studies on optimal designing of certain special...
TRANSCRIPT
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 Nk and Tk and are used to calculate the
probability of acceptance for different combinations of the consumer’s confidence
levels and the producer’s 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 producer’s and consumer’s
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 1n , 2n 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 producer’s and consumer’s 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 consumer’s 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
4.2 Conditions of Application 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
5.2 Conditions of Application 101
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
6.2 Conditions of Application 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
Specification Limits (Symmetric Fraction Non-conforming)
3.4 Optimal Parameters of Unknown Sigma VQSS with Double 76
Specification Limits (Symmetric Fraction Non-conforming)
3.5 Average Sample Number of Variables Single Sampling Plans 78
and VQSS with Double Specification Limits (Symmetric
Fraction Non- conforming)
xx
TABLE NO. TITLE PAGE NO. 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
and LQL for =1% and =1%
5.3 Variables Single Sampling Plans Indexed by AQL 119
and LQL for =1% and =1%
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
TABLE NO. TITLE PAGE NO.
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 iσ 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)
α Producer’s risk
β Consumer’s 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
Nk Acceptance criteria of normal inspection
Tk Acceptance criteria of tightened inspection
σ Population standard deviation
S2 Sample variance
U Upper specification limit
L Lower specification limit
nσ Sample size for known sigma plan
kσ Acceptance criteria for known sigma plan
k’σ Rejection criteria for known sigma plan
ns Sample size for unknown sigma plan
ks Acceptance criteria for unknown sigma plan
xxiv
k’s Rejection criteria for unknown sigma plan
푖휎 Number of preceding lots considered for accepting current lot for
known sigma plan
푖s 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
휎 The population standard deviation
Φ(⋅) 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 company’s 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 μ0 of the mean
from its true value μ. 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
nN
xnmN
xm
xXP
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:
nxppxn
xXP xnx .........3,2,1,0,1
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
.......3,2,1,0,!
xx
exXPx
λ 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:
0,
,,2
1,, 2
2
22
xexfx
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 consumer’s requirement, the producer’s
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-α) and (p2, β) 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.
α = the producer’s risk of rejecting p1 quality and
β = the consumer’s 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 Producer’s Quality Level
(PQL) and p2 as the Consumer’s 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 α and β 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
σ 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 nxxx ,...., 21 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 Ukx , then accept the lot and if
Ukx , 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 Dodge’s 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: TN cc )
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
TN
Ta PP
PpP
1)(
where is the probability of accepting a lot based on a
single sample with parameters
and is the probability of accepting a lot based on a
single sampling plan with parameters
NN kvP Pr
TT kvP Pr
),( Nkn
),( Tkn
26
2. The Average Sample Number (ASN) is
3. The Average Outgoing Quality (AOQ) is
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 0c , 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
npASN
)( ppPpAOQ a
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
innna PPPpP ,0,1,0
nP ,0 = Probability of getting exactly 0 defective in a sample of size n
nP ,1 = Probability of getting exactly 1 defective in a sample of size n
2. The Average Sample Number (ASN) is 3 The Average outgoing Quality (AOQ) is given by
npASN
)( ppPpAOQ a
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 consumer’s 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 consumer’s 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
where
and
2. The Average Sample Number (ASN) is
1,)(
mmnnpASN
where 1s ,)1)(1(
2
22
2
s
s
PPP
and st ,)1(
1
11
1
PPP
t
t
3. The Average outgoing Quality (AOQ) and
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,
12)( PPpPa
1s ,)1)(1(
2
22
2
s
s
PPP
s t , )1(
1
11
1
PPP
t
t
)( ppPpAOQ a
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
)()(1....)()(1)()(1)()( 12 pPpPpPpPpPpPpPpL am
aaaaaa
ma pP )(11
where kvpPa Pr)(
2. The Average Sample Number (ASN) is npPnpPnpPnpASN m
aaa12 )(1....)(1)(1)(
)(
)(11pP
pPn
a
ma
3. The Average outgoing Quality (AOQ) ppPpAOQ a
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 σ. Then the following procedure of the VQSS is proposed.
Step 1: Start with normal inspection. During normal inspection, take a random
sample of size n, say nXXX ..., 21 and compute
XUv
, where
n
iiX
nX
1
1 .
Step 2: Accept the lot if Nkv and reject the lot if Nkv . 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
nXXX ..., 21 and compute
XUv
, where
n
iiX
nX
1
1 .
