.75 2.408 127.01 367 9.75 2.410 495.754
0.01 67 10.0 2.376 90.486 256 10.0 2.378 346.408
0.012 42 8.5 2.325 54.707 156 8.5 2.329 203.051
0.015 27 7.25 2.262 33.683 96 7.25 2.268 120.313
0.020 16 7.75 2.169 20.327 54 7.75 2.179 68.599
0.03 9 8.0 2.032 11.487 28 8.0 2.051 35.649
0.02 0.03 140 9.5 1.916 187.378 397 9.5 1.917 531.761
0.04 46 9.25 1.813 60.920 122 9.25 1.817 161.455
0.05 26 8.0 1.734 33.213 65 8.0 1.741 83.536
0.06 18 7.0 1.669 22.251 43 7.0 1.679 53.613
0.07 13 8.0 1.602 16.626 30 8.0 1.616 38.332
0.08 10 9.5 1.537 13.324 22 9.5 1.555 29.444
95
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n m k ASN nS m kS ASNS
0.02 0.09 9 6.25 1.512 10.906 19 6.25 1.533 23.726
0.1 7 9.5 1.439 9.390 14 9.5 1.467 19.496
0.11 7 5.25 1.438 8.184 14 5.25 1.466 16.981
0.12 6 6.0 1.389 7.197 12 6.0 1.421 14.467
0.03 0.04 252 8.75 1.778 328.964 650 8.75 1.779 849.342
0.05 75 9.5 1.693 100.604 182 9.5 1.695 245.182
0.06 39 10.0 1.620 52.979 90 10.0 1.625 122.889
0.07 26 8.75 1.561 33.970 58 8.75 1.568 75.723
0.08 19 8.5 1.506 24.577 41 8.5 1.515 52.798
0.09 15 7.75 1.460 19.018 31 7.75 1.472 39.630
0.10 12 8.5 1.409 15.515 24 8.5 1.424 31.255
0.11 10 8.5 1.366 12.996 33 8.5 1.385 25.464
0.12 9 7.0 1.337 11.131 17 7.0 1.358 21.398
0.15 6 8.5 1.216 7.794 10 8.5 1.251 13.891
0.04 0.05 370 10.0 1.666 502.286 883 10.0 1.666 1199.73
0.06 110 9.25 1.595 145.605 250 9.25 1.597 331.187
0.07 56 9.0 1.533 73.817 122 9.0 1.536 160.914
0.08 38 6.5 1.487 46.417 80 6.5 1.492 98.06
0.09 25 9.5 1.424 33.323 50 9.5 1.431 67.455
0.1 19 9.75 1.376 25.517 37 9.75 1.386 50.011
0.11 16 7.5 1.344 20.170 30 7.5 1.356 38.704
0.12 13 8.25 1.299 16.755 24 8.25 1.313 31.203
0.13 11 8.5 1.259 14.262 20 8.5 1.276 25.868
0.14 10 7.0 1.234 12.347 18 7.0 1.252 22.03
0.15 9 6.25 1.208 10.886 16 6.25 1.228 19.101
0.05 0.06 511 9.75 1.573 691.102 1143 9.75 1.573 1546.47
0.07 148 9.0 1.511 195.102 317 9.0 1.512 418.172
0.08 72 10.0 1.453 97.839 148 10.0 1.455 201.468
0.09 46 9.0 1.404 60.393 91 9.0 1.408 120.248
0.10 33 7.75 1.362 41.982 64 7.75 1.367 81.231
0.11 27 5.5 1.332 31.890 51 5.5 1.339 60.464
0.12 19 9.25 1.272 25.316 34 9.25 1.282 46.109
0.13 16 8.25 1.238 20.639 28 8.25 1.250 36.751
96
MinASN(p1) MinASN(p1)
Known Sigma Unknown Sigma
p1 p2 n m k ASN nS m kS ASNS
0.05 0.14 13 10.0 1.193 17.650 22 10.0 1.207 30.514
0.15 12 7.25 1.174 14.961 20 7.25 1.190 25.548
0.06 0.08 184 10.0 1.435 250.769 373 10.0 1.436 509.305
0.09 91 9.75 1.384 122.634 178 9.75 1.386 240.413
0.10 61 6.5 1.346 74.263 116 6.5 1.349 141.827
0.11 41 7.25 1.301 51.333 76 7.25 1.305 95.065
0.12 30 8.0 1.258 38.452 54 8.0 1.264 69.167
0.13 24 7.5 1.222 30.177 42 7.5 1.229 52.985
0.14 19 8.75 1.180 24.763 32 8.75 1.896 42.284
0.15 16 8.25 1.148 20.645 27 8.25 1.159 34.514
0.07 0.09 229 9.75 1.368 308.023 443 9.75 1.369 596.562
0.10 112 9.0 1.322 147.810 210 9.0 1.324 277.280
0.11 75 6.0 1.287 89.608 137 6.0 1.289 164.092
0.12 46 10.0 1.235 62.231 81 10.0 1.239 109.986
0.13 34 9.75 1.197 45.974 58 9.75 1.202 79.200
0.14 29 6.5 1.174 35.430 49 6.5 1.180 60.101
0.15 22 8.75 1.129 28.840 36 8.75 1.137 47.484
0.20 11 7.5 0.983 13.792 16 7.5 0.9996 20.682
