PH.D Thesis M. Usha

STUDIES ON OPTIMAL DESIGNING OF CERTAIN
SPECIAL PURPOSE SAMPLING PLANS FOR

VARIABLES INSPECTION
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
Acceptance sampling is one of the major areas of statistical

quality control (also familiarly known as Industrial Statistics). Acceptance
sampling is the methodology that deals with procedures by which decision to
accept or reject the lot based on the results of the inspection of samples.
Acceptance sampling prescribes a procedure that, if applied to a series of lots, will
give a specified risk of accepting lots of given quality. In other words, acceptance
sampling yields quality assurance. Implementation of acceptance sampling in
industries through the operation of sampling plan yields quality assurance. Use of
acceptance sampling is essential to secure ISO certification which gives a passport
for larger exports.
In general, the acceptance sampling plans are classified in to

attribute sampling plans and variables sampling plans. Many quality characteristics
cannot be conveniently represented numerically. In such cases, we usually classify
each item inspected as either conforming (non-defective) to the specification on
that quality characteristics or non-conforming (defective) to those specifications.
Quality characteristics of this type are called attributes. Sampling plans applied to
such quality characteristics are called attributes sampling plans. Several sampling
plans are available in the literature for the application of attributes quality
characteristics. For example, single sampling plan, double sampling plan, multiple
sampling plan etc.
Variables sampling plans specify the number of items to be

sampled and the criterion for sentencing lots when measurements data are
collected on the quality characteristic of interest. These plans are generally based
vii
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.
The special purpose sampling plans such as Chain Sampling Plan

(ChSP) of Dodge (1955), Quick Switching System (QSS) of Romboski (1969),
Tightened-Normal-Tightened (TNT) sampling scheme developed by Calvin (1977)
etc., so far developed only for application of attributes quality characteristics. But
for the inspection of measurable characteristics, no such special purpose plan is
available in the literature except chain sampling plan which was developed by
viii
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.
Chapter 1 of this thesis comprises of sections that consists of basic concepts

of quality control, acceptance sampling, reliability, lifetime distributions and the
review of sampling plans which are relevant to this thesis. In Chapter 2, the
optimal designing of variables quick switching system is proposed in which the
quality characteristic under study follows a normal distribution with known and
unknown standard deviations. The minimum sample size n is determined for the
predefined acceptance criteria k N and k T and are used to calculate the
probability of acceptance for different combinations of the consumers confidence
levels and the producers confidence levels. The results are presented in tables and
explained with figures and examples.
Chapter 3 of the thesis investigates the variables quick switching sampling

system when a measurable quality characteristic has double specification limits
beyond which an item is considered to be a non-conforming. The quality
characteristic of interest is assumed to follow the normal distribution. The optimal
parameters of the variables quick switching system are determined for both known
and unknown standard deviations which satisfy the producers and consumers
risks at the corresponding specified quality levels. Symmetric and asymmetric
cases based on the fraction non-conforming by the lower and the upper
specification limits are also considered. The problem is formulated as a nonlinear
programming where the objective function to be minimized is the average sample
ix
number and the constraints are related to lot acceptance probabilities at acceptable
quality level and limiting quality level under the operating characteristic curve.
The Chapter 4 of the thesis deals with the optimal designing of

Tightened-Normal-Tightened sampling scheme with sample sizes n 1 , n 2 and the
acceptance criteria k . The advantages of the proposed variables scheme over
variables single, double sampling plans and attributes sampling scheme are
discussed. Tables are also constructed for the selection and application of
parameters of known and unknown standard deviation variables sampling schemes
for specified two points on the operating characteristic curve. The problem is
formulated as a nonlinear programming with minimizing the average sample
number as the objective function and the constraints are related to lot acceptance
probabilities at acceptable quality level and limiting quality level based on the
operating characteristic curve.
In Chapter 5 of the thesis, we investigate the optimal designing of chain

sampling plan for the application of normally distributed quality characteristics.
The advantages of this proposed variables chain sampling plan over variables
single sampling plan and variables double sampling plan are discussed. Tables are
also constructed for the selection of optimal parameters of known and unknown
standard deviation variables chain sampling plan for specified two points on the
operating characteristic curve, namely, the acceptable quality level and the limiting
quality level, along with the producers and consumers risks. A non-linear
optimization problem is formulated in which the average sample number is
minimized subject to the constraints of satisfying the producer and consumer risks
at their respective quality levels.
x
In Chapter 6, an optimal designing of variables sampling plan which can

be applied for sampling inspection of resubmitted lots when the quality
characteristic of interest follows the normal distribution is proposed. The
advantages of this proposed variables sampling plan over the existing single
sampling variables plan are discussed. Tables are also constructed for the selection
of optimal parameters of known and unknown standard deviation variables
resampling scheme for specified two points on the operating characteristic curve
namely the acceptable quality level and the limiting quality level along with the
producer and consumers risks. The optimization problem is formulated as a
nonlinear programming for finding the optimal parameters satisfying both
producer and consumer risks.
xi
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
like to express my deepest gratitude.
I am deeply indebted to Dr. S. BALAMURALI, Ph.D., Professor,
Department of Mathematics, Kalasalingam University, Krishnankoil, TamilNadu,
India for his valuable Supervision, continuous guidance, expert consultancy,
unstinted support and enormous encouragement to bring out this thesis. In his I
experience a personification of excellence, dedication and commitment. I also
appreciate and thank, for his free availability and approach. I pray to God for his
well being.
I wish to record my gratitude to the Management and the authorities of
Kalasalingam University, for granting me permission and providing necessary
facilities to carry out the research. I am extremely grateful to the Chairman and
members of the Doctoral Committee for their support.
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
and give her best suggestions.

xii
Most importantly, none of this would have been possible without the
love and patience of my family. I would like to express my heart-felt gratitude to
my family.
I would like to thank my spouse , Mr. T. Sermaraj, and my children
S. Divyamaki and S. Lakinyamaki for their constant love and support.
Last but not the least, I would like to thank my parents
Mr. G. Mahalingam and Mrs. M. Jeyasamvarthini , for giving birth to me at
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
me with their best wishes.
M. Usha
xiii
TABLE OF CONTENTS
CHAPTER TITLE PAGE

NO. NO.
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
2. OPTIMAL DESIGNING OF VARIABLES QUICK 34

SWITCHING SAMPLING SYSTEM (VQSS) BY MINIMIZING
THE AVERAGE SAMPLE NUMBER
2.1 Introduction 34
2.2 Conditions of Application 34
2.3 Operating Pocedure of Known Sigma Case 35
2.4 Operating Characteristic Function of VQSS 36
2.5 Designing of a Known Sigma VQSS 37
xiv
CHAPTER TITLE PAGE

NO. NO.
2.6 Average Sample Number 37

2.7 Optimal Designing of Known Sigma VQSS 38
2.8 Optimal Designing of Unknown Sigma VQSS 38
2.9 Examples 42
2.9.1 Selection of Known Sigma VQSS Indexed by 42
AQL and LQL
2.9.2 Selection of Unknown Sigma VQSS Indexed 43
by AQL and LQL
2.10 Advantages of the VQSS 43
2.11 Comparison 44
3. OPTIMAL DESIGNING OF VARIABLES QUICK SWITCHING 53

SYSTEM WITH DOUBLE SPECIFICATION LIMITS
3.1 Introduction 53
3.2 Conditions for Application of VQSS 53
3.3 Operating Procedure of a Known Sigma VQSS 54
3.4 OC Function of a Known Sigma VQSS with Double 55
Specification Limits
3.4.1 Known Sigma VQSS with Symmetric Fraction 56
Non- conforming
3.4.2 Known Sigma VQSS with Asymmetric 58
Fraction Non-conforming
3.5 Designing of Unknown Sigma VQSS having Double 59
Specification Limits
3.6 Determination of the Optimal Parameters of VQSS 61
xv
CHAPTER TITLE PAGE

NO. NO.
3.7 Illustrative Examples 63

3.7.1 Symmetric Fraction Non-conforming Case 63
3.7.2 Asymmetric Fraction Non-conforming Case 63
3.8 Industrial Application of the Proposed VQSS 65
3.9 Comparisons 68
3.10 Non-normality in VQSS 68
3.11 Conclusions 69
4. OPTIMAL DESIGNING OF VARIABLES TIGHTENED 81

NORMAL TIGHTENED (TNT) SAMPLING SCHEME BY
MINIMZING THE AVERAGE SAMPLE NUMBER
4.1 Introduction 81
4.3 Operating Procedure of Known Sigma Variables 82
TNT Scheme
4.4 OC Function of Known Sigma Variables TNT Scheme 83
4.5 Designing of a Known Sigma Variables TNT Scheme 85
4.6 Designing of Unknown Sigma Variables TNT Scheme 87
4.7 Examples 89
4.7.1 Selection of Known Sigma TNT Scheme Indexed 89
by AQL and LQL
4.7.2 Selection of Unknown Sigma Variables TNT 90

Scheme Indexed by AQL and LQL
xvi
CHAPTER TITLE PAGE

NO. NO.
4.7.3 Advantages of the Variables TNT Sampling 90
Scheme
4.8 Comparisons 91
4.8.1 Comparison Thorugh OC Curves 91
4.8.2 Comparison Through ASN 92
4.9 Conclusions 93
5. OPTIMAL DESIGNING OF VARIABLES CHAIN SAMPLING 100

PLAN BY MINIMZING THE AVERAGE SAMPLE NUMBER
5.1 Introduction 100
5.3 Operating Procedures of Variables ChSP 101
5.3.1 Known Sigma Case 101
5.3.2 Unknown Sigma Case 102
5.4 Designing Methodology of Variables ChSP 103
5.4.1 Known Sigma Case 103
5.4.2 Unknown Sigma Case 106
5.5 Designing Examples 109
5.5.1. Selection of Known Sigma Variables ChSP 109
for Specified AQL and LQL
5.5.2. Selection of Unknown Sigma Variables ChSP 110
for Specified AQL and LQL
5.6 Illustrative Example 110
5.7 Advantages of the Variables ChSP 112
xvii
CHAPTER TITLE PAGE

NO. NO.
5.8 Average Run Length of Variables ChSP 114

5.9 Conclusions 115
6. OPTIMAL DESIGNING OF VARIABLES SAMPLING 122

PLAN FOR RESUBMITTED LOTS
6.1 Introduction 122
6.3 Operating Procedure of the Known Sigma Variables 123
Resampling Scheme
6.4 Designing of Variables Resampling Scheme with 124
Known Standard Deviation
6.5 Operating Procedure of Variables Resampling 127
Scheme with Unknown Sigma
6.6 Designing of Variables Resampling Scheme with 128
Unknown Standard Deviation
6.7 Designing Examples 131
6.7.1. Selection of Known Sigma Variables 131
Resampling Scheme Indexed by AQL
and LQL
6.7.2. Selection of Unknown Sigma Variables 131
Resampling Scheme Indexed by AQL
and LQL
6.8 Merits of the Variables Resampling Scheme 132
6.9 Comparison with Attributes Resampling Scheme 134
6.10 Conclusions 134
xviii
CHAPTER TITLE PAGE

NO. NO.
7. CONCLUSIONS AND FUTURE WORK 141

REFERENCES 146
LIST OF PUBLICATIONS 159
CURRICULUM VITAE 161
xix
LIST OF TABLES
TABLE NO. TITLE PAGE NO.
2.1 Variables Quick Switching Sampling Systems 49

Indexed by AQL and LQL for =5% and =10%
Involving Minimum ASN
2.2 ASN Values of the Known Sigma Variables Single Sampling 54
Plan, Variables Double Sampling Plan and VQSS
2.3 ASN Values of the Unknown Sigma Variables Single 54
Sampling Plan, Variables Double Sampling Plan and VQSS
2.4 Parameters of Known Sigma Variables QSS for Some 55
Selected Combination of AQL and LQL Values
2.5 Parameters of Unknown Sigma Variables QSS for Some 55
Selected Combination of AQL and LQL Values
3.1 Optimal Parameters of Known Sigma VQSS with Double 70
Specification Limits (Symmetric Fraction Non-conforming)
3.2 Optimal Parameters of Known Sigma VQSS with Double 72
Specification Limits (Asymmetric Fraction Non-conforming)
3.3 Optimal Parameters of Unknown Sigma VQSS with Double 74
3.4 Optimal Parameters of Unknown Sigma VQSS with Double 76
3.5 Average Sample Number of Variables Single Sampling Plans 78
and VQSS with Double Specification Limits (Symmetric
Fraction Non- conforming)
xx
4.1. Variables Tightened-Normal-Tightened Sampling Scheme 93

Indexed by AQL and LQL for =5% and =10% Involving
Minimum ASN
4.2 ASN Values of the Known Sigma Variables Single Sampling 97
Plan, Variables Double Sampling Plan and Variables TNT
Scheme
4.3 ASN Values of the Unknown Sigma Variables Single 97
Sampling Plan, Variables Double Sampling Plan and
Variables TNT Scheme
4.4 Parameters of Known Sigma Variables TNT Scheme 98
for different AQL and LQL Values
4.5 Parameters of Unknown Sigma Variables TNT Scheme 98
for different AQL and LQL Values
5.1 Variables Chain Sampling Plans Indexed by AQL 116
and LQL for =5% and =10%
5.2 Variables Chain Sampling Plans Indexed by AQL 119
5.3 Variables Single Sampling Plans Indexed by AQL 119
5.4 ASN Values of the Variables SSP, DSP and Variables 120
Chain Sampling Plans
6.1 Variables Resampling Scheme (with m=2) Indexed by 134
AQL and LQL for =5% and =10%
6.2 ASN Values of the Variables Single, Double Sampling 138
Plans and Resampling Schemes
xxi
6.3 Sample Size of the Attributes and Variables Resampling 138

Schemes
6.4 Average Sample Number of the Variables resampling 139
Schemes for different m values
6.5 Variables Resampling Schemes (with m=2) Involving 140
Minimum ASN Indexed by AQL and LQL
6.6 Variables Resampling Schemes (with m=2) Involving 141
Minimum Sum of ASN Indexed by AQL and LQL
xxii
LIST OF FIGURES
FIGURE NO. FIGURE CAPTIONS PAGE NO.
1.1 Operating Characteristic Curve

2.1 OC Curves of Single Sampling Normal Plan (10, 1.754), 53
Quick Switching System (10; 1.754, 2.179) and
Single Sampling Tightened Plan (10, 2.179)
4.1 OC Curves of Single Sampling Normal Plan (12, 1.857), 96

TNT Scheme (63, 12; 1.857) and Single Sampling
\ Tightened Plan (63, 1.857)
5.1 OC Curves of a Variables Chain Sampling Plan for 121

Different i Values
6.1 ASN Curves of a Variables Single Sampling Plan and 142

Resampling Sampling Scheme
6.2 OC curves of a Variables Resampling Scheme for 143

different m values
xxiii
LIST OF SYMBOLS AND ABBREVIATIONS
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
ks Rejection criteria for unknown sigma plan

Number of preceding lots considered for accepting current lot for
known sigma plan
s Number of preceding lots considered for accepting current lot for
unknown sigma plan
v The value which is to be compared with acceptance criterion for
making decision
The population mean
X The sample mean
S The sample standard deviation
The population standard deviation
() The cumulative distribution function of standard normal
distribution
ASN Average Sample Number
SSP Single Sampling Plan
DSP Double Sampling Plan
QSS Quick Switching System
TNT Tightened Normal - Tightened
ChSP Chain Sampling Plan
CHAPTER 1
INTRODUCTION
Inspection of raw materials, semi finished products, or finished products

are one aspect of quality assurance. Whenever a statistical technique is used to
control, maintain and improve the quality, it is termed as statistical quality control.
When inspection is for the purpose of acceptance or rejection of a
product, based on adherence to a standard, the type of procedure employed is
usually called acceptance sampling. Acceptance sampling is one of the major
components in the field of Statistical Quality control.
A company receives a shipment of product from a vendor. This product
is often a component or raw material used in the companys manufacturing
process. A sample is taken from the lot, and some quality characteristic of the units
in the sample is inspected for a specified period of time. On the basis of the
information in this sample, a decision is made regarding lot disposition. Usually,
this decision is either to accept or to reject the lot. Accepted lots are put into
production; rejected lots may be returned to the vendor or may be subjected to
some other lot disposition action.
This chapter comprises of the following sections
SECTION 1 BASIC CONCEPTS OF QUALITY CONTROL
SECTION 2 BASIC CONCEPTS OF ACCEPTANCE SAMPLING
SECTION 3 A REVIEW ON ACCEPTANCE SAMPLING PLAN BY
VARIABLES
SECTION 4 A REVIEW ON CERTAIN SPECIAL PURPOSE SAMPLING
PLANS BY ATTRIBUTES
SECTION 5 A REVIEW ON SPECIAL PURPOSE SAMPLING PLANS
BY VARIABLES
2
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.
International Organization for Standardization (ISO) 8402-1986

standard defines quality as the totality of features and characteristics of a product
or service that bears its ability to satisfy stated or implied needs. ISO 9000
defines quality as "degree to which a set of inherent characteristics fulfills
requirements".
1.2 QUALITY CONTROL

American Society for Quality (ASQ) defines, Quality Control as the
operational techniques and activities used to fulfill requirements for quality.
1.3 STATISTICAL QUALITY CONTROL

Whenever a statistical technique is employed to control, improve and
maintain the quality or to solve quality problem it is termed as Statistical Quality
Control (SQC). The new era of quality control development began during the
World War II when SQC was much needed due to mass production. It is used
throughout the quality system at various stages of production such as
3
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.
1.4 TOOLS OF SQC

The SQC is the term used to describe a set of statistical tools used by
quality professionals. The following are the statistical tools used generally for the
purpose of exercising control, improvement of quality, enhancement of
productivity, creation of consumer confidence and development of the industrial
economy of the country.
Frequency Distribution: It is a tabulation or tally of the number of times a
given quality characteristic occurs within the samples. Graphic
representation of frequency distribution will show the average quality,
4
spread of quality, comparison with specific requirements and process

capability.
Control Chart: It is a graphical representation of quality characteristics,
which indicates whether the process is under control or not.
Acceptance Sampling: In order to maintain the quality of purchased lots,
two major alternatives are open to a buyer. One, complete inspection: every
single item in the lot is inspected and tested. Two, partial inspection: a
sample of items is taken, the sampled items are inspected and tested, and
the lot as a whole is accepted or rejected depending on whether few or
many non-conforming items are found in the sample. This type of sampling
called acceptance sampling which is the process of randomly inspecting a
sample of goods and deciding whether to accept the entire lot based on the
results. Acceptance sampling determines whether a batch of goods should
be accepted or rejected.
Analysis of the data: This includes techniques such as analysis of
correlation, analysis of variances, analysis for engineering design, problem
solving technique to eliminate cause of troubles.
1.5 BENEFITS OF STATISTICAL QUALITY CONTROL

SQC ensures rapid and efficient inspection at a minimum cost. It finds out
the cause excessive variability in manufactured products by forecasting
trouble before rejections occur and reducing the amount of spoiled work.
It exerts more effective pressure for quality improvement than that of a 100
percent inspection.
It easily detects faults. For example, using control charts one can easily
examine the deterioration in quality by verifying whether the points fall
above the upper control limits or below the lower control limits.
5
So long as the statistical control continues, specifications can be accurately

predicted for future, by which it is possible to assess whether the production
processes are capable of producing the products with the given set of
specifications.
Increases output and reduces wasted machines and materials resulting in
higher productivity.
Better customer relations through general improvement in product and
higher share of the market.
It provides a common language that may be used by designers, production
personnel and inspectors.
It says when and where 100 percent inspection is required.
Creates quality awareness in employees.
6
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.
1.2.2 NECESSITY OF ACCEPTANCE SAMPLING PLANS