Step 4: Accept the lot if Tkv and reject the lot if Tkv )( NT kk .
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 kNσ=kTσ, 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
)Pr()Pr(1
)Pr(1
)(
TN
T
TN
Ta kvkv
kvPP
PpP
(2.1)
where NN kvP Pr is the probability of accepting a lot based on a single
sampling plan with parameters (n, kN) and TT kvP Pr 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
)()(11 vvUp
(2.2)
where )(y is given by
y
dzzy2
exp21)(
2
, (2.3)
provided that the quality characteristic of interest is normally distributed with
mean µ and standard deviation σ, 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
)()(1
)()(TN
Ta ww
wpP
(2.4)
where nkvw TT and nkvw NN
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 producer’s risk α and
the consumer’s risk β are prescribed then the OC function can be expressed as
1)()(1
)(
11
1
TN
T
www (2.5)
and
)()(1
)(
22
2
TN
T
www (2.6)
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, nkvw TT )( 11 , nkvw NN )( 11
nkvw TT )( 22 and nkvw NN )( 22 (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
nnnpASN
)( (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, kNσ, 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)= nσ
Subject to
1)( 1pPa
)( 2pPa
0 ,1 TNσ k kn (2.9)
where )( 1pPa and )( 2pPa 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 σ. 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
lot if NSkv and reject the lot if NSkv , where
SXU
v
,
Sn
ii
S
Xn
X1
1 and 1
)( 2
S
i
nXX
S . 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 TSkv and reject the lot if TSkv , where
SXU
v
,
Sn
ii
S
Xn
X1
1 and 1
)( 2
S
i
nXX
S . 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 Hamaker’s (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
SkX NS is approximately normally distributed with mean )(SEkNS and
variance )(2
SVarkn NS
S
(see Duncan (1986)).
That is.,
SNS
SNSNS n
kn
kNSkX2
,~2
22
Therefore, the probability of accepting a lot under normal inspection is given by
pSkUXP NS pUSkXP NS
21)/(
2NS
S
NS
kn
kU
21
)( 2NS
SNS k
nkv
If we let
21
)( 2NS
SNSNS k
nkvw then the probability of acceptance under
tightened inspection is considered )( NSw .
41
Similarly if we let
21
)( 2TS
STSTS k
nkvw then the probability of acceptance
under tightened inspection is taken as )( TSw . Hence the lot acceptance
probability of the proposed system for sigma unknown case under two-points on
the OC curve is given by
)()(1
)()(
11
11
STSN
STa ww
wpP
(2.10)
and )()(1
)()(
22
22
STSN
STa ww
wpP
(2.11)
We obtain STSNSTSN wwww 2211 ,,, corresponding to 2211 ,,, TNTN wwww respectively by
21
)( 211NS
SNSSN k
nkvw ,
21
)( 211TS
STSST k
nkvw
21
)( 222NS
SNSSN k
nkvw and
21
)( 222TS
STSST k
nkvw
In this case, the optimization problem becomes,
Minimize ASN(p1) = nS
Subject to
1)( 1pPa
)( 2pPa
0 ,1 TSNSs k kn (2.12)
42
where )( 1pPa and )( 2pPa 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 producer’s and consumer’s 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
and compute
XUv , where
12
1121
iiXX . Accept the lot if 552.1v
and reject the lot if 552.1v . 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
XUv , where
12
1121
iiXX . Accept the lot if 817.1v and reject the lot if 817.1v .