97
0.9
0.7
P r o b a b i li ty o f A c c e p ta n c e , P a (p )
0.5
0.4
Variables Tightened SSP
0.3
0.2
0.1
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Fraction Nonconforming, p
Figure 4.1: OC Curves of Single Sampling Normal Plan (12, 1.857), TNT
Scheme (63, 12; 1.857) and Single Sampling Tightened
Plan (63, 1.857)
98
Table 4.2: ASN Values of the Known Sigma Variables Single Sampling Plan,
Variables Double Sampling Plan and Variables TNT Scheme
ASN
Variables Variables Variables* Variables **
p1 p2 SSP DSP TNT TNT
0.001 0.002 191 154.9 179.792 123.244
0.001 0.003 74 59.4 68.673 47.538
0.005 0.010 138 112.0 126.310 90.486
0.005 0.012 85 69.5 76.809 54.707
0.03 0.04 506 434.6 471.546 328.964
0.03 0.06 81 127.7 72.746 52.979
0.04 0.07 114 180.6 108.116 73.817
0.04 0.08 72 58.4 68.675 46.417
0.05 0.07 300 246.7 271.860 195.102
0.05 0.08 149 122.3 137.421 97.839
* ASN given in Senthilkumar and Muthuraj (2010)
** ASN given in Table 4.1
Table 4.3: ASN Values of the Unknown Sigma Variables Single Sampling
Plan, Variables Double Sampling Plan and Variables TNT Scheme
ASN
Variables Variables Variables* Variables **
p1 p2 SSP DSP TNT TNT
0.001 0.002 1032 829.1 971.031 650.977
0.001 0.003 381 302.4 355.665 236.957
0.005 0.010 547 437.1 500.564 346.408
0.005 0.012 327 263.0 296.943 203.051
0.03 0.04 1333 1138.7 1238.397 849.342
0.03 0.06 197 156.5 176.451 122.889
0.04 0.07 258 205.0 245.252 160.914
0.04 0.08 159 125.8 151.353 98.06
0.05 0.07 660 535.4 598.751 418.172
0.05 0.08 319 258.0 294.171 201.468
* ASN given in Senthilkumar and Muthuraj (2010)
** ASN given in Table 4.1
99
Table 4.4: Parameters of Known Sigma Variables TNT Scheme for different
AQL and LQL Values
*Parameters of **Parameters of
p1 p2 Variables TNT Variables TNT
n m k ASN n m k ASN
0.001 0.002 146 2.0 2.955 151.596 97 7.75 2.925 123.244
0.005 0.007 451 2.0 2.503 469.248 294 10.0 2.481 400.85
0.01 0.020 83 2.0 2.150 86.288 56 10.0 2.108 75.757
0.02 0.03 224 2.0 1.941 233.170 140 9.5 1.916 187.378
0.03 0.04 387 2.0 1.796 401.699 252 8.75 1.778 328.964
0.04 0.06 163 2.0 1.621 169.272 110 9.25 1.595 145.605
0.05 0.07 224 2.0 1.541 233.170 148 9.0 1.511 195.102
0.06 0.08 328 2.0 1.460 341.510 184 10.0 1.435 250.769
0.07 0.10 163 2.0 1.351 169.301 112 9.0 1.322 147.810
*Parameters of **Parameters of
p1 p2 Variables TNT Variables TNT
nS m kS ASN nS m kS ASN
0.001 0.002 728 2.0 2.951 771.730 97 7.75 2.925 123.244
0.005 0.007 265 2.0 2.503 469.248 294 10.0 2.481 400.85
0.01 0.020 83 2.0 2.150 86.288 56 10.0 2.108 75.757
0.02 0.03 224 2.0 1.941 233.170 140 9.5 1.916 187.378
0.03 0.04 387 2.0 1.796 401.699 252 8.75 1.778 328.964
0.04 0.06 163 2.0 1.621 169.272 110 9.25 1.595 145.605
0.05 0.07 224 2.0 1.541 233.170 148 9.0 1.511 195.102
0.06 0.08 328 2.0 1.460 341.510 184 10.0 1.435 250.769
0.07 0.10 163 2.0 1.351 169.301 112 9.0 1.322 147.810
CHAPTER 5
OPTIMAL DESIGNING OF
VARIABLES CHAIN SAMPLING PLAN BY MINIMIZING THE
AVERAGE SAMPLE NUMBER
5.1 INTRODUCTION
In addition, the usual conditions for the application of single sampling variables
plans with known or unknown standard deviation should also be valid.