Acceptance sampling plan is an essential tool in the SQC and is a
methodology which deals with quality contracting on product orders between the
producers and the consumers and thus allows the producers to take decision to
accept or reject the manufactured products based on the inspection of samples. It is
the process of evaluating a portion of the product/material in a lot for the purpose
7
of accepting or rejecting the lot as either conforming or not conforming to a quality

specification
Acceptance sampling is necessary to limit the cost of inspection and is

the only available method to appraise the quality in destructive testing. Acceptance
sampling itself does not improve quality, but whenever the lot is rejected it
indicates the instability of the production process. Acceptance sampling is cost
efficient and the only admissible method of efficient tests with quick results.
1.2.3 MAJOR AREAS OF ACCEPTANCE SAMPLING

Acceptance sampling deals with procedures/algorithms by which
decision to accept or reject a lot is based on the results of the inspection of
samples.
According to Duncan (1986), an acceptance sampling plan is likely to

be implemented when the following holds:
When the cost of inspection is high and the loss arising from the passing of a
non-conforming unit is not great.
When a 100 percent inspection is fatiguing.
When inspection is destructive i.e., a situation where inspection is not
possible without destroying the article chemically or physically.
Where there are great quantities or areas to be inspected.
When it is desired to stimulate the maker and/or the buyer.
According to Dodge (1969), the major areas of acceptance sampling

may be classified under the following four broad categories,
8
1. Lot-by-Lot acceptance sampling by the method of attributes, in which each

unit in a sample is inspected on a go-not-go basis for one or more
characteristics.
2. Lot- by-Lot sampling by the method of variables, in which each unit in a

sample is measured for a single characteristic, such as weight or strength, etc.
3. Continuous sampling of flow of units by the method of attributes and
4. Special purpose plans including chain sampling, skip-lot sampling and small
sample plans etc.
1.2.4 BASIC TERMINOLOGIES AND DEFINITIONS
SAMPLING PLAN, SCHEME AND SYSTEM

American National Standards Institute / American Society for Quality
Control (ANSI / ASQC) Standard A2 (1987) defines an acceptance sampling plan
as a specific plan that states the sample size or sizes to be used and the associated
acceptance and non-acceptance criteria It defines an acceptance-sampling scheme
as a specific set of procedures which usually consists of acceptance sampling
plans in which lot sizes, sample sizes and acceptance criteria or the amount of
100% inspection and sampling are related. The MIL-STD-105 D (1963) tables
and procedures are the examples for sampling scheme. Stephens and Larson
(1967) define a sampling system as an assigned grouping of two or three
sampling plans and the rules for using (that is, switching between) these plans for
sentencing lots or batches of articles to achieve blending of the advantageous
features of the sampling plan. Quick Switching System (QSS) of Romboski
(1969) is an example for a sampling system.
9
CUMULATIVE AND NON CUMULATIVE RESULTS SAMPLING

PLANS
Dodge and Stephens (1966) defines a non cumulative sampling plan as

one which uses the current sample information from the process or current product
entity in making decisions about process or product quality. Single and double
sampling plans are examples of non cumulative sampling. Cumulative results
sampling inspection is one which uses the current and past information from the
process in making a decision about the process. Chain sampling plan of Dodge
(1955) is an example for cumulative results sampling plan.
INSPECTION
ANSI / ASQC Standard A2 (1987) defines the term inspection as

activities, such as measuring, examining, testing, gauging one or more
characteristics of a product or service and comparing them with specified
requirements to determine conformity. A sampling scheme or a sampling system
may contain three types of inspections viz normal, tightened and reduced
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.
OPERATING CHARACTERISTIC (OC) CURVE

The Operating Characteristic curve (OC) is a picture of a sampling
plan. Each sampling plan has a unique OC curve. The probability of acceptance
can be regarded as a function of the deviation of the specified value 0 of the mean
from its true value . This function is called OC function of the sampling plan. An
OC curve is developed by determining the probability of acceptance for several
values of incoming quality.
Figure 1.1 Operating Characteristic (OC) Curve

11
The OC curves are generally classified as Type A and Type B OC

curves. ANSI/ASQC Standard A2 (1987) defines the terms as follows:
Type A OC curve is used for isolated or unique lots, or a lot from an
isolated sequence. A curve showing, for a given sampling plan, the
probability of accepting a lot as a function of lot quality.
Type B OC curve is used for a continuous stream of lots. A curve
showing, for a given sampling plan, the probability of accepting a lot
as a function of the process average.
In sampling systems or schemes, one will have a composite OC curve

which gives the steady state probability of acceptance under the switching rules of
the system or scheme as a function of process quality.
To evaluate the probability of acceptance, Pa (p), hypergeometirc model

is exact for type A situation (when sampling attribute characteristics from a finite
lot without replacement). Under type B situation, binomial model will be accurate
for the case of non-confirming units to calculate Pa (p). Binomial model is also
correct in case of sampling from a finite lot with replacement.
Poisson model is accurate in calculating Pa (p), which specifies a given

number of non-conformities per unit (or non-conformities per hundred units). In
case of variable sampling plans normal distribution (Gaussian) is widely used to
compute relevant measures of sampling plans. Hyper geometric, binomial, Poisson
and normal distributions are the distributions commonly used in the development
of acceptance sampling plans. Schilling (1982) has given the conditions under
which each of these is to be used.
12
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.
The probability of getting exactly x defectives in a sample of size n is

given by the probability mass function:
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 .
The probability density of the normal distribution is:

x 2
1
f x, , e 2 2
, x ,
2
2
, 0
Here, is the mean
is its standard deviation
AVERAGE SAMPLE NUMBER (ASN)

ANSI / ASQC Standard A2 (1987) defines ASN as the average
sample units per lot used for making decisions either acceptance or non-
acceptance. A plot of ASN against process quality is called ASN curve. ASN will
14
be affected according to the type of curtailment of inspection (on acceptance and

rejection decisions). Sampling inspection is to be called fully curtailed if sampling
is stopped whenever decision could be reached on acceptance (or rejection) before
reaching the prescribed sample size.
AVERAGE OUTGOING QUALITY (AOQ)

ANSI / ASQC Standard A2 (1987) defines AOQ as the expected
quality of outgoing product following the use of an acceptance sampling plan for
a given value of incoming product quality.
AVERAGE OUTGOING QUALITY LIMIT (AOQL)

The maximum AOQ over all possible levels of incoming quality is
termed as AOQL. The assumption underlying in this expression is that for all
accepted lots the average proportion non-conforming is assumed to be p and for all
rejected lots the entire units are being screened and non-conforming units are
replaced. A plot of AOQ against p is called AOQ curve.
AVERAGE TOTAL INSPECTION (ATI)

According to ANSI / ASQC Standard A2 (1987), ATI is the average
number of units inspected per lot based on the sample size for accepted lots and all
inspected units in rejected lots. ATI is not applicable whenever testing is
destructive. A plot of ATI against p is called ATI curve.
ACCEPTABLE QUALITY LEVEL (AQL)

ANSI/ASQC Standard A2 (1987) defines AQL as the maximum
percentage or proportion of variant units in a lot or batch that, for the purpose of
acceptance sampling, can be considered satisfactory as a process average.
15
LIMITING QUALITY LEVEL (LQL)

ANSI/ASQC Standard A2 (1987) defines LQL as the percentage or
proportion of variant units in a batch or lot for which, for the purposes of
acceptance sampling, the consumer wishes the probability of acceptance to be
restricted to a specified low value.
INDIFFERENCE QUALITY LEVEL (IQL)

The percentage of variant units in a batch or lot for which, for purposes
of acceptance sampling, the probability of acceptance to be restricted to a specific
value namely 0.50.
1.2.5 DESIGNING METHODOLOGY

In designing a sampling plan, one has to accomplish a number of
different purposes. According to Hamaker (1960), the most important are
1. To strike a balance between the consumers requirement, the producers

capabilities and the inspectors capacity.
2. To separate bad lots from good one.
3. Simplicity of procedures and administration.
4. Economy in number of observations.
5. To reduce the risk of wrong decisions with increasing lot size.
6. To use accumulated sample data as valuable source of information.
7. To exert pressure on the producer or supplier when the quality of the lot
received is unreliable up to standard.
8. To reduce sampling when the quality is reliable and satisfactory.
Hamaker (1960) also pointed out that these aims are partly conflicting
and all of them cannot be simultaneously realized.
16
The design methodologies of acceptance sampling may be categorized

as in the following.
Risk Based Economical

Based
Non Bayesian 1 2
Bayesian 3 4
Risk based sampling plans are traditional in nature, drawing upon

procedure and consumer type of risks as depicted by the OC curve. Economically
based sampling plans explicitly consider such factors as costs of inspections,
accepting a non conforming unit and rejecting a conforming unit in an attempt to
design a cost effective plan. Bayesian plan design takes into account the past
history of similar lots submitted previously for inspection purposes.
Non Bayesian plan design is not explicitly based upon the past lot history.
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,
p1= fraction non-conforming that is considered to be good so that

producer expects lot of quality p1 to be accepted most of the time.
p2= fraction non-conforming that is considered to be poor so that the
consumer expects lot of quality p2 to be rejected most of the time.
= the producers risk of rejecting p1 quality and
= the consumers risk of accepting p2 quality.
Sampling Plans of Cameron (1952) are the examples of this type of

designing. Schilling and Sommers (1981) term p1 as the Producers Quality Level
(PQL) and p2 as the Consumers Quality Level (CQL). Earlier literature calls p1 as
the Acceptable Quality Level (AQL) and p2 as the Limiting Quality Level (LQL)
or Rejectable Quality Level (RQL) or Lot Tolerance Percent Defective (LTPD).
Traditionally the values of and are assumed to be 95 % and 10 % respectively.
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.
3. The plan is specified by imposing upon the OC curve two or more

independent conditions none of which is explicitly involve the OC curves.
Dodge and Romig (1959) AOQL Sampling plans are the examples for this
type of designing.
18
1.2.5.1 DESIGNING METHODOLOGY FOLLOWED IN THIS THESIS
In this thesis, Search procedure has been followed. In this approach,

the parameters of a sampling plan are chosen, by trial and error by varying the
parameters in a uniform fashion depending upon the properties of OC function.
An example for this approach is the one followed by Guenther (1969,1970) while
determining the parameters of single and double sampling plans under the
conditions for application of binomial, Poisson and hyper geometric models of OC
curve. The advantage of search procedure is that the sample sizes need not be
rounded.
1.2.6 CLASSIFICATIONS OF ACCEPTANCE SAMPLING PLANS

The acceptance sampling plans are generally divided into two major
categories namely, attributes sampling plan and variables sampling plan.
1.2.6.1 ATTRIBUTES SAMPLING PLANS

Many quality characteristics cannot be conveniently represented
numerically. In such cases, we usually classify each item inspected as either
conforming (non-defective) to the specifications on that quality characteristics or
non conforming (defective) to those specifications. Quality characteristics of this
type are called attributes. Sampling plans applied to such quality characteristics are
called attributes sampling plans. Several sampling plans are available in the
literature for the application of attributes quality characteristics. For example,
single sampling plan, double sampling plan, multiple sampling plan, etc. (see Hald
(1981) and Schilling (1985)).
19
1.2.6.2 VARIABLES SAMPLING PLAN

Variables sampling plan specifies the number of items to be sampled
and the criterion for sentencing lots when measurements data are collected on
quality characteristic of interest. These plans are generally based on sample
average and sample standard deviations of the quality characteristic. When the
distributions 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 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
As many manufacturers begin to emphasize allowable numbers of non-

conforming parts per million (ppm), variables sampling becomes very attractive.
There are two cases in variables sampling plans.
(i) Sampling plans with known standard deviation

(ii) Sampling plans with unknown standard deviation
In these plans the decision on acceptance or rejection of the lot is

based on sample average alone. The decision specifications are associated with
each inspection characteristics. Many specification are one sided. (i.e.) The
specification merely states a lower specification limit L or an upper specification
limit U to apply to individual article.
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 .
1.3.1 LITERATURE REVIEW OF VARIABLES SAMPLING PLANS

Wallis (1947) suggested an approximation for finding the parameters
for unknown standard deviation plan from that of known plan. Military Standard
414, Department of Defense in 1957 issued Sampling procedures and Tables for
Inspection by variables for Percent Defective which was the culmination of the
developments in variables sampling plans. Lieberman and Resnikoff (1955)
developed tables and procedures for the selection of variable sampling plan
parameters for various AQL values given in the MIL-STD 414 scheme. They have
considered variables sampling plans for assuring the product quality when the
quality characteristic of the product follows normal distribution with unknown
standard deviation and provided a procedure for calculating the non-central t-
distribution. Owen (1967) developed variables sampling plans based on normal
distribution when the process standard deviation is unknown. Bender (1975)
considered variables sampling plans for assuring the product quality when the
quality characteristic of the product following normal distribution with unknown
standard deviation and provided a procedure for calculating the non-central t-
distribution. Hamaker (1979) has given a procedure of finding the parameters of
the unknown sigma variables sampling plans from the known sigma variables
sampling plans. Schneider and Wilrich (1981) investigated the robustness of
variables sampling plans. Govindaraju and Soundararajan (1986) developed tables
for selecting the parameters of variables single sampling plans that match with the
OC curves of MIL-STD 105D (1963) schemes. Kao (1971) provided the
comparison between the attribute acceptance sampling plans and the variable
22
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.
Recently, there are developments in designing various variables

sampling plans. Pearn and Wu (2006) investigated the variables sampling plans
for very low fraction non-conforming. Pearn and Wu (2007) proposed an effective
decision making method for product acceptance based on measurement data. Sheu
et al. (2014) developed a variables sampling plan based on incapability index Cpp
proposed to deal with lot sentencing. Yen et al. (2014) developed variable
sampling plan using the exponentially weighted moving average (EWMA)
statistic based on the yield index for lot sentencing.
23
SECTION 4
A REVIEW ON CERTAIN SPECIAL PURPOSE SAMPLING PLANS BY
ATTRIBUTES
The special purpose sampling inspection plans often known as special

purpose plans is one of the major areas of acceptance sampling which is classified
under fourth category of Dodges classification (1969), are tailored for special
applications as against general or universal use.
Special purpose sampling plans are also known as conditional sampling

plans were developed to overcome the short comings of zero acceptance single
sampling plans whenever samples of small sizes only are practically possible for
disposition of lots. Zero acceptance single sampling plan results in rejection of a
lot even if there is only one nonconforming unit is observed in the sample thereby
resulting in a poor operating characteristic (OC) curve. This is applicable
whenever high quality product is desirable.
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.
Balamurali and Kalyanasundaram (1997) determined a new sampling

scheme called an attribute single sampling scheme. Balamuali and
.Kalyanasundaram (1999) introduced conditional double sampling scheme and
they have made comparison with the single sampling scheme.
24
1.4.1 QUICK SWITCHING SYSTEM

A sampling system consists of two or more sampling plans and the
rules for switching between the sampling plans to achieve a blending of the
advantageous features of each of the sampling plans. In general, any sampling
system of inspection involving only normal and tightened inspection will be
referred to as a two-plan system.
Quick switching system (QSS) developed by Dodge (1967) is one of

the two-plan systems for the application of attributes quality characteristics. In
any two plan system, the tightened inspection can be used when the quality of a
product deteriorated and normal inspection is used when the quality is found to be
good. Dodge (1965), Hald and Thyregod (1965) and Stephen and Larson (1967)
have investigated the two-plan systems using different switching criteria to
achieve the desired discrimination on the operating characteristic (OC) curve.
Romboski (1969) has investigated the QSS of type QSS-1 by taking attributes
single sampling plan as the reference plan. Arumainayagam and Soundararajan
(1994, 1995) have constructed quick switching double sampling system by
tightening the acceptance number and tightening the sample sizes respectively.
Balamurali and Kalyanasundaram (1996) introduced procedures and constructed
tables for the selection of zero acceptance number quick switching systems.
Govindaraju (2011) designed zero acceptance number chained QSS. Balamurali
and Usha (2013) have investigated the QSS under the Weibull life time model.
The application of the system is as follows.

(1) Adopt a pair of sampling plans i.e., a normal plan (N) and a tightened plan (T).
(2) Use plan N for the first lot.
(3) For each lot inspected, if the lot is accepted, then use the plan N for the next
lot; if the current lot is rejected, then use plan T for the next lot.
25
OPERATING PROCEDURE OF QUICK SWITCHING SYSTEM

The operating procedure of attributes QSS-1 is as follows.
Step 1: Start with normal inspection. During normal inspection, take a random
sample of size n and inspect. Observe the number of non-conforming items
in the sample say d.
Step 2: Accept the lot if d cN and reject the lot if d > cN. If a lot is rejected on
normal inspection, then switch to tightened inspection as in Step 3.
Otherwise continue the normal inspection for the next lot.
Step 3: During tightened inspection, take a random sample of size n and inspect.
Observe the number of non-conforming items in the sample says d.
Step 4: Accept the lot if d cT and reject the lot if d > cT. If a lot is accepted on
tightened inspection, then switch to normal inspection as in Step 1.
Otherwise continue the same tightened inspection for the next lot.
( Note: c N cT )
MEASURES OF QSS
The important measures of QSS that describe the operation of an
acceptance sampling plan for various fraction nonconforming are,
1. The OC function (see Dodge (1967)) is
PT
Pa ( p )
1 PN PT
where PN Pr v k N is the probability of accepting a lot based on a
single sample with parameters (n , k N )
and PT Pr v k T is the probability of accepting a lot based on a
single sampling plan with parameters (n , k T )

26
2. The Average Sample Number (ASN) is

ASN p n
3. The Average Outgoing Quality (AOQ) is

AOQ p pPa ( p )
1.4.2 CHAIN SAMPLING PLANS (ChSP)

One of the cumulative results plans is the chain sampling plan (ChSP)
introduced by Dodge (1955), which made use of previous lots results, combining
with the current lot information, to achieve a reduction of sample size while
maintaining or even extending protection.
The ChSP was first conceived to overcome the problem of lack of

discrimination of the single sampling plan with acceptance number c 0 , and had
been received wide applications in industries where the test is either costly or
destructive.
Soundararajan (1978a, 1978b) had carried out further evaluations of