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) nσ = 116 and k = 2.17
(iii) n = 28, kNσ = 1.923 and kTσ = 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-α = 0.95) and (p2 = 0.045, β = 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 producer’s 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
p1
p2
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n kTσ kNσ nS kTS kNS 0.001
0.0025
0.005
0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010 0.012 0.015 0.025
0.004 0.005 0.006
0.0075 0.010 0.015 0.020 0.025 0.030 0.040
0.0075 0.010 0.012 0.015 0.020 0.030 0.035 0.040 0.060
134 45 31 14 11 10 14 8 7 9 6 4
96 35 26 23 14 9 8 6 8 5
100 31 23 19 9 9 7 7 6
2.998 2.958 2.899 2.994 2.985 2.939 2.769 2.897 2.899 2.710 2.751 2.672
2.841 2.917 2.878 2.773 2.762 2.693 2.572 2.569 2.353 2.374
2.626 2.683 2.641 2.549 2.657 2.377 2.385 2.298 2.109
2.943 2.833 2.784 2.599 2.535 2.519 2.639 2.457 2.409 2.525 2.381 2.222
2.596 2.432 2.398 2.413 2.302 2.193 2.182 2.079 2.213 2.039
2.361 2.183 2.146 2.139 1.897 1.982 1.895 1.913 1.884
234 108 104 71 60 45 38 35 32 28 24 16
326 156 113 82 56 37 33 28 19 15
330 123 89 63 45 28 23 20 15
3.193 3.166 2.999 2.998 2.955 2.985 2.982 2.947 2.926 2.880 2.819 2.715
2.917 2.945 2.910 2.861 2.809 2.710 2.563 2.490 2.564 2.485
2.679 2.705 2.664 2.617 2.533 2.443 2.424 2.394 2.224
2.773 2.666 2.704 2.628 2.600 2.525 2.482 2.467 2.446 2.415 2.379 2.260
2.542 2.445 2.410 2.371 2.309 2.235 2.233 2.200 2.074 2.010
2.329 2.205 2.169 2.122 2.073 1.978 1.929 1.894 1.834
47
Table 2.1. Contd….
p1
p2
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n kTσ kNσ nS kTS kNS 0.0075
0.010
0.015
0.010 0.012 0.015 0.020 0.025 0.030 0.035 0.040 0.050 0.060 0.070
0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050 0.060 0.070 0.080 0.090 0.100
0.020 0.025 0.030 0.035 0.040 0.045 0.050 0.060 0.070 0.080
149 58 34 23 14 11 11 11 8 7 6
68 28 21 15 13 14 10 14 7 6 6 6 3
129 43 26 23 22 21 13 11 8 7
2.499 2.527 2.485 2.391 2.402 2.367 2.263 2.179 2.151 2.086 2.036
2.427 2.418 2.334 2.314 2.255 2.141 2.179 2.005 2.123 2.084 1.987 1.904 1.906
2.236 2.265 2.252 2.163 2.082 2.012 2.083 2.009 2.021 1.952
2.244 2.137 2.080 2.041 1.922 1.867 1.893 1.909 1.816 1.781 1.686
2.047 1.923 1.899 1.829 1.810 1.856 1.754 1.875 1.648 1.599 1.617 1.629 1.557
1.971 1.835 1.757 1.768 1.795 1.787 1.658 1.629 1.526 1.526
575 209 107 64 47 36 31 26 20 17 14
243 94 62 47 38 32 27 24 20 17 14 11 11
402 121 78 59 48 39 34 26 22 18
2.502 2.549 2.549 2.486 2.426 2.395 2.337 2.313 2.266 2.196 2.175
2.434 2.441 2.393 2.344 2.304 2.268 2.247 2.214 2.148 2.100 2.089 2.138 2.027
2.248 2.317 2.279 2.233 2.188 2.167 2.131 2.088 2.025 2.005
2.252 2.144 2.059 1.996 1.951 1.900 1.877 1.838 1.776 1.741 1.685
2.064 1.946 1.893 1.854 1.819 1.788 1.752 1.729 1.693 1.655 1.599 1.513 1.532
1.973 1.827 1.779 1.748 1.723 1.687 1.666 1.613 1.585 1.535
48
Table 2.1 Contd….
p1
p2
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n kTσ kNσ nS kTS kNS 0.015
0.020
0.030
0.040
0.090 0.100
0.030 0.035 0.040 0.045 0.050 0.060 0.070 0.080 0.090 0.100 0.120
0.040 0.045 0.050 0.060 0.070 0.080 0.090 0.100 0.120 0.150 0.200
0.060 0.070 0.080 0.090 0.100 0.110 0.120 0.140 0.160
7 5
87 33 26 21 17 17 17 12 8 7 5
116 57 33 22 17 14 12 10 11 6 5
58 29 21 17 14 14 14 10 8
1.888 1.856
2.070 2.158 2.108 2.076 2.060 1.926 1.813 1.817 1.877 1.851 1.700
1.934 1.957 1.998 1.947 1.894 1.844 1.800 1.781 1.597 1.647 1.470
1.801 1.826 1.796 1.754 1.725 1.639 1.565 1.546 1.512
1.508 1.376
1.840 1.673 1.653 1.621 1.580 1.618 1.638 1.552 1.417 1.376 1.345
1.679 1.592 1.493 1.442 1.404 1.374 1.345 1.296 1.362 1.152 1.110
1.476 1.361 1.311 1.279 1.240 1.264 1.280 1.191 1.127
16 14
170 96 68 54 50 33 26 24 19 16 18
269 133 89 54 40 31 26 29 24 13 10
104 63 47 37 30 26 24 17 17
1.956 1.927
2.167 2.172 2.153 2.120 2.046 2.044 2.004 1.910 1.911 1.897 1.664
1.965 1.999 1.999 1.974 1.925 1.893 1.846 1.695 1.598 1.644 1.458
1.892 1.869 1.819 1.778 1.749 1.704 1.644 1.634 1.195
1.511 1.477
1.772 1.702 1.658 1.630 1.636 1.554 1.510 1.510 1.451 1.407 1.474
1.665 1.584 1.535 1.469 1.430 1.388 1.361 1.415 1.388 1.224 1.173
1.427 1.369 1.339 1.308 1.274 1.254 1.249 1.169 1.485
49
Table 2.1 Contd….