Step 1: From each submitted lot, take a random sample of size n , say
and compute v U X ,
n
1
X 1 , X 2 ... X n
where X X i .
n i 1
Step 2: Accept the lot if v k and reject the lot if v k ' . If k ' v k , then
accept the current lot provided that the preceding i lots were accepted on
the condition that v k but reject, otherwise. (Note: k k ' ).
102
Step 1: From each lot, take a random sample of size n , say X 1 , X 2 ... X n
and
X L , 1 n
compute v
where X
n
X
i 1
i .
Step 2: Accept the lot if v k and reject the lot if v k ' . If k ' v k , then
accept the current lot provided that the preceding i lots were accepted on
the condition that v k but reject, otherwise.
Step 1: From each submitted lot, take a random sample of size n S , say
1 2 nS i and
S i 1
(X i X )2
S .
nS 1
Step 2: Accept the lot if v k S and reject the lot if V k ' S . If k ' S v k S , then
accept the current lot provided that the preceding i S lots were accepted
on the condition that v k S .
Thus, the proposed unknown sigma variables chain sampling plan is
characterized by four parameters namely nS , i S , k S and k' S . If k S k ' S , then the
proposed plan reduced to the variables single sampling plan with unknown
standard deviation.
approach and economic based approach. For further detail, readers may refer
Chen and Lam (1999), Ferrell and Chhoker (2002), Chen (2005), Chen et al.
(2007), Balamurali and Subramani (2010), Vijayaraghavan and Sakthivel (2011),
Balamurali et al. (2012) and Fallahnezhad and Aslam (2013). In this chapter, we
have followed the designing methodology based on two points on the OC curve
approach. The variables chain sampling plan is designed based on the two points
on the OC curve in the following manner.
Then the OC function of the variables chain sampling plan, which gives the
proportion of lots that are expected to be accepted for given product quality, p is
given by (see Govindaraju and Balamurali (1998)
Pa ( p) Pr(v k ) Pr(v k ' ) Pr(v k )Pr(v k )
i
(5.4)
105
where Prv k is the probability of accepting a lot based on a single sample with
parameters (n, k ) and Prv k ' is the probability of rejecting a lot based on a
single sample with parameters (n, k ' ) . Under type B situation (i.e. a series of lots
of the same quality), forming lots of N items from a process and then drawing
random sample of size n from these lots is equivalent to drawing random samples
of size n directly from the process. Hence the derivation of the OC function is
straightforward.
The probability of acceptance of the chain sampling plan can also be written as
Pa ( p ) ( w2 ) ( w1 ) (w2 ) (w2 )
i
(5.5)
The OC function given in (5.5) under the specified AQL and LQL conditions can
be written as
The parameters of the known sigma variables chain sampling plan are
denoted by ( n , k ' , k ).The following optimization problem is considered to
determine the optimal parameters of known sigma variables sampling plan such
as n , i , k and k ' .