ChSP-1 type sampling plans. Since the invention of ChSP-1, numerous works had
been done on the extensions to chain sampling plans. These included the plans
designated as ChSP-2 and ChSP-3, which were developed by Dodge (1958) but
kept unpublished, partly due to the complexities of its operating procedures.
Frishman (1960) developed extended chain sampling plans designated as ChSP-4
and ChSP-4A (perhaps contemplating publication of designations 2 and 3 by
Dodge). These plans were developed from an application in the sampling
inspection of torpedoes for Naval Ordnance as a check on the control of the
production process and test equipment (including 100% inspection). Features of
these plans included a basic acceptance number greater than zero, an option for
27
forward or backward accumulation of results for an acceptance-rejection decision

on the current lot, and provision for rejecting a lot on the basis of the results of a
single sample (ChSP-4A). Some variations of chain sampling for which
cumulative results were used in the sentencing of lots had also been developed by
Anscombe et al. (1947), Page (1954), Hill et al. (1959), Ewan and Kemp (1960),
Kemp (1962), Beattie (1962), Wortham and Mogg (1970), Soundarajan (1978a,
1978b) and Vaerst (1982).
Further extensions to a general family of chain sampling inspection

plans had been developed by Dodge and Stephens (1966) and published in
numerous technical reports, conference papers, and journal articles. Raju (1996a,
1996b, 1991,1995, 1997) did extensive research work on chain sampling plans
both cooperatively and independently. His contribution included extending idea of
ChSP-1 and devising tables based on the Poisson model for the construction of
two stage chain sampling plans ChSP-(0, 2) and ChSP-(1, 2) under difference sets
of criteria, outlining the structure of a generalized family of three- stage chain
sampling plans, which extended the concept of two-stage chain sampling plans of
Dodge and Stephens (1966). He also authored a series of 5 papers, which
presented procedures and tables for the construction, and selection of chain
sampling plans ChSP-4A (c1, c2). Balamurali et al. (2008) have explained the
concepts of skip-lot sampling and chain sampling. Balamurali and Palaniswamy
(2012) have determined the minimum variance outgoing quality limit (VOQL)
chain sampling plans for compliance testing.
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,
1. The OC function(see Dodge (1955)) is given by
Pa p P0 , n P1, n P0 , n
i
P 0 , n = Probability of getting exactly 0 defective in a sample of size n
P1 , n = Probability of getting exactly 1 defective in a sample of size n
ASN p n
3 The Average outgoing Quality (AOQ) is given by
AOQ p pPa ( p )
29
1.4.3 TIGHTENED-NORMAL-TIGHTENED (TNT) SCHEME

The tightened-normal-tightened (TNT) sampling procedure developed
by Calvin (1977) is a particular case of the general two-plan system for the
inspection of attributes quality characteristics. This procedure is particularly
appropriate when the product is forthcoming in a stream of lots and a zero
acceptance number is to be maintained. This scheme utilizes two zero acceptance
number single sampling plans of different sample sizes namely n1 and n2 (< n1)
together with the switching rules and this scheme is designated as TNT-(n1, n2; 0).
Calvin (1977) has pointed out that, while increasing the protection to the
producer, the switching rules have no real effect on consumers quality level
namely LTPD or LQL which remains essentially that of the tightened plan. This
implies that the TNT scheme provides more protection to the producer while
safeguarding the consumers protection. Soundararajan and Vijayaraghavan
(1990) investigated the TNT scheme of type TNT-(n1, n2;c).
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
MEASURES OF TNT SCHEME

The important measures of the TNT scheme that describe the operation
of an acceptance sampling plan for various fraction non-conforming.
1. The OC function (see Calvin (1977)) is given by
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)
3. The Average outgoing Quality (AOQ) and
AOQ p pPa ( p)
1.4.4 RESAMPLING SCHEME

Govindaraju and Ganesalingam (1997) has proposed an attribute
sampling plan which can be applied in situations where resampling is permitted on
lots not accepted on original inspection. They have derived the performance
measures of the resampling scheme having single sampling attributes plan as the
reference plan. In this plan, it is assumed that during the course of resubmission,
31
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.
Recently, some of the authors have investigated the impact of

resampling scheme under various situations. For example, Aslam et al. (2011)
developed group acceptance sampling plan for resubmitted lots under Burr type
XII distribution. Aslam et. al .(2012) have developed Bayesian sampling
inspection for resubmitted lots under gamma-Poisson distribution.
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
MEASURES OF RESAMPLING SCHEME

The important measures that describe the operation of an acceptance
sampling plan for various fraction non-conforming, p are
1. The OC function (see Govindaraju and Ganesalingam (1997)) is given by

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
where Pa ( p ) Pr v k
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)
3. The Average outgoing Quality (AOQ)
AOQ p pP a p
33
SECTION 5
A REVIEW ON SPECIAL PURPOSE SAMPLING PLANS BY VARIABLES
The special purpose sampling plans were initially developed by Dodge

(1955) for attributes. Kuralmani and Govindaraju (1993) have investigated
conditional sampling plans for given AQL and LQL. Soundararajan and Palanivel
(1997)) have investigated on quick switching variables single sampling system
indexed by AQL and LQL by tightening acceptance criterion.
Govindaraju and Balamurali (1998) extended the idea of chain
sampling plans to variable inspection and examined the related properties and
listed the desired table. Balamurali and Jun (2006) have developed repetitive
group sampling procedure for variables inspection. Balamurali and Subramani
(2010) presented the procedures for designing of variables repetitive group
sampling plan indexed by indifference quality level and the relative slope on the
operating characteristic curve. Vijayaraghavan and Sakthivel (2011) have
developed chain sampling plans based on Bayesian methodology for variables
inspection.
Balamurali et al. (2005) have designed repetitive group sampling plan
for variables involving minimum average sample number. Balamurali and Jun
(2007) have developed multiple dependent state sampling plans for lot acceptance
based on the measurement data. Balamurali and Jun (2009) have designed a
variables two- plan system by minimizing the average sample number. Balamurali
and Subramani (2010) have designed of variables repetitive group sampling plans
indexed by point of control. Wu et al. (2012) and Aslam et al. (2013) investigated
the variables sampling plan for resubmitted lots based on the process capability
index Cpk. Balamurali et al. (2015) developed attribute-variable inspection lots
policy using resampling based on EWMA.
34
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.
2.2 CONDITIONS OF APPLICATION

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.
(iii) Inspection is by measurements, with quality is defined as the
fraction non-conforming, p.
In addition, the usual conditions for the application of variables single sampling
plans with known or unknown standard deviation should also be valid.
35
2.3 OPERATING PROCEDURE OF KNOWN SIGMA VQSS

The operating procedure of the VQSS is as follows.
Suppose that the quality characteristic of interest has the upper

specification limit U and follows a normal distribution with known standard
deviation . Then the following procedure of the VQSS is proposed.
sample of size n, say X 1 , X 2 ...X n

and compute v U X , where
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.
Step 3: During tightened inspection, take a random sample of size n, say
X , X U X , 1 n
1 2 ... X n and compute v

where X
n
X
i 1
i .
Step 4: Accept the lot if v k T and reject the lot if v k T (k T k N ) .

If a lot is rejected on tightened inspection, then immediately switch to
normal inspection as in Step 1.
Thus, the VQSS system is characterized by three parameters, namely

n, kN and kT. If kN=kT, then the system will reduce to the variables single
sampling plan.
36
2.4 OPERATING CHARACTERISTIC FUNCTION OF VQSS
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 )
where PN Pr v k N is the probability of accepting a lot based on a single
sampling plan with parameters (n, kN) and PT Pr v k T is the probability of
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.
The fraction non-conforming in a lot will be determined as
U
p 1 1 (v ) ( v ) (2.2)

where ( y ) is given by
y
1 z2
( y) exp dz , (2.3)
2 2
provided that the quality characteristic of interest is normally distributed with

mean and standard deviation , and the item is classified as non-conforming if it
exceeds the upper specification limit U.
37
Then its probability of acceptance is written as
( wT )
Pa ( p ) (2.4)
1 (w N ) ( wT )
where wT v k T n and wN v k N n
2.5 DESIGNING OF A KNOWN SIGMA VQSS

The OC function of a known sigma VQSS is given in (2.4). If two
points on the OC curve namely, AQL(=p1), LQL(=p2), the producers risk and
the consumers risk are prescribed then the OC function can be expressed as
( wT 1 )
1 (2.5)
1 ( w N 1 ) ( wT 1 )
( 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.
That is, wT 1 (v1 k T ) n , w N 1 (v1 k N ) n
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.
2.6 AVERAGE SAMPLE NUMBER
The average sample number (ASN), by definition, means the expected

number of sampled units required for making decisions about the lot. The concept
38
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)

2.7 OPTIMAL DESIGNING OF KNOWN SIGMA VQSS
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.
2.8 OPTIMAL DESIGNING OF UNKNOWN SIGMA VQSS
Whenever the standard deviation is unknown, we should use the sample

standard deviation S instead of . In this case, the operation of the proposed
system is as follows.
39
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
inspection, then switch to tightened inspection.
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
on tightened inspection, then immediately switch to normal inspection as
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.
Hamaker (1979) has given an approximation for finding the parameters

of the unknown sigma single sampling plan from the parameters of the known
sigma single sampling plan. The relationship between known and unknown sigma
plan parameters is true only for single sampling plan.
40
Soundarajan and Palanivel (1997) have followed the same approximation

for selecting the parameters of unknown sigma VQSS. However it is to be
pointed out that Hamakers (1979) results must be extended to VQSS system
rather than wrongly assuming that the same approximation is valid for VQSS. So
the entire design of unknown sigma schemes provided in Soundararajan and
Palanivel (1997) seems faulty. So we will follow a different procedure for the
unknown sigma case.
The determination of parameters for the unknown sigma case namely

(nS, kNS, kTS) is slightly different from the known sigma case. It is known that
X k NS S is approximately normally distributed with mean k NS E (S ) and
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 )
We obtain w N 1S , wT 1S , w N 2 S , wT 2 S corresponding to w N 1 , wT 1 , w N 2 , wT 2 respectively by

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
where Pa ( p1 ) and Pa ( p 2 ) are the lot acceptance probabilities of the proposed

sampling system at AQL and LQL respectively and are described in (2.10) and
(2.11).
We may determine the parameters of the known sigma and
unknown sigma VQSS by solving the nonlinear equation given in (2.9) and
(2.12) respectively. There may exist multiple solutions since there are three
unknowns with only two equations. Generally a sampling would be desirable if
the required number of sampled is small. So, in this chapter, we consider the
ASN as the objective function to be minimized with the probability of
acceptance along with the corresponding producers and consumers risks as
constraints. To solve the above nonlinear optimization problems given in (2.9)
and (2.12), the sequential quadratic programming (SQP) proposed by Nocedal
and Wright (1999) can be used. The SQP is implemented in Matlab software
using the routine fmincon. By solving the nonlinear problem mentioned above,
the parameters (n, kN and kT) for known sigma plan and the parameters (nS, kNS
and kTS) for unknown sigma plan are determined and these values are tabulated in
Table 2.2.
2.9 EXAMPLES
2.9.1 SELECTION OF KNOWN SIGMA VQSS INDEXED BY AQL AND

LQL
Table 2.1 is used to determine the parameters of the known sigma

VQSS for specified values of AQL and LQL when = 5% and = 10%. For
example, if p1 = 2%, p2 = 8%, = 5% and = 10%, Table 2.1 gives the
parameters as n = 12, kN = 1.552 and kT = 1.817.
43
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.
Step 2: Select a random sample of size 12 and compute v

U X , where

1 12
X X i . Accept the lot if v 1.817 and reject the lot if v 1.817 .
12 i 1
Switch to normal inspection as in step 1, if a lot accepted in the tightened
inspection phase.
2.9.2 SELECTION OF UNKNOWN SIGMA VQSS INDEXED BY AQL

AND LQL
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.
2.10 ADVANTAGES OF THE VQSS
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
Suppose that for given values of AQL=0.01, =5%, LQL=0.02 and =

10%, one can find the parameters of the attributes QSS under the application of
Poisson model as
(i) n = 251, cT = 5 and cN = 9
For the same AQL and LQL, we can determine the parameters of the
variables single sampling plan (from Sommers (1981)) and VQSS (from Table
2.1) respectively as follows.
(ii) n = 116 and k = 2.17
(iii) n = 28, kN = 1.923 and kT = 2.418
By comparing the above, it is clear that the VQSS achieves a reduction
of over 89% in sample size than the attributes QSS and about 76% than the
variables single sampling plan with same AQL and LQL conditions. In order to
show the better efficiency of the VQSS, three OC curves are considered.
Figure 2.1 shows the OC curves of the variables single sampling plans
with parameters (10, 1.754) and (10, 2.179) and the VQSS with parameters (10;
1.754, 2.179). The VQSS (10; 1.754, 2.179) is selected in such a way that it
satisfies the two-points (p1 = 0.01, 1- = 0.95) and (p2 = 0.045, = 0.10) on the
OC curve.
2.11. COMPARISON
In this section, we compare the parameters of VQSS with those of

VQSS given in Soundararajan and Palanivel (1997). It is to be pointed out that the
ASNs of VQSS provided in Soundararajan and Palanivel (1997) are equal or
greater than the ASN of single sampling plans for some combinations of p1 and
p2. For example, for given p1=0.01 and p2=0.04, ASN of variables single sampling
plan is 506, but the ASN of VQSS given in Soundararajan and Palanivel (1997) is
1856. This is a contradiction, since in the attributes case, QSS will always have
45
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.
Further, it is also to be pointed that the VQSS is economically superior

to the variables double sampling plan in terms of ASN. Obviously, a sampling
plan having smaller ASN would be more desirable. The variables double or
multiple sampling plans are not practically very useful. Variables sampling
Standards avoid presenting such plans due to increased complexity involved in
operating them.
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
0.0025 0.004 96 2.841 2.596 326 2.917 2.542

0.005 35 2.917 2.432 156 2.945 2.445
0.006 26 2.878 2.398 113 2.910 2.410
0.0075 23 2.773 2.413 82 2.861 2.371
0.010 14 2.762 2.302 56 2.809 2.309
0.015 9 2.693 2.193 37 2.710 2.235
0.020 8 2.572 2.182 33 2.563 2.233
0.025 6 2.569 2.079 28 2.490 2.200
0.030 8 2.353 2.213 19 2.564 2.074
0.040 5 2.374 2.039 15 2.485 2.010
0.005 0.0075 100 2.626 2.361 330 2.679 2.329

0.010 31 2.683 2.183 123 2.705 2.205
0.012 23 2.641 2.146 89 2.664 2.169
0.015 19 2.549 2.139 63 2.617 2.122
0.020 9 2.657 1.897 45 2.533 2.073
0.030 9 2.377 1.982 28 2.443 1.978
0.035 7 2.385 1.895 23 2.424 1.929
0.040 7 2.298 1.913 20 2.394 1.894
0.060 6 2.109 1.884 15 2.224 1.834
47
Table 2.1. Contd.
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
0.010 0.015 68 2.427 2.047 243 2.434 2.064

0.020 28 2.418 1.923 94 2.441 1.946
0.025 21 2.334 1.899 62 2.393 1.893
0.030 15 2.314 1.829 47 2.344 1.854
0.035 13 2.255 1.810 38 2.304 1.819
0.040 14 2.141 1.856 32 2.268 1.788
0.045 10 2.179 1.754 27 2.247 1.752
0.050 14 2.005 1.875 24 2.214 1.729
0.060 7 2.123 1.648 20 2.148 1.693
0.070 6 2.084 1.599 17 2.100 1.655
0.080 6 1.987 1.617 14 2.089 1.599
0.090 6 1.904 1.629 11 2.138 1.513
0.100 3 1.906 1.557 11 2.027 1.532
0.015 0.020 129 2.236 1.971 402 2.248 1.973

0.025 43 2.265 1.835 121 2.317 1.827
0.030 26 2.252 1.757 78 2.279 1.779
0.035 23 2.163 1.768 59 2.233 1.748
0.040 22 2.082 1.795 48 2.188 1.723
0.045 21 2.012 1.787 39 2.167 1.687
0.050 13 2.083 1.658 34 2.131 1.666
0.060 11 2.009 1.629 26 2.088 1.613
0.070 8 2.021 1.526 22 2.025 1.585
0.080 7 1.952 1.526 18 2.005 1.535
48
Table 2.1 Contd.
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
0.020 0.030 87 2.070 1.840 170 2.167 1.772

0.035 33 2.158 1.673 96 2.172 1.702
0.040 26 2.108 1.653 68 2.153 1.658
0.045 21 2.076 1.621 54 2.120 1.630
0.050 17 2.060 1.580 50 2.046 1.636
0.060 17 1.926 1.618 33 2.044 1.554
0.070 17 1.813 1.638 26 2.004 1.510
0.080 12 1.817 1.552 24 1.910 1.510
0.090 8 1.877 1.417 19 1.911 1.451
0.100 7 1.851 1.376 16 1.897 1.407
0.120 5 1.700 1.345 18 1.664 1.474
0.030 0.040 116 1.934 1.679 269 1.965 1.665

0.045 57 1.957 1.592 133 1.999 1.584
0.050 33 1.998 1.493 89 1.999 1.535
0.060 22 1.947 1.442 54 1.974 1.469
0.070 17 1.894 1.404 40 1.925 1.430
0.080 14 1.844 1.374 31 1.893 1.388
0.090 12 1.800 1.345 26 1.846 1.361
0.100 10 1.781 1.296 29 1.695 1.415
0.120 11 1.597 1.362 24 1.598 1.388
0.150 6 1.647 1.152 13 1.644 1.224
0.200 5 1.470 1.110 10 1.458 1.173
0.040 0.060 58 1.801 1.476 104 1.892 1.427

0.070 29 1.826 1.361 63 1.869 1.369
0.080 21 1.796 1.311 47 1.819 1.339
0.090 17 1.754 1.279 37 1.778 1.308
0.100 14 1.725 1.240 30 1.749 1.274
0.110 14 1.639 1.264 26 1.704 1.254
0.120 14 1.565 1.280 24 1.644 1.249
0.140 10 1.546 1.191 17 1.634 1.169
0.160 8 1.512 1.127 17 1.195 1.485
49
Table 2.1 Contd.
0.18 7 1.460 1.090 12 1.521 1.096
0.200 4 1.317 0.882 10 1.496 1.046
0.050 0.060 280 1.666 1.521 462 1.708 1.488

0.070 66 1.723 1.373 132 1.761 1.366
0.080 35 1.735 1.280 73 1.768 1.293
0.090 25 1.711 1.231 53 1.730 1.260
0.100 23 1.630 1.245 40 1.709 1.219
0.120 14 1.619 1.134 28 1.640 1.170
0.140 11 1.560 1.085 21 1.585 1.120
0.160 9 1.511 1.036 16 1.560 1.060
0.200 6 1.432 1.006 15 1.328 1.083
0.250 5 1.324 0.860 8 1.375 0.895
50
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
0.8 Variables Normal SSP

0.7
Probability of Acceptance, Pa(p)
0.6
VQSS
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.2.1. OC Curves of Single Sampling Normal Plan (10, 1.754),

Quick Switching System (10; 1.754, 2.179) and Single
Sampling Tightened Plan (10, 2.179)
51
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
0.001 0.002 191 154.9 191 134

0.001 0.003 74 59.4 74 45
0.005 0.010 138 112.0 138 31
0.005 0.012 85 69.5 85 23
0.03 0.04 506 434.6 506 116
0.03 0.06 81 127.7 81 22
0.04 0.07 114 180.6 114 29
0.04 0.08 72 58.4 72 21
0.05 0.07 300 246.7 300 66
0.05 0.08 149 122.3 149 35
* ASN given in Soundararajan and Palanivel (1997)
** ASN given in Table 2.1
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)
52
Table 2.4: Parameters of Known Sigma VQSS for Some Selected

Combinations of AQL and LQL Values
*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 from Soundararajan and Palanivel (1997)

**Parameters from Table 2.1
Table 2.5: Parameters of Unknown Sigma VQSS for Some Selected

Combinations of AQL and LQL Values
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
*Parameters from Soundararajan and Palanivel (1997)