p1
p2
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n kTσ kNσ nS kTS kNS
0.050
0.18 0.200
0.060 0.070 0.080 0.090 0.100 0.120 0.140 0.160 0.200 0.250
7 4
280 66 35 25 23 14 11 9 6 5
1.460 1.317
1.666 1.723 1.735 1.711 1.630 1.619 1.560 1.511 1.432 1.324
1.090 0.882
1.521 1.373 1.280 1.231 1.245 1.134 1.085 1.036 1.006 0.860
12 10
462 132 73 53 40 28 21 16 15 8
1.521 1.496
1.708 1.761 1.768 1.730 1.709 1.640 1.585 1.560 1.328 1.375
1.096 1.046
1.488 1.366 1.293 1.260 1.219 1.170 1.120 1.060 1.083 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
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1Fraction Nonconforming, p
Prob
abili
ty o
f Acc
epta
nce,
Pa(
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)
Variables Normal SSP
VQSS
Variables Tightened SSP
51
Table 2.2: ASN Values of the Known Sigma Variables Single Sampling Plan, Variables Double Sampling Plan and VQSS
p1
p2
ASN Variables
SSP Variables
DSP Variables*
QSS
Variables ** QSS
0.001 0.001 0.005 0.005 0.03 0.03 0.04 0.04 0.05 0.05
0.002 0.003 0.010 0.012 0.04 0.06 0.07 0.08 0.07 0.08
191 74 138 85 506 81 114 72 300 149
154.9 59.4
112.0 69.5
434.6 127.7 180.6 58.4
246.7 122.3
191 74
138 85
506 81
114 72
300 149
134 45 31 23
116 22 29 21 66 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
p1
p2
ASN Variables
SSP Variables
DSP Variables*
QSS
Variables ** QSS
0.001 0.001 0.005 0.005 0.03 0.03 0.04 0.04 0.05 0.05
0.002 0.003 0.010 0.012 0.04 0.06 0.07 0.08 0.07 0.08
1032 381 547 327
1333 197 258 159 660 319
829.1 302.4 437.1 263.0 1138.7 316.5 417.6 125.8 535.4 258.0
1032 412 941 823 945 357 263 201 768 572
234 108 123 89
269 89
104 47
132 73
* ASN given in Soundararajan and Palanivel (1997) ** ASN given in Table 2.1
52
Table 2.4: Parameters of Known Sigma VQSS for Some Selected Combinations of AQL and LQL Values
p1
p2
*Parameters of VQSS
**Parameters of VQSS
n kNσ kTσ n kNσ kTσ 0.001 0.001 0.005 0.005 0.03 0.03 0.04 0.04 0.05 0.05
0.002 0.003 0.010 0.012 0.04 0.05 0.06 0.08 0.06 0.07
370 95 399 126
1856 515 965 121
1311 853
2.93 2.87 2.40 2.39 1.73 1.73 1.62 1.55 1.55 1.53
3.10 3.00 2.60 2.50 2.00 1.90 1.78 1.68 1.66 1.66
134 45 31 23
116 33 58 21
280 66
2.943 2.833 2.183 2.146 1.679 1.493 1.476 1.311 1.521 1.373
2.998 2.958 2.683 2.641 1.934 1.998 1.801 1.796 1.666 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
p1
p2
*Parameters of VQSS
**Parameters of VQSS
nS kNS kTS nS kNS kTS 0.001 0.001 0.005 0.005 0.03 0.03 0.04 0.04 0.05 0.05
0.002 0.003 0.010 0.012 0.04 0.05 0.06 0.08 0.06 0.07
2053 504
1647 502
5100 1365 2362 279
3001 1939
2.93 2.87 2.40 2.39 1.73 1.73 1.62 1.55 1.55 1.53
3.10 3.00 2.60 2.50 2.00 1.90 1.78 1.68 1.66 1.66
234 108 123 89
269 89
104 47
462 132
2.773 2.666 2.205 2.169 1.665 1.535 1.427 1.339 1.488 1.366
3.193 3.166 2.705 2.664 1.965 1.999 1.892 1.819 1.708 1.761
*Parameters from Soundararajan and Palanivel (1997) **Parameters from Table 2.1
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.