Minimize ASN ( p1 ) n
Subject to Pa ( p1 ) 1
Pa ( p 2 )
n 2, i 1, k k ' 0, 0, n N, i N (5.6)
2
and variance k S Var (S ) (see Duncan (1986), Balamurali and Jun (2007)).
n
That is,
2 2
2
X k S S ~ N k S , kS
n 2n
U k S
2
kS
( / n S ) 1
2
nS
( Z p k S ) (5.7)
k
2
1 S
2
nS
If we let, w2 S (Z p k S ) then Pr(v k ) ( w ) . (5.8)
k
2 S 2S
1 S
2
nS
Similarly if we let, w1S ( Z p k ' S ) 2
then we have
k'S
1
2
Pr(v k ' S ) 1 ( w1S ) (5.9)
Hence the lot acceptance probability for sigma unknown case is given by
Pa ( p ) ( w2 S ) ( w1S ) ( w2 S ) (w2 S ) S
i
(5.10)
If (AQL, 1 ) and (LQL, ) are prescribed then the OC function can be written
as
(w21S ) ( w11S ) ( w21S ) ( w21S ) S 1
i
nS nS
w11S (Z p1 k ' S ) , w (Z k )
k'
2 21S p1 S
k
2
1 S 1 S
2 2
nS nS
w12 S (Z p 2 k 'S ) and w ( Z k ) (5.12)
k 'S
2 22 S p2 S
kS
2
1 1
2 2
where Z p 1 is the value of Z p at AQL and Z p 2 is the value of Z p at LQL.
For given AQL and LQL, the parametric values of the unknown sigma
variables chain sampling plan namely nS , i S , k S and k' S are determined by satisfying
the required producer and consumer conditions. Alternatively, we can determine
the above parameters of the variables chain sampling plan to minimize the ASN at
AQL, which analogous to minimizing the average sample number in the variables
repetitive group sampling plans and multiple dependent state sampling plan (see
Balamurali et al. (2005), Balamurali and Jun (2007)). Some of the authors have
investigated the designing of sampling plans by using some other optimization
techniques which are available in the literature (see for example, Feldmann and
Krumbholz (2002), Krumbholz and Rohr (2006,2009), Krumbholz et al. (2012),
Duarte and Sariava (2010, 2013)). The ASN for the chain sampling plan is the
sample size only.
109
nS 2, i S 1, k S k ' S 0, 0, n S N, i S N (5.13)
sigma plan are determined and these values are tabulated in Table 5.1.
From each submitted lot, take a random sample of size 18 and compute
U X , 1 18
v
where X X i . Accept the lot if v 1.779 and reject the lot if
18 i 1
v 1.544 . If 1.544 v 1.779 , then accept the current lot provided that the
preceding 3 lots were accepted on the condition that v 1.779 with the sample
size of 18.
can also be used for the selection of the parameters of the unknown variables
chain sampling plan for given values of AQL and LQL. Suppose that
AQL=0.0075, LQL=0.035, =5% and =10%. From Table 5.1, the parameters
of the variables chain sampling plan can be determined as n S = 47, i S = 1, k' S =
1.925 and k S = 2.190.
limit is given as 0.77 mm. Here we consider only 47 values randomly taken from
the original data given in Wu and Pearn (2008). The data are shown below.
0.717 0.698 0.726 0.684 0.727 0.688 0.708 0.703 0.694 0.713
0.730 0.699 0.710 0.688 0.665 0.704 0.725 0.729 0.716 0.685
0.712 0.716 0.712 0.733 0.709 0.703 0.730 0.716 0.688 0.688
0.712 0.702 0.726 0.669 0.718 0.714 0.726 0.683 0.713 0.737
Step 1: Take a random sample of size 47. The data are given above.
1 nS
Step 2: For this data, we calculate X X i 0.708915 and
n i 1
(X i X )2
S 0.017583 .
n 1
(U X ) (0.77 0.708915)
Step3: Calculate v 3.4741
S 0.017583
Just for sake of discussion, let us assume that the v value (for different
data set) is calculated as 2.15. In this case, we can accept the current lot provided
previous lot must have been accepted with the condition that v k ' S 1.925 .
Otherwise the current lot is rejected.
112
For the same values of the AQL and LQL, we can determine the parameters of the
single and double sampling variables plan (from Sommers (1981)) as
Figure 5.1 shows the OC curves of the variables chain sampling plans
with parameters n = 18, k ' = 1.544 and k = 1.779 for different values of i . This
figure apparently shows that the variables chain sampling plan increases the
probability of acceptance in the region of principal interest, i.e. for good quality
levels and maintains the consumers risk at poor quality levels. This is also an
important feature of the variables chain sampling plan.