53
CHAPTER 3
OPTIMAL DESIGNING OF VARIABLES QUICK SWITCHING SYSTEM

WITH DOUBLE SPECIFICATION LIMITS
3.1 INTRODUCTION
In this chapter, we have investigated VQSS when a measurable quality

characteristic has both upper and lower specification limits and follows normal
distribution. Most of the variables sampling plans are dealing with quality
characteristics having only one specification limit either lower or upper
specification limit. In the literature, there have been some studies available which
are dealing with sampling plans based on double specification limits. Military
Standard MIL-STD-414 (1957) described the procedure for the selection of a
variables single sampling plan involving the double specification limits. Schilling
(1982) suggested the use of two single sampling plans to handle the problem of
quality characteristics having double specification limits. Since the variables
sampling plans with double specifications are having wide applications, this
chapter deals with the designing methodology for determining the parameters of a
VQSS for normally distributed quality characteristics having double specification
limits.
3.2 CONDITIONS FOR APPLICATION OF VQSS
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
(iii) Inspection is by measurements, with quality is defined as the fraction

nonconforming, p.
In addition, the usual conditions for the application of variables single
sampling plans with known or unknown standard deviation should also be valid.
3.3 OPERATING PROCEDURE OF A KNOWN SIGMA VQSS

Suppose that the standard deviation of the normal distribution is
known. Then, the operating procedure of VQSS with double specification limits is
shown below.
and compute v and

X L
Sample of size n, say X 1 , X 2 ...X n 1
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
inspection as in Step 3, otherwise continue the normal inspection.
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
inspection, then immediately switch to normal inspection as in Step 1.
Thus, the VQSS with double specification limits is characterized by

five parameters, namely (n , k1T , k1N , k2T and k2 N ) . If k1T k1N k1 (say ) and
k2T k2 N k 2 ( say ) , then the proposed system will reduce to the variables single
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
nonconforming case. If there is only one specification limit either L or U as in

Balamurali and Usha ( 2012a), then the proposed system can be used with L=-
( v1 ) or U= ( v2 ). It is to be pointed out that the VQSS with single
specification limit presented in Chapter 2 is a special case of the VQSS with
double specification limits provided in this Chapter.
3.4 OC FUNCTION OF A KNOWN SIGMA VQSS WITH DOUBLE

SPECIFICATION LIMITS
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 ( p )
Pa ( p ) (3.1)
1 PN ( p) PT ( p )
Under double specification limits, the above can be written as
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
where PN P(v1 k1N , v 2 k 2 N ) is the probability of accepting a lot under normal
inspection based on a sample with parameters n , k1N , k2 N and
PT P (v1 k1T , v 2 k 2T ) is the probability of accepting a lot under tightened single
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.
The distributions of v1 and v2 follows normal distributions with the same

variance and means ( L) / and (U ) / respectively. The fraction non-
conforming p for the quality characteristic with double specification limits will be
defined by
p P{ X L} P{ X U } (3.3)
where X denotes the individual measurement of the quality characteristic under

study . When the fraction non-conforming beyond the lower limit and that beyond
the upper limit are specified separately, the means for v1 and v2 can be determined.
3.4.1 KNOWN SIGMA VQSS WITH SYMMETRIC FRACTION NON-

CONFORMING
In the case of symmetric fraction non-conforming, we first assume that

P{ X L} P{ X U } p / 2 (3.4)
Then, for the specified fraction non-conforming p,
U
zp/2 (3.5)

57
L
z1 p / 2 z p / 2 (3.6)

where z is the standard normal variate corresponding to the tail probability of .
In the case of symmetric fraction non-conforming, we have

k1N k 2 N k N and k1T k 2T kT . In this case, the design parameters will be
only n , k N and kT . Therefore, the probability of acceptance under tightened

inspection PT for the specified p reduces to

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 () is the cumulative standard normal distribution function and kT z p /2 .
Similarly, the probability of acceptance under normal inspection PN for

the specified p will be determined by
PN 2[1 ((z p / 2 k N ) n )] (3.8)
where k N z p /2 .
Hence, the probability of acceptance of the proposed VQSS under double

specification limits is given by
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
As mentioned earlier, the above OC function reduces to the OC

function of the variables single sampling plan with double specification limits
when kT k N k . Hence when kT k N k (3.9) becomes
58
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.
3.4.2 KNOWN SIGMA VQSS WITH ASYMMETRIC FRACTION

NON-CONFORMING
Generally, in some situations, the fraction non-conforming below the

lower specification limit and above the upper specification limit are different. In
such kind of situations, it will be assumed that
P{ X L} p L and P{ X U } pU such that ( p L pU p ) (3.11)
U
Then, z p (3.12)
U

L
and z1 pL z pL (3.13)

Therefore, the probability of acceptance under tightened inspection for the

specified p reduces to

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)
Similarly, the probability of acceptance based on normal inspection for the

specified p becomes
PN 1 (( z pU k 2 N ) n ) ((k1N z pL ) n ) (3.15)
It should be noted that the condition of k 2 N k 2T z p U

and k1N k1T z p are
L
needed for (3.14) and (3.15).

59
Therefore, the probability of acceptance of a lot under VQSS with double

specification limits is obtained by
(( z pU k 2T ) n ) ((k1T z pL ) n )
Pa ( p )
1 [ (( z pU k 2 N ) n ) ((k1N z pL ) n ] [ (( z pU k 2T ) n ) ((k1T z pL ) n )]
(3.16)
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.
3.5 DESIGNING OF UNKNOWN SIGMA VQSS HAVING DOUBLE

SPECIFICATION LIMITS
The standard deviation of a normal distribution is unknown in some of

the practical applications. In such cases, we should use the estimated standard
deviation from the sample. So, the unknown sigma VQSS with double
specification limits is operated as follows.
sample of size n, say X 1 , X 2 ... X n and compute v1

X L and
S
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
inspection as in Step 3, otherwise continue the normal inspection.

X L and v U X ,
X 1 , X 2 ... X n and compute v1 2 where
S S
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
inspection, then immediately switch to normal inspection as in Step 1.
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
PX kTS S U p PX kTS S L p (3.18)

61
If we use the result that X kS (for a constant k) is approximately distributed as

normal as follows (see Duncan (1986)):
2
2
k 2
X kS ~ N k , (3.19)
n 2n
Then the equation (3.18) reduces to
2n
PT 2 ( z p / 2 kTS ) 1 (3.20)
2 kTS
2

Similarly, the probability of accepting a lot under normal inspection can be

obtained similarly as follows:
2n
PN 2 2 ( z p / 2 k NS ) (3.21)
2 k NS
2

The condition of k NS kTS z p / 2 is needed in (3.20) and (3.21). Finally,
the probability of acceptance of a lot for the specified fraction non-conforming, p

will be obtained by using the equation (3.1), where PT and PN are given in (3.20)
and (3.21) respectively. As in the known sigma case, it reduces to the unknown
sigma single sampling plan with double specification limits when kTS k NS k S .
For asymmetric fraction non-conforming case, similar procedure can be adopted
for designing of unknown sigma VQSS.
3.6 DETERMINATION OF THE OPTIMAL PARAMETERS OF VQSS
The optimal parameters of the VQSS can be determined by using the

two-points on the OC curve approach that considers both the producers risk
and the consumers risk along with the corresponding quality levels. When the
product quality is at AQL, the probability of acceptance of a lot should be more
62
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)
We may evaluate the ASN at p1 or at p 2 , since ASN for VQSS is a

constant nothing but the sample size only, irrespective of the quality levels. Hence,
the optimization problem for determining the optimal parameters for the known
sigma asymmetric case is given by
Minimize ASN ( p )
Subject to
Pa ( p1 ) 1
Pa ( p2 )
n 1 , 0 k2 N k2T min( z pU , z pU )
1 2
0 k1N k1T min( z pL1 , z pL )

2
(3.24)
To solve the above nonlinear optimization problem, the sequential

quadratic programming (SQP) proposed by Nocedal and Wright (1999) can be
63
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.
3.7 ILLUSTRATIVE EXAMPLES
3.7.1 SYMMETRIC FRACTION NONCONFORMING CASE
Suppose a quality characteristic of interest follows a normal distribution

with =10 and double specification limits of L=75 and U=125. The inspector
wishes to adopt a VQSS, where AQL at =0.05 and LQL at =0.1 are specified
by p1 =0.01, p 2 =0.05, respectively. Then, Table 3.1 gives the parameters as n =4,
k N 1.559 and kT 1.932 .
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.
3.7.2 ASYMMETRIC FRACTION NONCONFORMING CASE
Suppose one wants to determine the optimal parameters of VQSS when

the quality characteristic of interest follows a normal distribution with =15 and
double specification limits of L=75 and U=125 for the specified AQL and LQL
conditions, Table 3.2 can be used. For example, the AQL and LQL are specified as
64
p1 =0.01 and p 2 =0.05 at =0.05 and at =0.1 respectively. Then, Table 3.2 gives
the optimal parameters as n =4, k1N 1.759 , k 2 N 1.443 , k1T 2.168
and k 2T 1.802 .
For the above example, the system operates as follows.
sample of size 4. Now, compute v1

X L and v U X , where
2

1 4
X Xi .
4 i 1
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.
3.8 INDUSTRIAL APPLICATION OF THE PROPOSED VQSS

In order to apply the proposed VQSS with double specification limits in
real-life situations, we consider an industrial case study example as provided by
Wu and Pearn (2008).
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.
0.642 0.720 0.726 0.684 0.727 0.632 0.657

0.630 0.748 0.635 0.688 0.665 0.710 0.633
0.712 0.633 0.712 0.733 0.739 0.700 0.699
0.640 0.645 0.731 0.669 0.659 0.641
0.632 0.651 0.658 0.768 0.656 0.712
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 2: Even though v 2 2.14657 k NS 1.329 but, v1 1.24622 k NS 1.329 , the

lot is rejected. Since the lot is rejected on normal inspection, then immediately
switch to tightened inspection as in Step 3, for the next lot.
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
Step 2: Since v 2 2.00547 k TS 1.632 but, v1 1.18614 k TS 1.632 , the lot is

rejected. Because the lot is rejected on tightened inspection also, the same
tightened inspection will be done for the next lot.
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.
3.10 NON-NORMALITY IN VQSS

The variables sampling system developed in this chapter is based on the
assumption that the quality characteristic of interest follows a normal distribution.
Whenever the normality assumption is not true, using of any variables sampling
plans can be quite misleading (refer Sahli et al. (1997)). However it is to be
69
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
Table 3.1. Optimal Parameters of Known Sigma VQSS with Double

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
0.0025 0.005 34 2.803 2.682

0.010 9 2.548 2.313
0.015 5 2.401 2.119
0.020 3 2.289 1.848
0.025 3 2.190 1.862
0.030 2 2.132 1.579
0.050 1 1.918 0.953
0.005 0.010 30 2.563 2.439

0.015 12 2.408 2.218
0.020 7 2.303 2.043
0.030 4 2.139 1.802
0.040 3 2.024 1.642
0.050 2 1.918 1.368
0.10 1 1.591 0.761
0.010 0.02 25 2.312 2.182
0.03 10 2.149 1.943
0.04 6 2.028 1.748
0.05 4 1.932 1.559
0.10 2 1.603 1.153
0.15 1 1.389 1.002
0.20 1 1.221 0.123
0.03 0.060 21 1.872 1.723

0.090 7 1.669 1.402
0.120 4 1.528 1.149
72
Table 3.1 Contd.
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
0.05 0.200 4 1.249 0.963

0.250 1 1.088 0.072
0.500 1 0.591 0.032
73
Table 3.2 Optimal Parameters of Known Sigma VQSS with Double

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
0.0025 0.005 35 3.001 2.674 2.884 2.563

0.010 8 2.758 2.428 2.498 2.198
0.015 5 2.612 2.291 2.312 2.012
0.020 3 2.498 2.179 2.019 1.748
0.025 3 2.414 2.088 2.042 1.763
0.030 2 2.343 2.022 1.749 1.488
0.050 1 2.128 1.809 1.093 0.893
0.005 0.010 30 2.782 2.432 2.653 2.312

0.015 11 2.629 2.289 2.399 2.069
0.020 7 2.521 2.178 2.242 1.918
0.030 4 2.358 2.011 1.993 1.682
0.040 3 2.239 1.889 1.828 1.533
0.050 2 2.142 1.798 1.543 1.269
0.10 1 1.819 1.477 0.902 0.689
0.01 0.02 26 2.553 2.169 2.399 2.039

0.03 10 2.379 2.012 2.152 1.812
0.04 6 2.261 1.889 1.958 1.629
0.05 4 2.168 1.802 1.759 1.443
0.10 2 1.839 1.468 1.342 1.039
0.15 1 1.632 1.262 0.668 0.448
0.20 1 1.456 1.089 0.713 0.469
0.03 0.060 18 2.134 1.712 1.939 1.542

0.090 7 1.939 1.509 1.642 1.258
0.120 4 1.789 1.368 1.373 1.019
74
Table 3.2 Contd.
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
0.05 0.100 15 1.913 1.458 1.692 1.269

0.150 5 1.698 1.253 1.283 0.903
0.200 3 1.542 1.079 1.002 0.648
0.250 2 1.397 0.949 0.708 0.402
0.500 2 0.862 0.432 0.838 0.409
75
Table 3.3 Optimal Parameters of Unknown Sigma VQSS with Double

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
0.0025 0.005 147 2.788 2.568

0.010 30 2.548 2.147
0.015 16 2.401 1.949
0.020 11 2.278 1.823
0.025 8 2.193 1.649
0.030 6 2.122 1.338
0.005 0.050 4 1.892 1.419

0.010 112 2.562 2.319
0.015 39 2.407 2.059
0.020 22 2.301 1.863
0.030 12 2.129 1.714
0.040 8 2.012 1.548
0.050 6 1.911 1.432
0.010 0.10 3 1.569 1.188

0.02 83 2.312 2.062
0.03 28 2.149 1.767
0.04 16 2.022 1.612
0.05 11 1.918 1.489
0.10 4 1.576 1.088
0.15 3 1.358 1.067
0.20 2 1.188 0.849
76
Table 3.3 Contd.
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
0.05 0.100 33 1.632 1.329

0.150 10 1.408 0.879
0.200 6 1.244 0.853
0.250 4 1.103 0.690
0.500 2 0.591 0.482
77
Table 3.4 Optimal Parameters of Unknown Sigma VQSS with Double

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
0.0025 0.005 162 2.989 2.678 2.864 2.583

0.010 30 2.738 2.452 2.469 2.229
0.015 16 2.584 2.291 2.242 2.028
0.020 10 2.469 2.193 2.098 1.943
0.025 8 2.384 2.108 1.958 1.768
0.030 7 2.303 2.029 1.828 1.648
0.050 4 2.084 1.809 1.469 1.313
0.005 0.010 126 2.768 2.442 2.623 2.332

0.015 43 2.609 2.229 2.393 2.116
0.020 22 2.521 2.188 2.214 1.968
0.030 12 2.334 2.034 1.979 1.762
0.040 9 2.219 1.913 1.822 1.613
0.050 6 2.102 1.819 1.687 1.489
0.10 3 1.769 1.477 1.109 0.958
0.01 0.02 90 2.533 2.192 2.369 2.069

0.03 31 2.362 2.028 2.108 1.839
0.04 16 2.229 1.912 1.923 1.669
0.05 12 2.128 1.808 1.779 1.553
0.10 5 1.789 1.488 1.359 1.149
0.15 3 1.572 1.268 0.978 0.812
0.20 2 1.388 1.113 1.018 0.818
0.03 0.060 48 2.113 1.722 1.919 1.594

0.090 16 1.913 1.532 1.603 1.308
78
Table 3.4 Contd.
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
0.05 0.200 6 1.478 1.112 1.132 0.848

0.250 4 1.332 0.969 0.957 0.702
0.500 2 0.786 0.468 0.658 0.369
79
Table 3.5. Average Sample Number of Variables Single Sampling Plans and
VQSS with Double Specification Limits (Symmetric Fraction Non-
conforming)
Average Sample Number

p1 p2 Known Sigma Unknown Sigma
SSP VQSS SSP VQSS
0.001 0.002 84 42 481 200
0.003 33 17 173 72
0.004 20 10 101 42
0.006 12 6 54 23
0.008 9 4 37 15
0.010 7 3 28 12
0.015 5 2 18 8
0.020 4 2 13 6
0.0025 0.005 72 34 353 147

0.010 17 9 72 30
0.015 10 5 38 16
0.020 7 3 25 11
0.025 6 3 19 8
0.030 5 2 15 6
0.050 3 1 9 4
0.005 0.010 63 30 270 112

0.015 24 12 94 39
0.020 15 7 53 22
0.030 9 4 28 12
0.040 6 3 18 8
0.050 5 2 14 6
0.10 3 1 6 3
0.010 0.02 55 25 199 83

0.03 21 10 68 28
0.04 13 6 38 16
0.05 9 4 26 11
0.10 4 2 9 4
0.15 3 1 6 3
0.20 3 1 4 2
80
Table 3.5 Contd.
Average Sample Number

p1 p2 Known Sigma Unknown Sigma
SSP VQSS SSP VQSS
0.03 0.060 41 21 110 46
0.090 15 7 36 15
0.120 9 4 20 8
0.150 7 3 13 5
0.300 3 1 4 2
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
OPTIMAL DESIGNING OF VARIABLES

TIGHTENED-NORMAL-TIGHTENED SAMPLING SCHEME BY
MINIMZING THE AVERGE SAMPLE NUMBER
4.1 INTRODUCTION
The Tightened-Normal-Tightened (TNT) sampling scheme developed

by Calvin (1977), is a particular case of the general two-plan system for the
inspection of attributes characteristics. Balamurali and Jun (2009) developed a
designing methodology to determine the parameters of TNT sampling scheme
under variables sampling. Recently, Senthilkumar and Muthuraj (2010) provided
procedures for selecting parameters of the variables TNT scheme and constructed
tables for selecting parameters of the variables TNT scheme of type TNT (n1, n2;
k). However they didnt follow any optimization techniques. So, this chapter
attempts to design a variables TNT (n1, n2; k) scheme by minimizing average
sample number (ASN) as done by Balamurali and Jun (2009). Obviously, any
sampling plan having smaller ASN would be more desirable. For the selection of
the parameters of the variables TNT scheme, the problem is formulated by a
nonlinear programming where the objective function to be minimized is the ASN
and the constraints are related to lot acceptance probabilities at acceptable quality
level (AQL) and LQL.