Step 1: Start with normal inspection. During normal inspection, take a random
Sample of size n, say nXXX ..., 21 and compute
1
X Lv
and
2
U Xv
, where
n
iiX
nX
1
1
Step 2: Accept the lot if 1 1Nv k and 2 2Nv k . Reject the lot if 1 1Nv k and/or
2 2Nv k . If a lot is rejected on normal inspection, then switch to tightened
inspection as in Step 3, otherwise continue the normal inspection.
Step 3: During tightened inspection, take a random sample of size n, say
nXXX ..., 21 and compute
1
X Lv
and
2
U Xv
, where
n
iiX
nX
1
1 .
55
Step 4: Accept the lot if 1 1Tv k and 2 2Tv k . Reject the lot if 1 1Tv k and/or
2 2Tv k 1 1 2 2( )T N T Nk k and k k . 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 1 1 2 2( , , , )T N T Nn k k k and k . If 1 1 1 ( ) T Nk k k say and
2 2 2 ( )T Nk k k 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, 2/)( UL , and if
1 2N Nk k and 1 2T Tk k , 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=-∞
( 1v ) or U=∞ ( 2v ). 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
)()(1
)()(
pPpPpP
pPTN
Ta
(3.1)
Under double specification limits, the above can be written as
),(),(1
),()(22112211
2211
TTNN
TTa kvkvPkvkvP
kvkvPpP
(3.2)
56
where ),( 2211 NNN kvkvPP is the probability of accepting a lot under normal
inspection based on a sample with parameters 1 2, ,N Nn k k and
),( 2211 TTT kvkvPP is the probability of accepting a lot under tightened single
sampling plan with parameters 1 2, ,T Tn k 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 distributions of 1v and 2v 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
}{}{ UXPLXPp (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 1v and 2v 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
2/}{}{ pUXPLXP (3.4)
Then, for the specified fraction non-conforming p,
Uz p 2/ (3.5)
57
Lzz pp 2/2/1 (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
NNN kkk 21 and TTT kkk 21 . In this case, the design parameters will be
only n , Nk and Tk . Therefore, the probability of acceptance under tightened
inspection TP for the specified p reduces to
pkvkvPP TTT 21 , = pkUXkLP TT
1))((2 2/ nkzP TpT (3.7)
where )( is the cumulative standard normal distribution function and /2T pk z .
Similarly, the probability of acceptance under normal inspection NP for the specified p will be determined by
)])((1[2 2/ nkzP NpN (3.8)
where / 2N pk z .
Hence, the probability of acceptance of the proposed VQSS under double
specification limits is given by
1))((2))((2
1)])(([2)(
2/2/
2/
nkznkznkz
pPNpTp
Tpa
(3.9)
As mentioned earlier, the above OC function reduces to the OC
function of the variables single sampling plan with double specification limits
when kkk NT . Hence when kkk NT (3.9) becomes
58
1))((2)( 2/ nkzpP pa (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
LpLXP }{ and UpUXP }{ such that ( ppp UL ) (3.11)
Then,
Uz
Up (3.12)
and
Lzz
LL pp1 (3.13)
Therefore, the probability of acceptance under tightened inspection for the
specified p reduces to
pkvkvPP TTT 2211 , = pkUXkLP TT 21
))(())(( 12 nzknkzPLU pTTpT (3.14)
Similarly, the probability of acceptance based on normal inspection for the
specified p becomes
))(())((1 12 nzknkzPLU pNNpN (3.15)
It should be noted that the condition ofUpTN zkk 22 and
LpTN zkk 11 are
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
)])(())(([])(())(([1
))(())(()(
1212
12
nzknkznzknkz
nzknkzpP
LULU
LU
pTTppNNp
pTTpa
(3.16)
The above function reduces to the OC function of the single sampling plan with
double specifications when 111 kkk NT (say) and 222 kkk NT (say). In this
case, (3.16) reduces to
))(())(()( 12 nzknkzpPLU ppa (3.17)
where n is the sample size of the variables single sampling plan and 1k and 2k 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.