Schilling (2005) has pointed out that average run length (ARL) is a
missing and meaningful measure for characterizing and evaluating the sampling
plans under Type B situations as in the process control procedures. The ARL
gives an indication of the expected number of samples until a decision is made.
The ARL can be easily calculated once the probability of acceptance (Pa(p)) of
the plan is known for any process fraction nonconforming, p. It is clear that the
distribution of the run length, L follows the geometric distribution with
probability mass function
f G ( L) Pa ( p) 1 Pa ( p)
L 1
(5.17)
1
ARL E ( L) (5.18)
1 Pa ( p)
Pa ( p)
Var ( L) (5.19)
1 Pa ( p) 2
Table 5.5 gives the values of ARL of the chain sampling plan with
n = 16, k ' = 1.501 and k = 1.841 for different values of i . This table
apparently shows that when the process fraction nonconforming is small, the ARL
is high and for the increased values of fraction nonconforming the ARL is low.
By comparing the ARL values for different i values, when i increases, the ARL
values reduce even for the lower fraction nonconforming. Also it is clear from the
table that 95% of the lots will be accepted at fraction nonconforming 2% by the
variables chain sampling plan ( i =1) at an average rate of 20 inspections where as
115
5.9 CONCLUSIONS
Table 5.3. Variables Single Sampling Plans Indexed by AQL and LQL
for =1% and =1%
Table 5.4: ASN Values of the Variables SSP, DSP and Variables Chain
Sampling Plans
ASN
Known Sigma Unknown Sigma
p1 p2 SSP DSP ChSP SSP DSP ChSP
0.001 0.003 74 59.4 46 381 302.4 242
0.0025 0.0075 62 50.1 39 267 214.2 171
0.005 0.015 53 43.0 33 196 157.6 128
0.0075 0.025 39 31.3 29 129 102.1 83
0.01 0.05 19 14.9 13 54 41.5 39
0.02 0.08 21 16.9 14 50 39.6 36
0.03 0.09 30 24.0 23 66 52.5 44
0.04 0.10 39 31.3 27 82 65.1 53
0.05 0.12 39 32.0 26 76 60.9 49
0.9
i =1
0.8 i =2
0.7
Probability of Acceptance, Pa(p)
i =3
0.6
i =4
0.5 i =
0.4
0.3
0.2
0.1
0
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Fraction Non-conforming, p
CHAPTER 6
OPTIMAL DESIGNING OF
VARIABLES SAMPLING PLAN FOR RESUBMITTED LOTS
6.1 INTRODUCTION
(i) It is required to discard the results of the original inspection that resulted in
non-acceptance of the lot. This may be by the provisions of a Contract or
Statute.
(ii) The consumer has confidence in the producer and producer will not
deliberately take advantage of the resampling option.
In addition, the usual conditions for the application of variables single sampling
plans with known or unknown standard deviation should also be valid.
123
Step 1: Perform original inspection by using variables single sampling plan as the
reference plan, i.e From each submitted lot, take a random sample of size
and compute v U X ,
n
1
n, say X 1 , X 2 ...X n
where X X i .
n i 1
Step 1: Perform original inspection by using variables single sampling plan as the
reference plan, i.e From each submitted lot, take a random sample of size
and compute v X L ,
n
1
n, say X 1 , X 2 ...X n
where X X i .
n i 1
U
p 1 (6.1)
where ( y ) is given by
y
1 z2
( y) exp dz (6.2)
2 2
z p 1 (1 p ) (6.3)
L( p ) Pa ( p) 1 Pa ( p)Pa ( p) 1 Pa ( p ) Pa ( p ) .... 1 Pa ( p)
2 m 1
Pa ( p)
1 1 Pa ( p )
m
(6.4)
single sampling plan with parameters (n, k). Under type B situation (i.e. a series
of lots of the same quality), forming lots of N items from a process and then
drawing random sample of size n from these lots is equivalent to drawing
random samples of size n directly from the process. Hence the derivation of the
OC function is straightforward.