In order to apply the variables TNT sampling scheme, the following
assumptions should be valid.
(i) Production is in a steady state, so that results of past, present and
future lots are broadly indicative of a continuing process.
82
(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.
plan with known or unknown standard deviation should also be valid (see for
further details Schilling (1982), Grant and Leavenworth (1996) and Montgomery
(2005)).
4.3 OPERATING PROCEDURE OF KNOWN SIGMA VARIABLES TNT

SCHEME
Suppose that the quality characteristic of interest has the upper specification limit
U and follows a normal distribution with known standard deviation .
variables plan with a sample size n1 and the acceptance criterion k.
Accept the lot if v k and reject the lot if v k , where v

U X and
1

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.
variables plan with a sample size n2 (<n1) and the acceptance criterion k.
83
Accept the lot if v k and reject the lot if v k , where v

U X and
2

n2
1
X2
n 2
X
i 1
i .
Step 4: Switch to tightened inspection after a rejection of lot if an additional lot is

rejected in the next s lots.
Thus, the variable TNT scheme has five parameters namely the
tightened plan sample size n1, the normal plan sample size n2, the acceptance
criterion k and the switching parameters s and t. In the attributes case, the normal
plan sample size is taken as n (=n2) and the tightened plan sample size is
considered as mn (=n1) and m>1. We have followed the similar way in this
chapter. In this chapter, the value of m is considered in the interval 1.25(0.25)10.00
for constructing tables. Also the switching parameters are fixed as 4 and 5
respectively since, when s=4 and t=5, the sampling procedures correspond to the
procedures of MIL-STD 105D (1963) scheme involving only normal and tightened
inspections. Also Soundararajan and Vijayaraghavan (1992) observed that when
s=4 and t=5 gives more discriminating OC curve than any other combinations s
and t. It is to be pointed out that the VQSS presented in Chapter 2 is a special case
of the TNT scheme when s=0 and t=1 with reverse switching order. However, the
VQSS given in Chapter 2 has single sample size with two acceptance criteria
where as the TNT scheme proposed in this Chapter involves two sample sizes with
single acceptance criterion.
4.4 OC FUNCTION OF KNOWN SIGMA VARIABLES TNT SCHEME
The OC function of the variables TNT scheme, which gives the

proportion of lots that are expected to be accepted for given product quality, p
under known sigma case is given by
P2 P1
Pa ( p) (4.1)

84
s
2 P2
where s
, s 1
(1 P2 )(1 P2 )
t
1 P1
t
, ts
P1 (1 P1 )
Here P1= Pr(v1 k) is the probability of acceptance under tightened inspection

and P2 = Pr(v2 k) is the probability of acceptance under normal inspection.
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.

U
p 1 1 ( v) ( v) (4.2)

y
1 z2
( y ) exp dz , (4.3)
2 2
provided that the quality characteristic of interest is normally distributed with
mean and standard deviation , and the unit is classified as non-conforming if it
exceeds the upper specification limit U.
Then its probability of acceptance is written as

( w2 ) (w1 )
Pa ( p ) (4.4)

where
2 ( w2 )
s
s
, s 1
(1 ( w2 ))(1 ( w2 ) )
85
1 ( w1 )
t
, ts
( w1 )t (1 ( w1 ))
Here w1 v k mn and w2 v k n where n1 mn and n2 n
4.5 DESIGNING OF A KNOWN SIGMA VARIABLES TNT SCHEME
The OC function of a known sigma variables TNT scheme is given in

(4.4). If two points on the OC curve namely, AQL(=p1), LQL(=p2), the producers
risk and the consumers risk are prescribed then the OC function can be
expressed as
( 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,
w11 (v1 k ) mn , w21 (v1 k ) n
w12 (v 2 k ) mn and w22 (v 2 k ) n
where v1 is the value of v at AQL and v2 is the value of v at LQL.
The ASN, by definition, means the expected number of sampled units

required for making decisions about the lot. The concept of ASN is meaningful
under Type B sampling situations. The average sample number of the TNT
variables scheme is given by
86
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
4.6 DESIGNING OF UNKNOWN SIGMA VARIABLES TNT SCHEME

Whenever the standard deviation is unknown, we should use the sample
standard deviation S instead of . In this case, the operation of the scheme is as
follows.
variables plan with a sample size n1 and the acceptance criterion kS.
Accept the lot if v k S and reject the lot if v k S , where v

U X ,
1
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.
variables plan with a sample size n2 (<n1) and the acceptance criterion kS.
Accept the lot if v k S and reject the lot if v k S , where v

U X ,
2
1 n2 S
(X i X 2 )2
X2
n2S
X i and S
i 1 n2S 1
.
Step 4: Switch to tightened inspection after a rejection of lot if an additional lot is

rejected in the next s lots.
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
unknown standard deviation. Hamaker (1979) has given an approximation for

finding the parameters of the unknown sigma single sampling plan from the
parameters of the known sigma single sampling plan. Senthilkumar and Muthuraj
(2010) have followed the similar approximation for selecting the parameters of
unknown sigma variables TNT scheme. In this chapter, we follow the same
approximation for finding the parameters of unknown sigma variables TNT
scheme. That is, one can determine the sample size for the unknown sigma TNT
scheme as
k 2
ns 1 n (4.9)
2

which is the normal plan sample size and the tightened plan sample size is
determined by multiplying the factor m with the normal plan sample size. The
acceptance criterion of unknown sigma TNT scheme is determined as
(4n 4)
k S k s (4.10)
4n s 5)
Also the ASN of unknown sigma TNT scheme namely ASNs can be determined as
k 2
ASN S ASN 1 s (4.11)
2

One can determine the parameters of the known sigma TNT scheme by
solving the nonlinear equation given in (4.8). There may exist multiple solutions
since there are three unknowns namely n, k and m with only two equations.
Generally a sampling would be desirable if the required number of sampled is
small. So, in this chapter, we consider the ASN as the objective function to be
minimized with the probability of acceptance along with the corresponding
producers and consumers risks as constraints. To solve the above nonlinear
optimization problems given in (4.8), the sequential quadratic programming (SQP)
proposed by Nocedal and wright (1999) can be used. The SQP is implemented in
89
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
Table 4.1 is used to determine the parameters of the known variables

TNT scheme for specified values of AQL and LQL when =5% and =10%. For
example, if p1=1%, p2=9%, =5% and =10%, Table 4.1 gives the parameters as
n = 5, m=5 and k = 1.598. The normal plan sample size is n2=n=5 and the
tightened plan sample size is obtained as n1=mxn= 5x5=25. The acceptance
criterion is same for both normal and tightened plans.
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
lot if v 1.598 . If 2 out of (s+1)=5 lots are rejected on normal inspection

then immediately revert to tightened inspection as in step 1.
4.7.2 SELECTION OF UNKNOWN SIGMA VARIABLES TNT SCHEME

INDEXED BY AQL AND LQL
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.
4.7.3 ADVANTAGES OF THE VARIABLES TNT SAMPLING SCHEME

This section describes the advantages of the variables TNT scheme over
attributes TNT scheme, variables single and double sampling plans. For the
purpose of comparison, we will consider the plans which have the same AQL and
LQL. Suppose that for given values of AQL=0.02, =5%, LQL=0.07 and =
10%, one can find the parameters of the attributes TNT scheme from
Soundararajan and Vijayaraghavan (1992) under the application of Poisson model
as
(i) n1 = 76, n2=38, c = 2 and ASN=38.609
91
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.
(ii) n = 26, k = 1.73 and ASN=26

(iii) n1 = n2 = 19, ka = 1.83, kr = 1.67 and ASN=21.2
(iv) n = 13, k = 1.602, m = 8 and ASN=16.626
By comparing the above, it is clear that the variables TNT scheme

achieves a reduction of over 57% in sample size than the attributes TNT scheme
and about 36% than the variables single sampling plan with same AQL and LQL
conditions. Further, it is also to be pointed that the variables TNT scheme is
economically superior to the variables double sampling plan in terms of ASN. The
proposed scheme also achieves a reduction of 22% in ASN over the variables
double sampling plan. Obviously, a sampling plan having smaller ASN would be
more desirable. The variables double or multiple sampling plans are not practically
very useful. Variables sampling standards avoid presenting such plans due to
increased complexity involved in operating them.
4.8 COMPARISONS
4.8.1 COMPARISON THORUGH OC CURVES
In order to show the better efficiency of the variables TNT scheme,

three OC curves are considered. Figure 4.1. shows the OC curves of the variables
TNT scheme with parameters n = 12, m = 5.25 and k = 1.857 along with two
variables single sampling plans (n =63, k =1.857) and (n = 12, k =1.857). The
variables TNT scheme is selected in such a way that it satisfies the two-points on
the OC curve condition (p1=0.01, 1- = 0.95) and (p2 = 0.045, = 0.10). From this
92
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.
4.8.2 COMPARISON THROUGH ASN

Table 4.2 shows the ASN values of the variables single sampling plan
and the variables double sampling plan along with the two variables TNT schemes
(one given by Senthilkumar and Muthuraj (2010) and the other given in Table 4.1)
for some arbitrarily selected combinations of AQL and LQL under known sigma
case. Table 4.3 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). To strengthen this point, two more tables are provided in this chapter.
Table 4.4 gives the parameters of known sigma variables TNT scheme along with
parameters given by Senthilkumar and Muthuraj (2010) and Table 4.5 gives the
parameters of unknown sigma variables TNT scheme.
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
Table 4.1. Variables Tightened-Normal-Tightened Sampling Scheme Indexed

by AQL and LQL for =5% and =10% Involving Minimum ASN
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
0.01 0.02 56 10.0 2.108 75.757 180 10.0 2.111 244.553

0.03 22 7.75 1.979 27.914 65 7.75 1.987 83.005
0.04 13 8.5 1.873 16.809 36 8.5 1.886 46.720
0.05 9 9.5 1.784 12.067 23 9.5 1.805 31.713
0.06 7 9.5 1.712 9.399 17 9.5 1.739 21.613
0.07 6 8.0 1.661 7.668 14 8.0 1.694 18.663
0.08 5 8.75 1.599 6.557 11 8.75 1.640 15.374
0.09 5 5.0 1.598 5.798 11 5.0 1.639 13.584
0.10 4 7.75 1.512 5.076 9 7.75 1.561 11.258
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
Table 4.1. Contd.
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
Table 4.1. Contd.
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
Variables Normal SSP

0.8
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 )
0.6 Variables TNT Scheme
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
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)
Table 4.3: ASN Values of the Unknown Sigma Variables Single Sampling
Plan, Variables Double Sampling Plan and Variables TNT Scheme
ASN
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)
99
Table 4.4: Parameters of Known Sigma Variables TNT Scheme for different
AQL and LQL Values
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 from Senthilkumar and Muthuraj (2010)

Table 4.5: Parameters of Unknown Sigma Variables TNT Scheme for

different AQL and LQL Values
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
*Parameters from Senthilkumar and Muthuraj (2010)

100
CHAPTER 5
VARIABLES CHAIN SAMPLING PLAN BY MINIMIZING THE
AVERAGE SAMPLE NUMBER
5.1 INTRODUCTION
Some examples of the use of cumulative results to achieve a reduction

of the size while maintaining or even extending protection can be found in chain
sampling plans introduced by Dodge (1955). These plans were finally conceived to
overcome the problem of lack of discrimination in c = 0 sampling plans. The
procedure was developed to chain together the most recent inspections in a way
that would build up the shoulder of the OC curve of c = 0 plans. This is especially
desirable in situations in which small samples are demanded because of the
economic or physical difficulty of obtaining a sample. The chain sampling plan is
one of the conditional sampling procedures and this plan under variables
inspection will be useful when testing is costly and destructive. Govindaraju and
Balamurali (1998) extended the concept of chain sampling to variables inspection
but they have not provided any tables for the selection of parameters and also they
have dealt only with known standard deviation case. This chapter attempts to
provide tables for the easy industrial application of this plan for both known and
unknown standard deviation cases. It is also to be pointed out that Govindaraju and
Balamurali (1998) have given an approximate solution for finding the plans of
unknown standard deviation case. But in this chapter, we provide a different
procedure for finding the unknown sigma chain sampling plan. The major
advantage of this plan is to achieve better protection to the producer with
minimum ASN.
101

variables chain sampling plan.
(i) Lots are submitted for inspection serially, in the order of production from a
process that turns out a constant proportion non-conforming items.
(ii) The consumer has confidence in the supplier and there should be no reason to
believe that a particular lot is poorer than the preceding lots.
In addition, the usual conditions for the application of single sampling variables
5.3 OPERATING PROCEDURES OF VARIABLES ChSP
5.3.1 KNOWN SIGMA CASE

specification limit U and follows a normal distribution with unknown mean and
known standard deviation . Then the operating procedure of the variables chain
sampling plan is proposed as follows.
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
In the case of lower specification limit L, the operating procedure is described as

follows.
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.
Thus, the proposed variables chain sampling plan is characterized by

four parameters, namely n , i , k and k ' . If k k ' , then the proposed plan will
reduce to the variables SSP. Also, when i tends to infinity, the proposed plan
becomes variables SSP with parameters n and k . It is to be pointed out that the
chain sampling plan can be applied for inspection of lots which are submitted
serially in the order of production or in the order of being submitted. The decision
of current lot depends on the results of preceding lots. So, when i 2, we need to
keep the records of results of previous lots. Of course, maintaining records of
preceding lots may be a drawback of the chain sampling plan over the single
sampling plan, however this can be compensated by minimizing the inspection
efforts in terms of minimum ASN with desired protection.
5.3.2 UNKNOWN SIGMA CASE

Whenever the standard deviation is unknown, we may use the sample
standard deviation S instead of . In this case, the plan operates as follows.
103
Step 1: From each submitted lot, take a random sample of size n S , say
X , X ...X and compute v U S X , where X n1 X

nS
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.
5.4 DESIGNING METHODOLOGY OF THE VARIABLES ChSP
5.4.1 KNOWN SIGMA CASE

Generally, variables sampling plans are designed based on two points
on the OC curve namely ( p1 ,1 ) and ( p 2 , ) , where p1 is called the AQL, p 2 is
the LQL, is the producers risk and is the consumers risk. Any well
designed sampling plan which must provide at least (1 )% probability of
acceptance of a lot when the process fraction nonconforming is at AQL level and
the sampling plan must also provide not more than % probability of acceptance
if the process fraction nonconforming is at the LQL level. Thus the acceptance
sampling plan must have its OC curve passing through two designated points
(AQL, 1 ) and (LQL, ). Some other strategies are also followed to design the
sampling plans besides the statistical based paradigm, which include Bayesian
104
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.
The fraction non-conforming in a lot is given as

U
p 1 (5.1)

where (Y ) is the cumulative distribution function of standard normal
distribution and is given by
Y
1 Z2
(Y ) exp dZ , (5.2)
2 2
Here the quality characteristic of interest is normally distributed with

mean and standard deviation , and the unit is classified as non-conforming if it
exceeds the upper specification limit U. So, the unknown mean can be
determined if p is specified. Let us define the standardized quality characteristic
corresponding to the fraction conforming as
Z p 1 (1 p) (5.3)
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)
where w1 Z p k ' n and w2 Z p k n
The OC function given in (5.5) under the specified AQL and LQL conditions can
be written as
(w21 ) ( w11 ) ( w21 ) ( w21 ) 1

i
(w22 ) ( w12 ) (w22 ) (w22 )

i
and
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 p 2 and w22 is the value of w2 at p p 2 . That is,
w11 ( Z p 1 k ' ) n , w21 ( Z p 1 k ) n
w12 ( Z p 2 k ' ) n and w22 ( Z p 2 k ) n
where Z p1 is the value of Z p at AQL and Z p 2 is the value of Z p at LQL.

106
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)
We may determine the parameters of the known sigma chain sampling

plan by solving the nonlinear equation given in (5.6).
5.4.2 UNKNOWN SIGMA CASE

The determination of parameters ( nS , k ' S , k S ) of unknown sigma
variables chain sampling plan is explained as follows. It is known that for large
samples, X k S S is approximately normally distributed with mean k S E (S )
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
Therefore, the probability of accepting a lot at each repetition is given by

P (v k S ) P X U k S S p

P X kS S U p
107

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)
where w1S Z p k ' S n S and w2 S Z p k S nS
If (AQL, 1 ) and (LQL, ) are prescribed then the OC function can be written
as
(w21S ) ( w11S ) ( w21S ) ( w21S ) S 1
i
(w22 S ) ( w12 S ) (w22 S ) (w22 S ) S

i
and (5.11)
108
Here w11S is the value of w1S at p p1 , w21S is the value of w2 S at p p1 , w12 S is

the value of w1S at p p 2 and w22 S is the value of w2 S at p p 2 .
We obtain w11S , w21S , w12 S , w22 S respectively by

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
Therefore, the following optimization problem is considered to determine those

parameters.
Minimize ASN ( p1 ) n S
Subject to Pa ( p1 ) 1
Pa ( p 2 )
nS 2, i S 1, k S k ' S 0, 0, n S N, i S N (5.13)
We may determine the parameters of the unknown sigma chain

sampling plan by solving the nonlinear equation given in (5.13). There may exist
multiple solutions since there are four unknowns with only two equations.
Generally a sampling would be desirable if the required number of sampled is
small. So, in this chapter, we consider the ASN as the objective function to be
minimized with the probability of acceptance along with the corresponding
producers and consumers risks as constraints. To solve the above nonlinear
optimization problems given in (5.6) and (5.12), the sequential quadratic
programming (SQP) proposed by Nocedal and Wright (1999) can be used. The
SQP is implemented in Matlab software using the routine fmincon. By solving
the nonlinear problem mentioned above, the optimal parameters ( n , i , k and
k ' ) for known sigma plan and the parameters ( nS , i S , k S and k' S ) for unknown
sigma plan are determined and these values are tabulated in Table 5.1.
5.5 DESIGNING EXAMPLES

5.5.1.SELECTION OF KNOWN SIGMA VARIABLES CHAIN SAMPLING
PLAN FOR SPECIFIED AQL AND LQL
Table 5.1 is used to determine the parameters of the known variables

chain sampling plan for specified values of AQL and LQL when = 5% and
110
= 10%. For example, if p1 = 2%, p 2 = 7%, = 5% and = 10%, Table 5.1

gives the parameters as n = 18, i =3, k ' = 1.544 and k = 1.779.
For the above example, the plan is operated as follows.
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.
5.5.2. SELECTION OF UNKNOWN SIGMA VARIABLES ChSP FOR

SPECIFIED AQL AND LQL
As mentioned earlier, the unknown sigma variables chain sampling plan

is operated as a known sigma variables chain sampling plan but the parameters
nS , i S , k S and k' S are used in the place of n , i , k and k ' respectively. Table 5.1
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.
5.6 ILLUSTRATIVE EXAMPLE
To illustrate the implementation of the proposed sampling plan for the

example given in section 5.5.2, we consider a case study data on STN-LCD
manufacturing process given in Wu and Pearn (2008). The upper specification
111
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
0.740 0.706 0.726 0.688 0.715 0.704 0.724
The implementation of the plan is shown below.
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
Step 4: Since v 3.4741 k S 2.190 , we accept the current lot without

considering the result of past lots.
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
Further it is to be pointed out that the proposed variables chain

sampling plan is more efficient in terms of minimum ASN than the variables SSP
for low values of producers risk () and consumers risk (). In order to prove
this, we provide two tables. Table 5.2 gives the optimal parameters of variables
chain sampling plan for some selected combinations of AQL and LQL and for
=1% and =1% and Table 5.3 gives the parameters of variables SSP under same
set of conditions. By comparing these two tables, one can easily observe that
variable chain sampling plan involves minimum ASN compared to the variables
SSP.
5.7 ADVANTAGES OF THE VARIABLES ChSP
This section describes the advantages of the variables chain sampling

plan over the conventional variables single sampling plan. Two acceptance
sampling plans will be called equivalent when they possess nearly identical OC
curves. A customary procedure for achieving such equivalency consists of
constructing the sampling plans so that their OC curves coincide in two suitably
chosen points namely ( p1 ,1 ) and ( p 2 , ) . Suppose that for given values of
p1 =0.5%, =5%, p 2 =1.5% and =10%, one can find the parameters of the
known sigma variables chain sampling plan from Table 5.1 as
(i) n = 33, i =1, k ' = 2.211 and k = 2.421; ASN = 33
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
(ii) n = 53 and k = 2.35; ASN=53