Step 1: Start with normal inspection. During normal inspection, take a random
sample of size n, say 1 2, ... nX X X and compute 1
X Lv
S
and
2
U Xv
S
, where 1
1 n
ii
X Xn
and 2
1
11
n
ii
S X Xn
.
60
Step 2: Accept the lot if NSkv 11 and NSkv 22 . Reject the lot if NSkv 11 and/or
NSkv 22 . If a lot is rejected on normal inspection, then switch to tightened
inspection as in Step 3, otherwise continue the normal inspection.
Step 3: During tightened inspection, take a random sample of size n, say
1 2, ... nX X X and compute 1
X Lv
S
and 2
U Xv
S
, where
1
1 n
ii
X Xn
and 2
1
11
n
ii
S X Xn
.
Step 4: Accept the lot if TSkv 11 and TSkv 22 . Reject the lot if TSkv 11 and/or
TSkv 22 ( NSTSNSTS kkandkk 2211 ). 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 , NSk1 , NSk2 , TSk1 and TSk2 .
Here we consider only the symmetric case for the purpose of explaining
the designing methodology. That is, it will be assumed that NSNSNS kkk 21 and
TSTSTS kkk 21 .
Then, the probability of accepting a lot based on the tightened inspection becomes
pSkUXSkLPP TSTST
pLSkXPpUSkXP TSTS (3.18)
61
If we use the result that kSX (for a constant k) is approximately distributed as
normal as follows (see Duncan (1986)):
nk
nkNkSX
2,~
222 (3.19)
Then the equation (3.18) reduces to
12
2)(2 22/
TSTSpT k
nkzP (3.20)
Similarly, the probability of accepting a lot under normal inspection can be
obtained similarly as follows:
22/ 2
2)(22NS
NSpN knkzP (3.21)
The condition of 2/pTSNS zkk 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 SNSTS kkk .
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 producer’s risk α
and the consumer’s risk β 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-α, whereas the probability should be less than β when the quality is at the
specified LQL.
If the AQL and LQL are designated as 1p and 2p , respectively, then the
probability of acceptance should satisfy the following conditions.
1( ) 1aP p
2( )aP p (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 1p or at 2p , 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
1( ) 1aP p
2( )aP p
1n , 1 22 20 min( , )
U UN T p pk k z z
1 21 10 min( , )
L LN T p pk k z z (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 1p =0.01, 2p =0.05, respectively. Then, Table 3.1 gives the parameters as n =4,
559.1Nk and 932.1Tk .
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, Lp and Up 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
1p =0.01 and 2p =0.05 at =0.05 and at =0.1 respectively. Then, Table 3.2 gives
the optimal parameters as n =4, 759.11 Nk , 443.12 Nk , 168.21 Tk
and 802.12 Tk .
For the above example, the system operates as follows.
Step 1: Start with normal inspection. During normal inspection, take a random
sample of size 4. Now, compute 1
X Lv
and
2
U Xv
, where
4
141
iiXX .
Step 2: Accept the lot if 759.11 v and 443.12 v . Reject the lot if 759.11 v and/or
443.12 v . 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
1
X Lv
and
2
U Xv
, where
4
141
iiXX .
Step 4: Accept the lot if 168.21 v and 802.12 v . Reject the lot if 168.21 v
and/or 802.12 v . 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 substrate’s thickness is 0.77 mm, the
lower specification limit (L) of a glass substrate’s 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, 632.1TSk and 329.1NSk . 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
For this data, calculate 681424.011
Sn
iiX
nX and 041264.0
1)( 2
nXX
S i .
Also calculate 041264.0051424.0)(
1
S
LXv = 1.24622
and 041264.0088576.0)(
2
S
XUv = 2.14657
Step 2: Even though 329.114657.22 NSkv but, 329.124622.11 NSkv , 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
For this data, calculate 68203.011
Sn
iiX
nX and 043865.0
1)( 2
nXX
S i .