ASN ( p ) n 1 Pa ( p )n 1 Pa ( p ) n .... 1 Pa ( p )
2 m 1
n
n 1 1 Pa ( p)
m
P ( p)
a (6.5)
126
L( p ) Pa ( p)2 Pa ( p ) (6.6)
ASN ( p ) n2 Pa ( p) (6.7)
The above measures of the resampling scheme under variables inspection can also
be written as
L( p ) ( w)2 ( w) (6.8)
ASN ( p ) n2 ( w) (6.9)
where w z p k n
(w1 )2 (w1 ) 1
Here w1 is the value of w at p=p1 and w2 is the value of w at p=p2. That is,
w1 ( z p 1 k ) n and w2 ( z p 2 k ) n (6.11)
For given AQL and LQL, the parametric values of the variables
resampling scheme namely n, k, and m are determined by satisfying the required
127
L( p 2 )
n 1, m 1, k 0 (6.12)
2 (w1 ) 2 ( w2 )
and Minimize ASN ( p1 ) ASN ( p 2 ) n
(w1 ) ( w2 )
Subject to L( p1 ) 1
L( p 2 )
n 1, m 1, k 0 (6.13)
Here L(p1) and L(p2) are the probability of acceptance of the lot at AQL and LQL
respectively based on resampling scheme.
Whenever the standard deviation is unknown, we may use the sample standard
deviation S instead of . In this case, the proposed scheme operates as follows.
Step 1: Perform original inspection by using variables single sampling plan as the
reference plan, i.e From each submitted lot, take a random sample of size
128
and compute v U X ,
nS
1
nS, say X 1 , X 2 ...X n where X X i and
S
S nS i 1
(X i X )2
S .
nS 1
2
k S Var ( S ) (see Duncan (1986)). That is,
n
2 2
2
X k S S ~ N k S , kS
n 2n
PX k S U p
S
129
U kS
2
( / n ) 1 k S
S
2
nS
( z p k S ) 2
(6.13)
k
1 S
2
nS
If we let, wS ( z p k S ) 2
then Pr(v k ) (w ) .
S S (6.14)
kS
1
2
Hence the lot acceptance probability of the resampling scheme under sigma
unknown case is given by
L( p ) ( wS )2 ( wS )
nS nS
w1S ( z p 1 k S ) 2
and w2 S ( z p 2 k S ) 2
(6.15)
kS kS
1 1
2 2
nS 1, m 1, k S 0 (6.16)
and
2 ( w1 ) 2 ( w2 )
Minimize ASN ( p1 ) ASN ( p 2 ) n S
( w1 ) ( w2 )
130
Subject to L( p1 ) 1
L( p 2 )
nS 1, m 1, k S 0 (6.17)
Here L(p1) and L(p2) are the probability of acceptance of the lot at AQL and LQL
respectively.
We have observed an interesting thing from Table 6.1 and Table 6.2 is
that we are getting almost same values of the parameters of the proposed variables
resampling scheme while minimizing ASN(p2) as well as minimizing the sum of
ASN at both AQL and LQL.
131
Step 1: From each submitted lot, take a random sample of size 35 and compute
X L , 1 35
v
where X Xi .
35 i 1
Step 2: Accept the lot if v 1.924 and reject the lot if v 1.924 . On non-
acceptance on the original inspection, apply the variables single sampling plan for
2 times and reject the lot if it is not accepted on the 1st inspection.
of the unknown variables resampling scheme for given values of AQL and
LQL. Suppose that AQL=1%, LQL=3%, =5% and =10%. From Table 6.1, the
parameters of the variables resampling scheme can be determined as nS = 99, kS =
2.183 and m = 2. The ASN for this plan is 192.950.
For the same values of the AQL and LQL, we can determine the parameters of the
single sampling variables plan from (Sommers (1981)) as
higher fraction non-conforming. This is logical since lots that are declared as not
acceptable are always resampled even though the original inspection showed the
evidence of poor quality. The main advantage and strength of the resampling
scheme lies in achieving smaller ASN at good quality that is, low fraction non-
conforming in which case the usual variables SSP requires a larger sample size or
ASN. This can easily be observed from Figure 6.1 and Figure 6.2 show the OC
curves of the variables resampling scheme with parameters n = 10 and k = 2.085
for different values of m such as m=2, m=3, m=4 and m=5. This figure shows that
when m increases the AQL values also increase.
reduced. Thus the variables resampling scheme provides better protection to the
producers than the variables single sampling plan.