(iii) n = 39, k a = 2.41 and k r = 2.31:ASN=43
113
It can be observed that variables chain sampling plan achieves a

reduction of over 38% in sample size than the variables SSSP and a reduction of
23% over DSP with same AQL and LQL conditions. It indicates that the variables
chain sampling plan achieves same OC curve with minimum sample size
compared to the variables single and double sampling plans.
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.
Further, it is also to be noted that the variables chain sampling plan is

economically superior to the double sampling plans in terms of average sample
number (ASN). Obviously, a sampling plan having smaller ASN would be more
desirable. The variables double or multiple sampling plans are not practically very
useful. Variables sampling Standards avoid presenting such plans due to increased
complexity involved in operating them, but the variables chain sampling plan has
no such complexity. Table 5.4 shows the ASNs for variables single sampling
plan, variables double sampling plan along with variables chain sampling plan for
some arbitrarily selected combinations of AQL and LQL. These ASN values are
calculated at the producers quality level for the known sigma plans. The ASN of
the variables single and double sampling plans can be found in Sommers (1981).
114
5.8 AVERAGE RUN LENGTH OF VARIABLES ChSP
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)
Its mean and variance are respectively given by
1
ARL E ( L) (5.18)
1 Pa ( p)
Pa ( p)
Var ( L) (5.19)
1 Pa ( p) 2
where Pa(p) is given in (5.5).
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
with the single sampling plan ( i = ), at 2% nonconforming only 80% of lots

will be accepted at an average rate of 5 inspections. Also, 90% of the lots will be
rejected at the fraction nonconforming 7% by the variables chain sampling plan at
an average rate of 1.11 inspections and at the fraction nonconforming, 93% of the
lots will be rejected by the single sampling plan at the rate of 1.08 inspections.
5.9 CONCLUSIONS
The purpose of this chapter is to develop conditional sampling

procedures for the inspection of normally distributed quality characteristics.
Variables sampling plans generally require a smaller sample size than do
attributes plans. If the OC curve of the variables plan is unsatisfactory, then its
shape can be improved by chaining the past lot results. The proposed variables
chain sampling plan is one of the variables conditional sampling plans which also
ensure the protection against the consumer point of view. This plan is also simple
to apply rather than double and multiple sampling variables plans. Also this plan
provides better protection than the conventional single and double sampling
variables plans with minimum sample size. Such a variables chain sampling plan
will be effective and useful for compliance testing. However, it is also to be
pointed out that the variables chain sampling plan developed in this chapter is
based on the assumption that the quality characteristic of interest follows a normal
distribution. Whenever the normality assumption is not true or invalid, using of
this variables chain sampling plan can be quite misleading.
116
Table 5.1.Variables Chain Sampling Plans Indexed by AQL and LQL

for =5% and =10%

p1 p2 n i k ' k nS iS k' S kS
0.001 0.0020 119 1 2.883 3.013 650 1 2.888 3.013
0.0025 68 2 2.810 2.965 361 1 2.834 2.984
0.003 46 1 2.776 2.961 242 1 2.774 2.964
0.004 31 2 2.740 2.885 147 1 2.716 2.921
0.005 21 2 2.610 2.860 103 1 2.637 2.897
0.006 19 4 2.597 2.807 84 2 2.599 2.829
0.007 15 3 2.534 2.789 72 3 2.532 2.792
0.008 13 3 2.495 2.765 62 3 2.518 2.768
0.009 12 3 2.521 2.736 51 2 2.470 2.765
0.010 12 6 2.517 2.697 45 2 2.425 2.750
0.012 9 4 2.331 2.686 40 3 2.414 2.694
0.015 7 3 2.237 2.657 34 4 2.387 2.637
0.0025 0.004 230 2 2.658 2.738 1131 2 2.676 2.736

0.005 105 2 2.598 2.703 474 2 2.585 2.705
0.006 64 2 2.530 2.675 282 1 2.533 2.698
0.0075 39 2 2.441 2.641 171 1 2.464 2.669
0.010 26 3 2.379 2.579 106 1 2.430 2.610
0.012 19 2 2.335 2.555 76 1 2.321 2.601
0.015 16 4 2.296 2.491 55 1 2.237 2.572
0.020 12 4 2.260 2.425 38 1 2.139 2.529
0.025 10 6 2.206 2.366 32 2 2.163 2.413
0.030 8 5 2.114 2.334 29 5 2.107 2.342
0.035 7 6 2.307 2.297 25 5 2.092 2.302
0.005 0.0075 259 1 2.434 2.524 1113 2 2.442 2.512

0.010 86 1 2.326 2.486 347 1 2.348 2.483
0.012 53 1 2.277 2.457 211 1 2.299 2.454
0.015 33 1 2.211 2.421 128 1 2.233 2.418
0.020 24 4 2.161 2.316 75 2 2.100 2.345
0.025 16 3 2.046 2.281 55 2 2.083 2.293
0.030 14 6 1.989 2.224 43 3 1.974 2.249
0.035 12 6 2.002 2.182 36 3 1.979 2.209
0.040 10 4 1.976 2.156 31 3 1.964 2.174
0.050 8 5 1.893 2.098 25 4 1.925 2.105
117
Table 5.1. Contd.

0.0075 0.010 505 2 2.339 2.384 1817 1 2.329 2.394
0.012 173 1 2.259 2.369 690 2 2.278 2.353
0.015 78 1 2.185 2.335 286 1 2.182 2.337
0.020 43 2 2.146 2.251 133 1 2.083 2.293
0.025 29 5 2.008 2.198 83 1 2.003 2.258
0.030 19 2 1.968 2.178 61 1 1.984 2.214
0.035 17 5 1.888 2.123 47 1 1.925 2.190
0.040 13 3 1.837 2.107 41 3 1.851 2.111
0.010 0.015 248 1 2.212 2.257 786 1 2.187 2.267

0.020 74 1 2.090 2.220 247 1 2.071 2.226
0.025 42 1 2.027 2.177 135 1 2.011 2.186
0.030 27 1 1.906 2.161 93 2 1.954 2.124
0.035 21 2 1.846 2.096 66 1 1.900 2.125
0.040 18 3 1.804 2.054 53 2 1.846 2.066
0.045 15 3 1.762 2.027 46 3 1.811 2.026
0.050 13 3 1.731 2.001 39 2 1.824 2.004
0.060 11 5 1.662 2.942 30 3 1.730 1.955
0.070 9 5 1.573 1.903 26 4 1.728 1.898
0.02 0.030 187 2 1.896 1.976 528 1 1.886 1.991

0.035 96 2 1.835 1.945 272 2 1.831 1.946
0.040 59 1 1.779 1.939 169 2 1.770 1.920
0.045 45 2 1.749 1.889 117 1 1.734 1.919
0.050 34 2 1.693 1.868 88 1 1.683 1.903
0.060 22 2 1.563 1.833 63 2 1.670 1.820
0.070 18 3 1.544 1.779 44 2 1.577 1.792
0.080 14 3 1.434 1.749 36 3 1.525 1.745
0.090 13 4 1.502 1.697 27 2 1.423 1.737
0.10 10 2 1.436 1.691 22 1 1.386 1.766
0.11 10 4 1.457 1.632 23 4 1.476 1.636
0.12 7 1 1.235 1.725 17 1 1.379 1.699
0.13 8 5 1.360 1.580 17 4 1.356 1.596
118
Table 5.1. Contd.

0.030 0.040 341 2 1.771 1.821 1027 5 1.771 1.816
0.045 157 1 1.711 1.811 416 2 1.704 1.799
0.050 96 1 1.650 1.795 253 1 1.679 1.789
0.060 52 2 1.566 1.736 126 1 1.585 1.760
0.070 33 1 1.508 1.728 81 1 1.539 1.724
0.080 26 3 1.427 1.657 62 2 1.514 1.659
0.090 23 5 1.449 1.609 44 2 1.401 1.641
0.100 17 3 1.378 1.593 36 2 1.384 1.609
0.110 13 2 1.268 1.588 29 2 1.305 1.590
0.120 11 1 1.270 1.605 26 3 1.274 1.549
0.130 11 4 1.204 1.514 23 3 1.289 1.519
0.150 10 6 1.273 1.443 18 3 1.230 1.475
0.04 0.060 145 2 1.568 1.663 339 2 1.564 1.664

0.070 73 1 1.513 1.643 166 1 1.507 1.647
0.080 50 2 1.478 1.588 102 1 1.425 1.625
0.090 34 2 1.394 1.564 71 1 1.374 1.600
0.100 27 3 1.324 1.529 53 1 1.315 1.580
0.110 24 5 1.309 1.489 42 1 1.294 1.554
0.120 18 3 1.233 1.478 34 1 1.234 1.539
0.130 14 2 1.130 1.475 31 2 1.271 1.461
0.140 14 4 1.158 1.423 27 3 1.193 1.433
0.150 11 2 1.108 1.428 23 3 1.130 1.415
0.170 10 4 1.070 1.360 17 2 1.061 1.396
0.050 0.070 188 1 1.487 1.582 427 2 1.485 1.570

0.080 93 1 1.408 1.558 203 1 1.422 1.557
0.090 60 2 1.359 1.509 127 2 1.365 1.510
0.100 46 3 1.331 1.471 86 1 1.331 1.506
0.110 33 2 1.288 1.453 63 1 1.261 1.491
0.120 26 1 1.270 1.450 49 1 1.213 1.473
0.130 20 1 1.157 1.452 45 3 1.235 1.395
0.140 19 3 1.160 1.375 34 2 1.148 1.393
0.150 17 4 1.108 1.348 30 2 1.165 1.365
0.160 13 2 1.011 1.356 25 1 1.134 1.384
0.170 12 2 1.045 1.330 24 3 1.112 1.312
119
Table 5.2:Variables Chain Sampling Plans Indexed by AQL and LQL

for =1% and =1%
Variables Chain Sampling Plan (Known Sigma)

p1 p2 n i k ' k Pa ( p1 ) Pa ( p 2 ) ASN
0.001 0.005 79 1 2.82801 2.83801 0.99007 0.00991 79
0.0025 0.01 91 1 2.56203 2.57203 0.99024 0.00956 91
0.005 0.02 77 1 2.31004 2.32004 0.99012 0.00974 77
0.0075 0.03 69 1 2.15205 2.16205 0.99003 0.00976 69
0.01 0.04 64 1 2.03504 2.04504 0.99007 0.00929 64
0.02 0.05 124 1 1.84403 1.85403 0.99018 0.00997 124
0.03 0.06 193 1 1.71300 1.73200 0.99007 0.00977 193
0.04 0.08 172 1 1.57299 1.58299 0.9906 0.00986 172
0.05 0.10 156 1 1.45798 1.46798 0.99017 0.00999 156
Table 5.3. Variables Single Sampling Plans Indexed by AQL and LQL
for =1% and =1%
Variables Single Sampling Plan (Known

p1 p2 Sigma)
n k Pa ( p1 ) Pa ( p 2 ) ASN
0.001 0.005 82 2.83305 0.99006 0.00992 82
0.0025 0.01 94 2.56703 0.99001 0.00980 94
0.005 0.02 80 2.31505 0.99016 0.00970 80
0.0075 0.03 72 2.15806 0.99004 0.00932 72
0.01 0.04 66 2.03907 0.99015 0.00957 66
0.02 0.05 130 1.84907 0.99018 0.00996 130
0.03 0.06 205 1.71806 0.99009 0.00971 205
0.04 0.08 182 1.57805 0.99007 0.00982 182
0.05 0.10 165 1.46305 0.99025 0.00984 165
120
Table 5.4: ASN Values of the Variables SSP, DSP and Variables Chain
Sampling Plans
ASN
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
Table 5.5. ARL of Variables Chain Sampling Plan for Different

i Values
ARL
p i = 1 i = 2 i = 3 i = 4 i =
0
0.005 88768.34 71697.5 60133.4 51781.5 607.80
0.01 868.88 554.66 410.18 327.19 38.29
0.015 81.89 50.25 37.21 30.13 10.64
0.02 20.07 12.66 9.77 8.25 5.06
0.025 8.12 5.42 4.41 3.92 3.15
0.03 4.41 3.14 2.70 2.50 2.29
0.035 2.88 2.19 1.98 1.89 1.83
0.04 2.14 1.73 1.62 1.58 1.56
0.045 1.73 1.47 1.41 1.39 1..39
0.05 1.48 1.32 1.28 1.28 1.28
0.055 1.33 1.22 1.20 1.20 1.20
0.06 1.23 1.15 1.15 1.14 1.14
0.065 1.16 1.11 1.11 1.11 1.11
0.07 1.11 1.08 1.08 1.08 1.08
0.075 1.08 1.06 1.06 1.06 1.06
0.08 1.06 1.04 1.04 1.04 1.04
0.085 1.04 1.03 1.03 1.03 1.03
0.09 1.03 1.02 1.02 1.02 1.02
0.10 1.02 1.01 1.01 1.01 1.01
121
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
Figure 5.1: OC Curves of a Variables Chain Sampling Plan for

Different i Values
122
CHAPTER 6
VARIABLES SAMPLING PLAN FOR RESUBMITTED LOTS
6.1 INTRODUCTION
Lot resubmissions are permitted in situations where the original

inspection results are suspected or when the producer or supplier is allowed to opt
for resampling. Moreover, in testing certain products, the test procedures are not
always accurate and also obtaining a random sample is difficult etc. In such cases,
resampling is preferable. Based on this idea, this chapter attempted to develop a
variables sampling plan for the inspection of resubmitted lots of normally
distributed quality characteristics.

variables resampling scheme.
(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.
123
6.3 OPERATING PROCEDURE OF THE KNOWN SIGMA VARIABLES

RESAMPLING SCHEME

specification limit U and follows a normal distribution with unknown mean and
known standard deviation . Then the following procedure is proposed for the
variables resampling scheme .
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 2: Accept the lot if v k and reject the lot if v k . On non-acceptance on

the original inspection, apply the variables single sampling plan for m
times and reject the lot if it is not accepted in the last stage of inspection.
In the case of lower specification limit L, the operating procedure is as follows.
and compute v X L ,
n
1
n, say 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 . On non-acceptance on

times and reject the lot if it is not accepted in the last stage of inspection.
124
Thus, the proposed variables resampling scheme is characterized by three

parameters, namely n, k and m. If m=1, then the proposed resampling scheme
will reduce to the variables single sampling plan.
6.4 DESIGNING OF VARIABLES RESAMPLING SCHEME WITH

KNOWN STANDARD DEVIATION
The variables sampling plans are, in general, designed based on two

points on the OC curve namely (p1, 1-) and (p2, ), where p1 is called the
acceptable quality level (AQL), p2 is the limiting quality level (LQL), is the
producers risk and is the consumers risk. A well designed sampling plan must
provide at least (1- ) probability of acceptance of a lot when the process fraction
nonconforming is at AQL and the sampling plan must also provide not more than
probability of acceptance if the process fraction nonconforming is at the LQL.
Thus the acceptance sampling plan must have its OC curve passing through two
designated points (AQL, 1-) and (LQL, ). The proposed variables resampling
scheme is designed based on the two points on the OC curve in the following
manner.
U
p 1 (6.1)

y
1 z2
( y) exp dz (6.2)
2 2
provided that the unit is classified as non-conforming if it exceeds the upper

specification limit U. So, the unknown mean can be determined if p is specified.
125
Let us define the standardized quality characteristic corresponding to

the fraction conforming as
z p 1 (1 p ) (6.3)
Govindaraju and Ganesalingam (1997) derived the performance

measures such as OC function, average sample number (ASN), average outgoing
quality (AOQ) etc for the attributes resampling scheme. Based on this, the
performance measures of the variables resampling scheme are given as follows.
The OC function of the variables resampling scheme, which gives the

proportion of lots that are expected to be accepted for given product quality, p is
given by
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)
where Pa ( p ) Pr v k is the probability of accepting a lot based on a variables
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.
The ASN function is given by
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
Govindaraju and Ganesalingam (1997) suggested that the preferred

value of m is two. That is, resampling is done on the resubmitted lot once if the lot
is not accepted on the original inspection. In this chapter, we also consider m=2
only for the proposed plan. So, when m=2, the probability of acceptance and ASN
of the resampling scheme become,
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
If (AQL, 1-) and (LQL, ) are prescribed then we have
(w1 )2 (w1 ) 1
and (w2 )2 ( w2 ) (6.10)
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)
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 variables
resampling scheme namely n, k, and m are determined by satisfying the required
127
producers and consumers conditions. Alternatively, we can determine the above

parameters of the variables resampling scheme to minimize the ASN at LQL,
which is analogous to minimizing the ASN in the variables repetitive group
sampling plans (see Balamurali et al. (2005). The ASN for the resampling scheme
is given in (6.5). Therefore, the following optimization problems are considered to
determine those parameters by minimizing the ASN at LQL and minimizing the
sum of ASN at AQL and LQL respectively.
Minimize ASN p 2 n 2 (w2 ) ( w2 )

Subject to L( p1 ) 1
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 )
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.
6.5 OPERATING PROCEDURE OF VARIABLES RESAMPLING

SCHEME WITH UNKNOWN SIGMA
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.
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
Step 2: Accept the lot if v k S and reject the lot if v k S . On non-acceptance on

times and reject the lot if it is not accepted on (m-1)st inspection.
Thus, the proposed unknown sigma variables resampling scheme is characterized

by three parameters namely nS, kS and m.
6.6 DESIGNING OF VARIABLES RESAMPLING SCHEME WITH

UNKNOWN STANDARD DEVIATION
The determination of parameters (nS, kS, m) of unknown sigma plan is

slightly different from the known sigma case. It is known that X k S S is
approximately normally distributed with mean k S E (S ) and variance
2
k S Var ( S ) (see Duncan (1986)). That is,
n
2 2
2
X k S S ~ N k S , kS
n 2n
Therefore, the probability of accepting a lot at each repetition is given by

P (v k S ) P X U k S S p
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 )
We obtain w1S and w2 S corresponding to AQL and LQL respectively by

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
In this case, the optimization problem becomes,

Minimize ASN p 2 n S 2 (w 2 ) ( w2 )
L( p 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
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 may determine the parameters of the known and unknown sigma

resampling schemes by solving the nonlinear equations given in [(6.12),(6.13)]
and [(6.16),(6.17)] respectively. Generally a sampling plan would be desirable if
the required number of sampled is small. So, this chapter considers the ASN as
the objective function to be minimized with the probability of acceptance along
with the corresponding producers and consumers risks as constraints. To solve
the above nonlinear optimization problems given in [(6.12),(6.13)] and
[(6.16),(6.17)], the sequential quadratic programming (SQP) proposed by Nocedal
and wright (1999) can be used. The SQP is implemented in Matlab software using
the routine fmincon. By solving the nonlinear problem mentioned above, the
parameters (n, k) for known sigma plan and the parameters (nS, kS) for unknown
sigma plan are determined and these values are tabulated in Table 6.1.
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
6.7 DESIGNING EXAMPLES