6.10 CONCLUSIONS
MinASN(p2)
Known Sigma Unknown Sigma
p1 p2 n k ASN(p2) nS kS ASN(p2)
0.001 0.0020 129 3.023 251.552 713 3.023 1389.883
0.004 30 2.951 58.477 160 2.951 311.820
0.006 18 2.902 35.118 90 2.906 175.410
0.008 13 2.862 25.335 64 2.873 124.766
Table 6.3: ASN Values of the Variables Single, Double Sampling Plans and
Resampling Schemes
ASN at AQL
Known Sigma
p1 p2 Single Double Resampling Scheme
0.001 0.004 45 36.8 36.69
0.0025 0.0075 62 50.1 50.15
0.005 0.015 53 43.0 42.81
0.0075 0.030 29 23.2 23.22
0.01 0.03 44 35.0 35.48
0.02 0.05 52 42.3 42.82
0.03 0.09 30 24.0 24.45
0.04 0.13 22 18.3 18.35
0.05 0.11 49 40.0 40.35
Table 6.4: Sample Size of the Attributes and Variables Resampling Schemes
Sample size
Attributes Resampling Variables Resampling
p1 p2 Scheme Scheme
0.001 0.01 470 10
0.001 0.02 148 6
0.001 0.03 98 4
0.0025 0.01 624 25
0.0025 0.02 234 11
0.0025 0.03 98 7
0.004 0.01 1177 231
0.004 0.02 311 65
0.004 0.03 156 38
0.0065 0.01 * 927
0.0065 0.02 249 117
0.0065 0.03 207 58
* sample size greater than 5000
138
Table 6.5: ASN of the Variables Resampling Schemes for Different m values
CHAPTER 7
attributes sampling scheme have been discussed. Tables have also been
constructed for the application of the proposed scheme.
normal distribution and has single specification limit. Resubmission of lots for
inspection is allowed in some situations where the original inspection results are
suspected or when the supplier or producer is allowed to opt for resampling as per
the provisions of the contract etc. We have considered both known and unknown
sigma cases. Non-linear optimization problem has been considered for the
selection optimal parameters. The advantages of the proposed variables sampling
plan over the existing variables single sampling plan have also been discussed.
Useful tables have also been constructed for the selection of optimal parameters of
known and unknown standard deviation cases of the proposed variables sampling
plan.
In this thesis, we have developed five different sampling
systems/plans for variables inspection. The conditions of application of each
sampling procedure have been given in the respective chapter. All the sampling
plans provided in this thesis can be applied for the inspection of normally
distributed quality characteristics. However, the following may be an additional
guideline for applying appropriate sampling system.
(i) If the quality characteristic of interest has single specification limit and if we
want to use same sample size in both normal and tightened inspections but
with two different acceptance criteria, then the VQSS presented in Chapter 2
can be used.
(ii) If the quality characteristic of interest has single specification limit but we
want to use different sample sizes in normal and tightened inspections and
same acceptance criterion, then the TNT scheme proposed in Chapter 4 can be
applied.
(iii) If the quality characteristic of interest has double specification limits, then the
VQSS developed in Chapter 3 can be implemented.
145
(iv) Whenever we want to take a decision on the current lot submitted for
inspection based on the history of the previous lot quality, one can utilize the
chain sampling plan proposed in Chapter 5.
(v) Whenever the original inspection results are suspected or when the producer
or supplier has provision of opting resampling, the variables resampling
scheme proposed in Chapter 6 can be opted for quality inspection.
REFERENCES
Aslam, M., Wu, C-W., Azam, M. and Jun, C-H. (2013),Variables Sampling
Inspection for Resubmitted Lots Based on Process Capability Index Cpk
for Normally Distributed Items, Applied Mathematical Modeling,
Vol.37, pp.667-675.
Baillie, D.H. (1992), Double Sampling Plans for Inspection by Variables when
the Process Standard Deviation is Unknown, International Journal of
Quality and Reliability Management, Vol.9, No.5, pp.59-70.
Balamurali, S and Jun, C-H.(2006),Repetitive Group Sampling Procedure for
Variables Inspection, Journal of Applied Statistics, Vol.33,
pp.327-338.
Balamurali, S. and Jun, C.-H. (2007), Multiple Dependent State Sampling Plans
for Lot Acceptance based on Measurement Data, European Journal of
Operational Research, Vol.180, pp.1221-1230.
Balamurali, S. and Jun, C.-H. (2009),Designing of a Variables Two-Plan System
by Minimizing the Average Sample Number, Journal of Applied
Statistics, Vol.36, No.10, pp.1159-1172.