6.7.1. SELECTION OF KNOWN SIGMA VARIABLES RESAMPLING
SCHEME INDEXED BY AQL AND LQL
Table 6.1 is used to determine the parameters of the known and

unknown variables resampling schemes for specified values of AQL and LQL
when = 5% and = 10% with minimum ASN at LQL. Similarly Table 6.2 can
be used for obtaining the optimal parameters of the proposed scheme for both
known and unknown sigma with minimum sum of ASN at AQL and LQL. For
example, if p1 = 2%, p2 = 5%, = 5% and = 10%, Table 6.1 gives the
parameters as n = 35, k = 1.924 and m = 2. The ASN for this plan is 68.274. At
the same time, the unknown sigma resampling scheme under the same AQL and
LQL conditions, is determined as n S = 99, kS = 1.922 and m=2. The ASN at LQL
for this plan is 192.942.
For the above example, the plan is operated as follows.
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.
6.7.2. SELECTION OF UNKNOWN SIGMA RESAMPLING SCHEME

INDEXED BY AQL AND LQL
The unknown sigma resampling scheme is operated as a known sigma

resampling scheme but the parameters nS, kS and m are used in the place of n, k
and m respectively. Table 6.1 can also be used for the selection of the parameters
132
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.
6.8 MERITS OF THE VARIABLES RESAMPLING SCHEME
In this section, we discuss the advantages of the variables resampling

scheme over the conventional variables SSP. Two acceptance sampling plans will
be called equivalent when they possess nearly identical OC curves. A customary
procedure for achieving such equivalency consists of constructing the sampling
plans so that their OC curves coincide in two suitably chosen points namely (p1,
1-) and (p2, ). Suppose that for given values of p1=0.5%, =5%, p2=1% and
=10%, one can find the parameters of the known sigma variables resampling
scheme from Table 6.1 as
(i) n = 94, m=2, k = 2.497 and ASN (at LQL) = 183.401
For the same values of the AQL and LQL, we can determine the parameters of the
single sampling variables plan from (Sommers (1981)) as
(ii) n = 138 and k = 2.44, ASN=138
It can be observed that variables resampling scheme achieves a

reduction over 46% in sample size and about 25% reduction in ASN than the
variables SSP with same AQL and LQL conditions. It indicates that the variables
resampling scheme achieves same OC curve with minimum sample size or ASN
compared to the variables SSP. Further, Figure 6.1 gives the ASN curves of the
above mentioned variables resampling scheme and variables SSP. The resampling
scheme requires more sample size or ASN when the quality is poor that is, for
133
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.
Further, it is also to be pointed out that the variables resampling scheme

is economically superior to the variables single sampling plans in terms of ASN.
Obviously, a sampling plan having smaller ASN would be more desirable. The
variables double or multiple sampling plans are not practically very useful.
Variables sampling Standards avoid presenting such plans due to increased
complexity involved in operating them, but the variables resampling scheme has
no such complexity. Table 6.3 shows the ASNs for variable single and double
sampling variables plan along with variables resampling scheme for some
arbitrarily selected combinations of AQL and LQL. These ASN values are
calculated at the producers quality level for the known sigma plans. The ASN of
the variables single and double sampling plans can be found in Sommers (1981).
From this table, one can easily understand that the variables resampling scheme
will have minimum ASN when compared to the variables single sampling plans
and achieves the almost the same ASN as the variables double sampling plans.
Similar reduction in the ASN can be achieved for any combination of AQL and
LQL values. This implies that variables resampling scheme will give desired
protection with minimum inspection so that the cost of inspection will greatly be
134
reduced. Thus the variables resampling scheme provides better protection to the
producers than the variables single sampling plan.
6.9 COMPARISON WITH ATTRIBUTES RESAMPLING SCHEME
This section compares the proposed variables resampling schemes with

the attributes resampling scheme of Govindaraju and Ganesalingam (1997). For
this purpose, Table 6.4 is presented which gives the samples sizes of both
attributes and variables resampling schemes for different combinations of AQL
and LQL. The sample sizes of the attributes resampling schemes are taken from
Govindaraju and Ganesalingam (2007). This table apparently shows that variables
resampling scheme achieves a great reduction in sample size over the attributes
resampling scheme. Also when m increases for fixed values of AQL and LQL, the
sample size of the resampling scheme decreases but ASN(p2) will increase. This
can be easily observed from Table 6.5. Table 6.5 displays the optimal sample size
and ASN(p2) for different values of m and for fixed AQL and LQL values.
6.10 CONCLUSIONS
The proposed variables resampling scheme gives more protection to the

producer at the same time it also ensures the protection against the consumer
point of view. This plan is also simple to apply rather than double and multiple
sampling variables plans. Also this plan provides better protection than the
conventional single sampling variables plans with minimum ASN.
135
Table 6.1. Variables Resampling Schemes (with m=2) Involving Minimum

ASN Indexed by AQL and LQL
MinASN(p2)
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
0.005 0.006 1445 2.555 2815.899 6164 2.555 12011.75

0.008 208 2.523 405.631 868 2.523 1692.537
0.01 94 2.497 183.401 381 2.496 742.489
0.012 57 2.475 111.153 230 2.474 448.209
0.01 0.02 78 2.240 152.108 272 2.239 530.048

0.025 43 2.210 83.827 148 2.209 288.427
0.03 29 2.184 56.514 99 2.183 192.950
0.035 22 2.163 42.905 73 2.161 142.264
0.02 0.03 194 1.999 378.350 581 1.998 1132.236

0.035 99 1.977 193.036 291 1.977 567.281
0.04 63 1.957 122.805 183 1.957 356.665
0.05 35 1.924 68.274 99 1.922 192.942
0.03 0.04 342 1.839 666.519 919 1.839 1790.912

0.045 168 1.822 327.547 446 1.822 869.483
0.05 104 1.805 202.676 273 1.806 532.261
0.06 54 1.777 105.235 140 1.777 272.895
0.04 0.05 512 1.717 997.757 1266 1.717 2467.042

0.06 151 1.688 294.335 364 1.688 709.339
0.07 76 1.663 148.102 182 1.663 354.743
0.08 49 1.639 95.514 113 1.641 220.263
0.05 0.06 711 1.616 1385.524 1640 1.616 3195.917

0.07 201 1.591 391.712 455 1.591 886.678
0.08 101 1.568 196.871 224 1.568 436.525
0.09 63 1.547 122.797 137 1.548 267.022
0.10 44 1.528 85.753 95 1.529 185.177
136
Table 6.2.Variables Resampling Schemes (with m=2) Involving Minimum

Sum of ASN Indexed by AQL and LQL
Min[ASN(p1)+ASN(p2)]
p1 p2 ASN(p1)+ ASN(p1)+
n k ASN(p 2 ) n S k S ASN(p2)
0.001 0.0020 129 3.023 409.290 713 3.023 2262.286
0.004 30 2.951 95.166 160 2.951 507.560
0.006 18 2.902 56.940 90 2.906 285.412
0.008 13 2.862 41.005 64 2.873 202.938
0.005 0.006 1445 2.555 4571.700 6164 2.555 19500.28

0.008 208 2.523 660.100 868 2.523 2754.619
0.01 94 2.497 298.320 381 2.496 1207.84
0.012 57 2.475 180.887 230 2.474 729.240
0.01 0.02 78 2.240 247.511 272 2.239 862.196

0.025 43 2.210 136.414 148 2.209 469.126
0.03 29 2.184 91.948 99 2.183 313.654
0.035 22 2.163 69.788 73 2.161 231.305
0.02 0.03 194 1.999 615.683 581 1.998 1840.637

0.035 99 1.977 314.102 291 1.977 923.313
0.04 63 1.957 199.763 183 1.957 580.282
0.05 35 1.924 111.029 99 1.922 313.605
0.03 0.04 342 1.839 1083.797 919 1.839 2912.349

0.045 168 1.822 533.050 443 1.821 1403.939
0.05 104 1.805 329.550 273 1.806 866.205
0.06 54 1.777 171.273 140 1.777 444.022
0.04 0.05 512 1.717 1624.070 1266 1.717 4015.786

0.06 151 1.688 478.664 364 1.688 1153.925
0.07 76 1.663 241.006 182 1.663 577.120
0.08 49 1.639 155.159 113 1.641 358.503
0.05 0.06 711 1.616 2253.381 1640 1.616 5197.656

0.07 201 1.591 637.426 455 1.591 1442.941
0.08 101 1.568 320.076 224 1.568 709.906
0.09 63 1.547 199.569 137 1.548 434.460
0.10 44 1.528 139.392 95 1.529 301.225
137
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
Variables Resampling Scheme

p1 p2 m
Sample Size ASN at LQL
0.001 0.004 1 226 226.0
2 160 311.822
3 136 394.123
4 121 465.480
5 112 537.145
0.002 0.005 1 439 439.0

2 306 596.375
3 255 739.072
4 228 877.211
5 209 1002.380
0.01 0.02 1 388 388.0

2 272 530.048
3 227 657.840
4 201 773.324
5 186 892.258
139
Figure. 6.1: ASN Curves of a Variables Single Sampling Plan and

Resampling Sampling Scheme
140
FIGURE 6.2: OC CURVES OF A VARIABLES RESAMPLING SAMPLING

SCHEME FOR DIFFERENT m VALUES
141
CHAPTER 7
CONCLUSIONS AND FUTURE WORK
This thesis is mainly focused on optimal designing of the special

purpose sampling plans for the application of variables inspection. In this thesis,
the optimal parameters of respective special purpose plans are determined based
on two-points on the OC curve approach namely (AQL, 1-) and (LQL, ). In
chapter 1, we have presented certain basic concepts and review of literature which
are related to this thesis.
The quick switching system developed by Dodge (1967) is one of the

two-plan systems for the application of attributes quality characteristics. In any
two-plan system, the tightened inspection can be used when the quality of the
product deteriorated and normal inspection is used when the quality is found to be
good. In Chapter 2, we have investigated the optimal designing of variables quick
switching system by minimizing the average sample number, where the quality
characteristic of interest has the single specification limit and follows a normal
distribution. We have considered both known and unknown sigma cases for the
designing the variables quick switching system. The advantages of the variables
quick switching system over the variables single and double sampling plans and
attributes quick switching system have also been discussed. Tables have also been
constructed for the selection of parameters of known and unknown standard
deviation variables quick switching system for given AQL and LQL. Comparisons
have been made in terms of ASN with the existing plans and proved that ASN
given in the proposed system is optimum.
Chapter 3 of this thesis deals with the optimal designing of variables

quick switching system with double specification limits by minimizing the ASN.
142
We have investigated the variables quick switching system when a measurable

quality characteristic has double specification limits beyond which an item is
considered to be a non-conforming. The quality characteristic of interest is
assumed to follow a normal distribution. In this chapter, we have considered two
cases of fraction non-conforming namely symmetric fraction non-conforming and
asymmetric fraction non-conforming. The probability of acceptance of the
proposed variables quick switching system under double specification limits for
symmetric and asymmetric fraction non-conforming cases have also been derived.
The optimal parameters of the proposed system have been determined based on
two points on the OC curve approach by solving the optimization problem for the
known sigma asymmetric and asymmetric fraction non-conforming cases
separately. Necessary tables have also been constructed for the determination of
optimal parameters of the proposed sampling system with double specification
limits. Based on the comparisons, it has been proved that the proposed variables
quick switching system with double specification limits is optimum.
The Tightened-Normal-Tightened (TNT) sampling scheme procedure

developed by Calvin (1977) is a particular case of the general two plan system for
the inspection of attributes characteristics. In Chapter 4 of this thesis, we have
presented the optimal designing of variables TNT scheme by minimizing the ASN.
The proposed scheme can be applied for measurable characteristics, where the
quality characteristic follows normal distribution and has upper or lower
specification limit. We have considered known sigma and unknown sigma cases
separately for designing the proposed variables TNT scheme. Non-linear
optimization problems have been used to determine the optimal parameters of the
proposed scheme under known and unknown sigma cases. The advantages of the
proposed variables scheme over variables single, double sampling plans and
143
attributes sampling scheme have been discussed. Tables have also been
constructed for the application of the proposed scheme.
The concept of chain sampling was first introduced by Dodge (1955)

for the application of attribute quality characteristics. Govindaraj and Balamurali
(1998) extended the concept of chain sampling to variables inspection. Chapter 5
of the thesis has investigated the optimal designing of variables chain sampling
plan by minimizing the ASN. The chain sampling plan is one of the conditional
sampling plans and this plan under variables inspection will be useful for costly
and destructive testing. We have formulated an optimization problem for
determining the parameters of known and unknown sigma chain sampling plans.
The advantages of this proposed variables chain sampling plan over variables
single sampling plan and variables double sampling plan have also been discussed.
Tables have also been constructed for the selection of optimal parameters of
known and unknown standard deviation variables chain sampling plan for
specified two points on the operating characteristic curve namely the AQL and the
LQL along with the producer and consumers risks.
Govindaraju and Ganesalingam (1997) has proposed an attribute

sampling plan which can be applied in situations where resampling is permitted on
lots not accepted on original inspection. They have derived the performance
measures of the resampling scheme having single sampling attributes plan as the
reference plan. In this plan, it is assumed that during the course of resubmission,
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. Chapter 6 of the thesis has dealt with the optimal designing of variables
sampling plan for resubmitted Lots. This sampling plan can be applied for
inspection of resubmitted lots when the quality characteristic of interest follows
144
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.
All the sampling systems/plans developed in this thesis are

applicable for the inspection of normally distributed quality characteristics. There
is no special purpose sampling plan available in the literature when the quality
characteristic under study follows other than a normal distribution. So, developing
special purpose sampling plans for other Gaussian family of distributions such as
Inverse Gaussian distribution, Half-nomal distribution and Folded Normal
distribution will be considered as future study.
146
REFERENCES
American National Standards Institute/American Society of Quality Control

(ANSI/ASQC) Standard A2 (1987): Terms, Symbols, and Definitions for
Acceptance Sampling, American Society for Quality Control,
Milwaukee, Wisconsin, USA.
Anscombe ,F.J. , Godwin , H. J. and Plackett, R.L. (1947) , Methods of Deferred
Sentencing in Testing the Fraction Defective of a Continuous Output ,
Supplement to the Journal of the Royal Statistical Society, Vol.9,
pp. 198-217.
Arumainayagam, S.D. and Soundararajan, V. (1994), Construction and Selection
of Quick Switching Double Sampling System- Acceptance Number
Tightening, Communications in Statistics- Theory and Methods,
Vol.23, No.7, pp.2079- 2100.
Arumainayagam, S. D. and Soundararajan, V. (1995),Construction and Selection
of Quick Switching Double Sampling System: Sample Size
Tightening, Journal of Applied Statistics, Vol.22, No.1, pp.105-119.
Aslam, M., Balamurali, S, Jun, C-H and Ahmad, M. (2012),Bayesian Sampling
Inspection for Resubmitted Lots Under Gamma-Poisson Distribution,
Research Journal of Applied Sciences, Engineering and Technology,
Vol. 4, No.23, pp.5171-5176.
Aslam, M., Jun, C.H., Lio, Y., Ahmad, M. and Rasool, M.(2011),Group
Acceptance Sampling Plans for Resubmitted Lots under Burr Type XII
Distributions, Journal of Chinese Institute of Industrial Engineers,
Vol.28, pp.606-615.
147
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 and Palaniswamy, R. (2012),Determination of Minimum

Variance and VOQL Chain Sampling Plans for Compliance Testing
International Journal of Statistics and Management Systems, Vol.7,
No.1-2, pp.1-17.
Balamurali, S. and Subramani, J. (2010),Designing of Variables Repetitive
Group Sampling Plans Indexed by Point of Control, Journal of
Manufacturing Engineering, Vol.5, No. 3, pp.166-169.
Balamurali, S. and Usha, M. (2012a),Optimal Designing of Variables Quick
Switching Variables Sampling System by Minimizing Average Sample
Number, Journal of Applied Statistical Science, Vol.19, No.3,
pp.51-66.
Balamurali, S. and Usha, M. (2012b),Variables Quick Switching System with
Double Specification Limits, International Journal of Reliability,
Quality and Safety Engineering, Vol.19, No.2, pp. 1250008-1-17.
Balamurali, S. and Usha, M (2013),A New Failure Censored Sampling System
for Weibull Distribution, International Journal of Performability
Engineering, Vol.9, No.1, pp.3-12.
Balamurali, S., Azam, M. and Aslam, M. (2015),Attribute-Variable Inspection
Lots Policy using Resampling Based on EWMA, Communications in
Statistics-Simulation and Computation (To Appear).
Balamurali, S., Jun, C.H., and Collani, E.V., (2008),Skip-lot Sampling and Chain
Sampling, Encyclopedia of Statistics in Quality and Reliability, (Eds.
F. Ruggeri et al.), Wiley.
Balamurali, S., Aslam, M., Jun, C-H. and Ahmad, M. (2012),Bayesian Double
Sampling Plan under Gamma-Poisson Distribution, Research Journal
of Applied Sciences, Engineering and Technology, Vol. 4, No.8,
pp.949-956.
149
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
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
Duarte, B. and Saraiva, P. (2010),An Optimization-based Framework for

Designing Acceptance Sampling Plans by Variables for Non-
conforming Proportions, International Journal of Quality and
Reliability Engineering, Vol.27, pp.794-814.
Duarte, B. and Saraiva, P. (2013),Optimal Design of Acceptance Sampling Plans
by Variables for Non-conforming Proportions When the Standard
Deviation is Unknown, Communications in Statistics-Simulation and
Computation, Vol.42, pp.1318-1342.
Duncan, A. J. (1986): Quality Control and Industrial Statistics, 5th Ed., Richard D.
Irwin, Homewood, Illinois.
Evan, D. and Kemp, K. W. (1960),Sampling Inspection of Continuous Process
with no Auto-correlation Between Successive results, Biometrika,
Vol.47, pp.363-380.
Fallahnezhad, M. S. and Aslam, M. (2013),A New Economical Design of
Acceptance Sampling Models Using Bayesian Inference,
Accreditation and Quality Assurance, Vol.18, No.3, pp.187-195.
Feldmann, B. and Krumbholz, W. (2002),"ASN-Minimax Double Sampling Plans
for Variables", Statistical Papers, Vol.43, pp.361-377.
Ferrell, W.G. and Chhoker A. (2002),Design of Economically Optimal
Acceptance Sampling Plans with Inspection Error, Computers and
Operations Research, Vol.29, No.10, pp.1283-1300.
Frishman, F. (1960), An Extended Chain Sampling Plan, Industrial Quality
Control, Vol. 17, No. 1, pp. 10-12.
Govindaraju, K., (2011),Zero Acceptance Number Chained Quick Switching
System, Communications in Statistics-Theory and Methods, Vol. 40,
No.12, pp.2104-2116.
Govindaraju, K and Balamurali, S. (1998),Chain Sampling Plan for Variables
Inspection, Journal of Applied Statistics, Vol.25, pp. 103-109.
152
Govindaraju, K. and Ganesalingam, S. (1997),Sampling Inspection for