Balamurali, S. and Kalyanasundaram, M. (1996),Procedures and Tables for the
Selection of Zero Acceptance Number Quick Switching Systems,
International Journal of Management and Systems, Vol.12, No.3,
pp.369-378.
Balamurali, S. and Kalyanasundaram, M. (1997),Determination of an
Attribute Single Sampling Scheme, Journal of Applied Statistics
Vol.24, No.6, pp.689-695.
Balamurali, S. and Kalyanasundaram, M. (1999),Determination of
Conditional Double Sampling Scheme, Journal of Applied Statistics,
Vol.26, No.8, pp.893-902.
148
Balamurali, S., Park, H., Jun, C.-H., Kim, K. J. and Lee, J. (2005),Designing of
Variables Repetitive Group Sampling Plan Involving Minimum
Average Sample Number, Communications in Statistics- Simulation
and Computation, Vol.34, pp.799-809.
Beattie, D. W. (1962), A Continuous Acceptance Sampling Procedure based
upon a Cumulative Sum Chart for the Number of Defectives, Applied
Statistics, Vol.11, No.3, pp.137-147.
Bender, A. J.(1975),Sampling by Variables to Control the Fraction Defective:
Part II, Journal of Quality Technology, Vol.7, pp.139-143.
Bravo, P. C. and Wetherill, G. B. (1980),The Matching of Sampling Plans and the
Design of Double Sampling Plans, Journal of the Royal Statistical
Society A, Vol.143, pp.49-67.
Bruhn-Suhr, M. and Krumbholz, W. (1990),A New Variable Sampling Plan for
Normally Distributed Lots with Unknown Standard Deviation and
Double Specification Limits, Statistical Papers, Vol.31, pp.195-207.
Calvin, T.W.(1977):TNT Zero Acceptance Number Sampling, American Society
for Quality Control Annual Technical Conference Transactions,
Philadelphia, PA, pp.35-39.
Cameron, J. M. (1952),Tables for Constructing and for Computing the Operating
Characteristics of Single Sampling Plans, Industrial Quality Control,
Vol.9, No.1, pp.37-39.
Chen, C.H. (2005),Economic Design of Dodge-Romig AOQL Single Sampling
Plans by Variables with the Quadratic Loss Function, Tamkang
Journal of Science and Engineering, Vol.8, No.4, pp.313-318.
Chen, J. and Lam, Y. (1999), Bayesian Variable Sampling Plan for Weibull
Distribution with Type I Censoring, Mathematicae Applicatae Sinica,
Vol.15, No.3,pp.269-280.
150
Chen, J., Li, K., and Lam, Y., (2007), Bayesian Single and Double Variable
Sampling Plans for the Weibull Distribution with Censoring,
European Journal of Operational Research, Vol. 177, No. 2,
pp.10621073.
Collani, E. V. (1990),A Note on Acceptance Sampling for Variables, Metrika,
Vol.38, pp.19-36.
Das, T. K. and Mitra, S. K. (1964),The Effect of Non-normality on Sampling
Inspection, Sankhya, Vol.26A, pp.169176.
Das, N. G., Gosavi, A. and Kanghibhatta, K. M. (2002),Optimal Design of Plans
for Acceptance Sampling by Variables with Inverse Gaussian
Distribution, Communications in Statistics-Simulation and
Computation, Vol.31,No.3, pp.463 488.
Dodge, H. F. (1955),Chain Sampling Inspection Plan, Industrial Quality
Control, Vol.11, No.4, pp.10-13.
Dodge, H. F. (1965),Evaluation of a Sampling System Having Rules for
Switching Between Normal and Tightened Inspection, Technical
Report No.14, Statistics Center, Rutgers University, Piscataway, NJ.
Dodge, H. F. (1967),A Dual System of Acceptance Sampling, Technical Report
No.16, The Statistics Center, Rutgers- The State University, New
Brunswick, New Jersey.
Dodge, H.F. (1969),Notes on the Evolution of Acceptance Sampling Plans
Part I, Journal of Quality Technology, Vol.1, No.2, pp.77-88.
Dodge, H.F., and Romig, H.G. (1959), Sampling Inspection Tables- Single and
Double Sampling , 2nd Edition, John Wiley and Sons , New York, NY.
Dodge, H.F., and Stephens K.S. (1966), Some New Chain Sampling Inspection
Plans , Industrial Quality Control, Vol.23 ,No.2, pp. 61-67
151
LIST OF PUBLICATIONS
CURRICULUM VITAE