Resubmitted Lots, Communications in Statistics-Simulation and
Computation, Vol. 26, pp.1163-1176.
Govindaraju, K and Kuralmani, V. (1998),A Note on Operating Characteristic
Curve of the Known Sigma Single Sampling Variables Plan,
Communications in Statistics-Theory and Methods, Vol.21, No.8,
pp.2339-2347.
Govindaraju, K., Soundararajan, V. (1986),Selection of Single Sampling Plans
for Variables Matching the MIL-STD-105 Scheme, Journal of Quality
Technology, Vol. 18, pp.234238.
Grant, E. and Leavenworth, R. (1996): Statistical Quality Control, 7th Ed.,
McGraw-Hill, New York, NY.
Guenther, W. C. (1969),Use of the Binomial, Hypergeometric and Poisson Tables
to Obtain Sampling Plans, Journal of Quality Technology, Vol.1,
No.2, pp.105109.
Guenther, W. C. (1970),A Procedure for Finding Double Sampling Plans for
Attributes, Journal of Quality Technology, Vol.2, No.4, pp.219-225.
Hald, A. (1981): Statistical Theory of Sampling Inspection by Attributes. London:
Academic Press, New York.
Hald, A and Thyregod, P. (1965),The Composite Operating Characteristic under
Normal and Tightened Sampling Inspection by Attributes, Bulletin of
the International Statistical Institute, Vol. 41, pp.517-529.
Hamaker, H. C. (1960),Attribute Sampling in Operation, Bulletin of the
International Statistical Institute, Vol.37, No.2, pp.265-281.
Hamaker, H. C. (1979),Acceptance Sampling for Percent Defective by Variables
and by Attributes, Journal of Quality Technology, Vol.11, pp.139-148.
153
Hamilton, D.C. and Lesperance, M.L. (1995),A Comparison of Methods for

Univariate and Multivariate Acceptance Sampling by Variables,
Technometrics, Vol.37, pp.329-339.
Hill, I.D, Horsnell, G. and Warner, B.T. (1959) , Deferred Sentencing Schemes,
Applied Statistics, Vol.8, No. 2, pp.76-91.
Jun, C-H, Balamurali, S and Lee, S-H. (2006),Variables Sampling Plans for
Weibull Distributed Lifetimes under Sudden Death Testing, IEEE
Transactions on Reliability,Vol.55, No.1, pp.53-58.
Kao, J. H. K. (1971),MIL-STD-414: Sampling Procedures and Tables for
Inspection by Variables for Percent Defective, Journal of Quality
Technology, Vol. 3, pp.2837.
Kemp, K. W. (1962), The Use of Cumulative Sums for Sampling Inspection
Schemes, Applied Statistics , Vol.11, pp. 16-31.
Kocherlakota, S. and Balakrishnan, N. (1986),One and Two Sided Sampling
Plans Based on the Exponential Distribution, Naval Research
Logistics, Vol.33, No.3, pp.513-522.
Krumbholz, W. and Rohr, A. (2006),"The Operating Characteristic of Double
Sampling Plans When the Standard Deviation is Unknown",
Allgemeines Statistisches Archiv, Vol.90, No.2, pp.233-251.
Krumbholz, W.and Rohr, A. (2009),"Double ASN Minimax Sampling Plans by
Variables When the Standard Deviation is Unknown", Asta-Advances
in Statistical Analysis, Vol.93, pp.281-294.
Krumbholz, W., Rohr, A. and Vangjeli, E. (2012),"Minimax Versions of the Two-
Stage Test", Statistical Papers, Vol.53, pp.311-321.
Kuralmani, V. and Govindaraju, K. (1993), Selection of Conditional Sampling
Plans for given AQL and LQL, Journal of Applied Statistics, Vol. 20,
pp. 467-479.
154
Lam, Y. (1994),Bayesian Variable Sampling Plan for Exponential Distribution

with Type I Censoring, Annals of Statistics, Vol.22,No.2, pp.696-711.
Lieberman, G. J., Resnikoff, G. J. (1955),Sampling Plans for Inspection by
Variables, Journal of the American Statistical Association, Vol.50,
pp.457516.
MIL-STD 105D (1963), Sampling Procedure and Tables for Inspection by
Attributes , U.S. Government Printing Office, Washington, D.C.
Montgomery, D. C. (1985),The Effect of Non-normality on Variables Sampling
Plans, Naval Research Logistics Quarterly, Vol.32, pp.27.33.
Montgomery, D. C. (2005): Introduction to Statistical Quality Control. 5th Ed.
New York: John Wiley.
Nocedal, J. and Wright, S.J. (1999): Numerical Optimization, Springer.
Owen, D. B. (1967),Variables Sampling Plans Based on the Normal
Distribution, Technometrics, Vol. 9, pp.417423.
Owen, D. B.(1969),Summary of Recent Work on Variables Acceptance
Sampling with Emphasis on Non-normality, Technometrics, Vol.11,
pp.631-637.
Page, E. S.(1954),Continuous Inspection Schemes, Biometrika, Vol.41,
pp.100-115.
Peach, P. (1947): An Introduction to Industrial Statistics and Quality Control, 2nd
Edition, Edwards and Broughton, Releigh, North Corolina.
Pearn, W.L. and Wu, C.-W. (2006),Critical Acceptance Values and Sample Sizes
of a Variables Sampling Plan for Very Low Fraction of Non-
conforming, Omega, Vol.34, pp.90-101.
Pearn, W.L. and Wu, C.-W. (2007),An Effective Decision Making Method for
Product Acceptance, Omega, Vol.35, pp.12-21.
Perry, R. L. (1973),Skip-lot Sampling Plans, Journal of Quality Technology,
Vol.5, No.3, pp.123-130.
155
Raju, C. (1991). Three Stage Chain Sampling Plans, Communications in

Statististics Theory and Methods, Vol. 20, No. 5&6, pp.1777-1801.
Raju, C. and Jothikumar, J. (1997). Procedures and Tables for the Construction
and Selection of Chain Sampling Plans Chsp-4A(C1, C2)R-Part 5,
Journal of Applied Statistics, Vol. 24, No.1, pp.49-76.
Raju, C. and Murthy, N. M. N. (1995). Minimum Risks Chain Sampling Plans
(ChSP- 4(c1, c2)) Indexed by Acceptable Quality Level and Limiting
Quality Level, Journal of Applied Statistics, Vol.22, No. 3,
pp.389-426.
Raju, C. and Murthy, N. M. N. (1995). Procedures and Tables for the
Construction and Selection of Chain Sampling Plans ChSP-4A(c1, c2)r-
Part 4, Journal of Applied Statistics, Vol. 22, No.2, pp.261-271.
Raju, C.and Murthy, N. M. N. (1996), Two-Stage Chain Sampling Plans ChSP-
(0,2) and ChSP-(1,2) Part 1, Communications in Statistics
simulation, Vol. 25, No.2, pp.557-572.
Romboski, L. D. (1969): An Investigation of Quick Switching Acceptance
Sampling System, Doctoral Dissertation, Rutgers- The State University,
New Brunswick, New Jersey.
Sahli, A., Trecourt, P. and Robin, S. (1997),One Sided Acceptance Sampling by
Variables: The Case of the Laplace Distribution, Communications in
Statistics-Theory and Methods, Vol.26, No.11, pp.2817-2834.
Schilling, E. G. (1982): Acceptance Sampling in Quality Control. New York:
Marcel Dekker Inc.
Schilling, E. G. (1985),The Role of Acceptance Sampling Plans in Modern
Quality Control, Communications in Statistics-Theory and Methods,
Vol.14, No.11, pp.2769-2783.
Schilling, E. G. (2005),Average Run Length and the OC Curve of Sampling
Plans, Quality Engineering, Vol.17, pp.399-404.
156
Schilling, E. G. and Sommers, D. J. (1981),Two Point Optimal Narrow Limit

Plans with Applications to MIL-STD 105 D, Journal of Quality
Technology, Vol13, No.2, pp.83-92.
Schneider, H. and Wilrich, P. (1981): The Robustness of Sampling Plans for
Inspection by Variables, Computational Statistics: Wolfgang Wetzel
Zur Vollendung Seines, Vol.60, pp.281-295.
Seidel, W. (1997),Is Sampling by Variables Worse than Sampling by Attributes?
A Decision Theoretic Analysis and a New Mixed Strategy for
Inspecting Individual Lots, Sankhya B, Vol. 59, pp.96-107.
Senthilkumar, D and Muthuraj, D. (2010), Construction and Selection of
Tightened-Normal-Tightened Sampling Inspection Scheme of Type
TNT-(n1, n2; k), Journal of Applied Statistics, Vol.37, pp.375-390.
Sherman, R.E.(1965),Design and Evaluation of Repetitive Group Sampling
Plans, Technometrics, Vol.7, pp.11-21.
Sheu, L.C., Yeh, C.H., Yen, C.H., and Chang, C.H., (2014),Developing
Acceptance Sampling Plans based on Incapability Index Cpp, Applied
Mathematics and Information Sciences.,Vol. 8, No. 5, pp. 2509-2514
Singh, H. R.(1966),Producer and Consumer Risks in Non-normal Populations,
Technometrics, Vol.8, pp.335-343.
Sommers, D. J. (1981), Two-Point Double Variables Sampling Plans, Journal of
Quality Technology, Vol. 13, pp.25-30.
Soundararajan, V. (1978a),Procedures and Tables for Construction and Selection
of Chain Sampling Plans (ChSP-1)-Part I, Journal of Quality
Technology, Vol.7, pp.56-60.
Soundararajan, V. (1978b),Procedures and Tables for Construction and Selection
of Chain Sampling Plans (ChSP-1), Part 2; Tables for Selection of
Chain Sampling Plans, Journal of Quality Technology, Vol.10, No.3,
pp.99-103.
157
Soundararajan, V and Arumainayagam, S. D. (1992),Some Sampling Plans with

Identical Operating Characteristics Curves, Journal of Applied
Soundararajan, V and Palanivel, M (1997),Quick Switching Variables Single
Sampling System Indexed by AQL and LQL: Acceptance Criterion
Tightening, Journal of Applied Statistical Sciences, Vol.6, No.1,
pp.45-57.
Soundararajan, V. and Vijayaraghavan, R. (1990),Construction and Selection of
Tightened- Normal-Tightened (TNT) Plans, Journal of Quality
Technology, Vol.22, pp.146-153.
Soundararajan, V. and Vijayaraghavan, R. (1992),Construction and Selection of
Tightened-Normal-Tightened Sampling Inspection Scheme of Type
TNT-(n1,n2;c), Journal of Applied Statistics, Vol.19, pp.339-349.
Srivastava, ABL. (1961),Variables Sampling Inspection for Non-normal
Samples, Journal of Science and Engineering Research, Vol.5,
pp.145-152.
Stephens, K. S. and Larson, K. E. (1967),An Evaluation of the MIL-STD 105D
System of Sampling Plans, Industrial Quality Control, Vol.23,
pp.310-319.
Suresh, R. P., Ramanathan, T. V. (1997),Acceptance Sampling Plans by
Variables for a Class of Symmetric Distributions, Communications in
Statistics-Simulation and Computation, Vol.26, No.4, pp.13791391.
Takagi, K. (1972),On Designing Unknown Sigma Sampling Plans Based on a
Wide Class of Non-normal Distributions, Technometrics, Vol.14,
pp.669-678.
U.S. Department of Defense (1957): Sampling Procedures and Tables for
Inspection by Variables for Percent Defective, Military Standard
MIL-STD-414, Washington D.C.
158
Vaerst, R.(1982),A Procedure to Construct Multiple Deferred State Sampling

Plan, Methods of Operations Research, Vol.37, pp.477-485.
Vijayaraghavan, R. and Sakthivel, K. M. (2011),Chain Sampling Inspection
Plans Based on Bayesian Methodology, Economic Quality Control,
Vol.26, No.2, pp.173-187.
Wallis, W.A. (1947): Use of Variables in Acceptance Inspection by Percent
Defective, in Selected Techniques of Statistical Analysis, McGraw-Hill,
New York.
Wortham, A.W. and Baker, R.C.(1976),Multiple Deferred State Sampling
Inspection, International Journal of Production Research, Vol.14,
pp.719-731.
Wortham, W. A., and Mogg, J. M., (1970). Dependent Stage Sampling
Inspection, International Journal of Production Research, Vol. 8,
No.4, pp. 385-395.
Wu, C. W. and Pearn, W.L.(2008),A Variables Sampling Plan Based on C pmk for
Product Acceptance Determination, European Journal of Operational
Research, Vol.184, pp.549-560.
Wu, C.-W., Aslam, M. and Jun, C.-H.(2012),Variables Sampling Inspection
Scheme for Resubmitted Lots Based on Process Capability Index Cpk,
European Journal of Operational Research, Vol.217, pp.560-566.
Yen, C.H. , Aslam, M., and Jun, C.H., (2014), A lot inspection sampling plan
based on EWMA yield index, International Journal of Advanced
Manufacturing Technology, Vol. 75, No. 5-8, pp 861-868.
Zimmer, W. J. and Burr, I.W. (1963),Variables Sampling Plans Based on Non-
normal Populations, Industrial Quality Control, Vol.20, No.1,
pp.18-26.
159
LIST OF PUBLICATIONS
1. Balamurali, S. and Usha, M. (2012),Optimal Designing of a Variables

Quick Switching Sampling System by Minimizing the Average Sample
Number, Journal of Applied Statistical Science, Vol.19, No.3, pp.51-66.
2. Balamurali, S. and Usha, M. (2012),Variables Quick Switching System with

Double Specification Limits, International Journal of Reliability,
Quality and Safety Engineering, Vol.19, No.2, pp.1250008-1-17.
3. Balamurali, S and Usha, M. (2013),A New Failure Censored Variables

Sampling System for Weibull Distribution, International Journal of
Performability Engineering, Vol.9, No.1, pp.3-12.
4. Balamurali, S. and Usha, M. (2013),Optimal Designing of Variables Chain

Sampling Plan by Minimizing the Average Sample Number,
International Journal of Manufacturing Engineering, Volume 2013,
Article ID 751807, 12 pages.
5. Balamurali, S. and Usha, M. (2014),Optimal Designing of Variables Quick

Switching System Based on the Process Capability Index Cpk, Journal
of Industrial and Production Engineering, Vol.31, No.2, pp.85-94.
6. Balamurali, S. and Usha, M.(2015),Optimal Designing of Variables

Sampling Plan for Resubmitted Lots, Communications in Statistics-
Simulation and Computation, Vol. 44, No.5, pp. 1210-1224.
160
7. Balamurali, S. and Usha, M. (2015),Optimal Designing of Variables

Tightened-Normal-Tightened Sampling Scheme by Minimizing
the Average Sample Number, International Journal of Industrial and
Systems Engineering, Vol.21, No.1, pp.99-118.
8. Balamurali, S. and Usha, M. (2014),A New System of Skip-lot Resampling

Schemes, Communications in Statistics-Simulation and Computation
(Accepted) .
9. Balamurali, S. and Usha, M. (2015),Designing of Variables Quick

Switching Sampling System by Considering Process Loss Functions,
Communications in Statistics-Theory and Methods (Accepted).
LIST OF PAPERS PRESENTED IN CONFERENCES

(1) Balamurali, S. and Usha, M (2011),Evaluation of MLP-rx1Continuous
Sampling Plan Using a Markov Chain Model, Paper Presented at the
International Conference on Mathematical Modelling and
Applications to Industrial Problems (MMIP 2011), NIT, Calicut,
March 28-31, 2011.
161
CURRICULUM VITAE
Mrs. M. Usha was an Assistant Professor at Kalasalingam University for 7

years. She has also more than 6 years of teaching experience at undergraduate and
postgraduate levels in various established Arts and Science Colleges in
Tamilnadu, India.
She has obtained her B.Sc (Mathematics) from Madurai Kamaraj

University with first class in the year 1997. She has obtained her Post Graduate
degree, M.Sc ( Mathematics) from Madurai Kamaraj University with first class in
the year 1999. She completed her M.Phil (Mathematics) with first class with
distinction in the year 2000.
She is a life member of Indian Society of Technical Education (ISTE).

Her areas of interest are Statistical Quality Control, Probability and Statistics and
Operations Research. She has published her research articles in refereed
International Journals and International conference.

PH.D Thesis M. Usha

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

PH.D Thesis M. Usha

Încărcat de

Drepturi de autor:

Formate disponibile

STUDIES ON OPTIMAL DESIGNING OF CERTAIN

SPECIAL PURPOSE SAMPLING PLANS FOR

Acceptance sampling is one of the major areas of statistical

In general, the acceptance sampling plans are classified in to

Variables sampling plans specify the number of items to be

The special purpose sampling plans such as Chain Sampling Plan

Chapter 1 of this thesis comprises of sections that consists of basic concepts

Chapter 3 of the thesis investigates the variables quick switching sampling

The Chapter 4 of the thesis deals with the optimal designing of

In Chapter 5 of the thesis, we investigate the optimal designing of chain

In Chapter 6, an optimal designing of variables sampling plan which can

like to express my deepest gratitude.

I am deeply indebted to Dr. S. BALAMURALI, Ph.D., Professor,

Department of Mathematics, Kalasalingam University, Krishnankoil, TamilNadu,

India for his valuable Supervision, continuous guidance, expert consultancy,

experience a personification of excellence, dedication and commitment. I also

I wish to record my gratitude to the Management and the authorities of

Kalasalingam University, for granting me permission and providing necessary

members of the Doctoral Committee for their support.

and give her best suggestions.

love and patience of my family. I would like to express my heart-felt gratitude to

I would like to thank my spouse , Mr. T. Sermaraj, and my children

S. Divyamaki and S. Lakinyamaki for their constant love and support.

Last but not the least, I would like to thank my parents

Mr. G. Mahalingam and Mrs. M. Jeyasamvarthini , for giving birth to me at

me with their best wishes.

CHAPTER TITLE PAGE

2. OPTIMAL DESIGNING OF VARIABLES QUICK 34

CHAPTER TITLE PAGE

2.6 Average Sample Number 37

3. OPTIMAL DESIGNING OF VARIABLES QUICK SWITCHING 53

CHAPTER TITLE PAGE

3.7 Illustrative Examples 63

4. OPTIMAL DESIGNING OF VARIABLES TIGHTENED 81

4.7.2 Selection of Unknown Sigma Variables TNT 90

CHAPTER TITLE PAGE

5. OPTIMAL DESIGNING OF VARIABLES CHAIN SAMPLING 100

CHAPTER TITLE PAGE

5.8 Average Run Length of Variables ChSP 114

6. OPTIMAL DESIGNING OF VARIABLES SAMPLING 122

CHAPTER TITLE PAGE

7. CONCLUSIONS AND FUTURE WORK 141

TABLE NO. TITLE PAGE NO.

2.1 Variables Quick Switching Sampling Systems 49

TABLE NO. TITLE PAGE NO.

4.1. Variables Tightened-Normal-Tightened Sampling Scheme 93

TABLE NO. TITLE PAGE NO.

6.3 Sample Size of the Attributes and Variables Resampling 138

FIGURE NO. FIGURE CAPTIONS PAGE NO.

1.1 Operating Characteristic Curve

4.1 OC Curves of Single Sampling Normal Plan (12, 1.857), 96

5.1 OC Curves of a Variables Chain Sampling Plan for 121

6.1 ASN Curves of a Variables Single Sampling Plan and 142

6.2 OC curves of a Variables Resampling Scheme for 143

LIST OF SYMBOLS AND ABBREVIATIONS

ks Rejection criteria for unknown sigma plan

Inspection of raw materials, semi finished products, or finished products

International Organization for Standardization (ISO) 8402-1986

1.2 QUALITY CONTROL

1.3 STATISTICAL QUALITY CONTROL

1.4 TOOLS OF SQC

spread of quality, comparison with specific requirements and process