Also calculate 043865.005203.0)(
1
S
LXv = 1.18614
and 043865.008797.0)(
2
S
XUv = 2.00547
Step 2: Since 632.100547.22 TSkv but, 632.118614.11 TSkv , 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 Specification Limits (Symmetric Fraction Non-conforming)
p1
p2
Optimal Parameters
n kTσ kNσ 0.001
0.0025
0.005
0.010
0.03
0.002 0.003 0.004 0.006 0.008 0.010 0.015 0.020
0.005 0.010 0.015 0.020 0.025 0.030 0.050
0.010 0.015 0.020 0.030 0.040 0.050 0.10 0.02 0.03 0.04 0.05 0.10 0.15 0.20
0.060 0.090 0.120
42 17 10 6 4 3 2 2
34 9 5 3 3 2 1
30 12 7 4 3 2 1 25 10 6 4 2 1 1
21 7 4
3.082 2.948 2.863 2.719 2.617 2.538 2.391 2.278
2.803 2.548 2.401 2.289 2.190 2.132 1.918
2.563 2.408 2.303 2.139 2.024 1.918 1.591 2.312 2.149 2.028 1.932 1.603 1.389 1.221
1.872 1.669 1.528
2.972 2.789 2.648 2.471 2.283 2.117 1.853 1.868
2.682 2.313 2.119 1.848 1.862 1.579 0.953
2.439 2.218 2.043 1.802 1.642 1.368 0.761 2.182 1.943 1.748 1.559 1.153 1.002 0.123
1.723 1.402 1.149
72
Table 3.1 Contd….
p1
p2
Optimal Parameters
n kTσ kNσ 0.03
0.05
0.150 0.300 0.100 0.150
0.200 0.250 0.500
3 1 16 6
4 1 1
1.401 0.991 1.632 1.408
1.249 1.088 0.591
1.001 0.121 1.452 1.128
0.963 0.072 0.032
73
Table 3.2 Optimal Parameters of Known Sigma VQSS with Double Specification Limits (Asymmetric Fraction Non-conforming)
p1
p2
Optimal Parameters n k1Tσ k2Tσ k1Nσ k2Nσ
0.001
0.0025
0.005
0.01
0.03
0.002 0.003 0.004 0.006 0.008 0.010 0.015 0.020
0.005 0.010 0.015 0.020 0.025 0.030 0.050
0.010 0.015 0.020 0.030 0.040 0.050 0.10
0.02 0.03 0.04 0.05 0.10 0.15 0.20
0.060 0.090 0.120
42 16 10 6 4 3 2 2
35 8 5 3 3 2 1
30 11 7 4 3 2 1
26 10 6 4 2 1 1
18 7 4
3.272 3.143 3.048 2.912 2.818 2.742 2.589 2.468
3.001 2.758 2.612 2.498 2.414 2.343 2.128
2.782 2.629 2.521 2.358 2.239 2.142 1.819
2.553 2.379 2.261 2.168 1.839 1.632 1.456
2.134 1.939 1.789
2.968 2.842 2.753 2.608 2.513 2.429 2.294 2.172
2.674 2.428 2.291 2.179 2.088 2.022 1.809
2.432 2.289 2.178 2.011 1.889 1.798 1.477
2.169 2.012 1.889 1.802 1.468 1.262 1.089
1.712 1.509 1.368
3.163 2.967 2.828 2.642 2.449 2.281 2.013 2.029
2.884 2.498 2.312 2.019 2.042 1.749 1.093
2.653 2.399 2.242 1.993 1.828 1.543 0.902
2.399 2.152 1.958 1.759 1.342 0.668 0.713
1.939 1.642 1.373
2.859 2.678 2.548 2.363 2.182 2.018 1.757 1.779
2.563 2.198 2.012 1.748 1.763 1.488 0.893
2.312 2.069 1.918 1.682 1.533 1.269 0.689
2.039 1.812 1.629 1.443 1.039 0.448 0.469
1.542 1.258 1.019
74
Table 3.2 Contd….
p1
p2
Optimal Parameters n k1Tσ k2Tσ k1Nσ k2Nσ
0.03
0.05
0.150 0.300
0.100 0.150 0.200 0.250 0.500
3 1
15 5 3 2 2
1.668 1.259
1.913 1.698 1.542 1.397 0.862
1.252 0.839
1.458 1.253 1.079 0.949 0.432
1.218 0.259
1.692 1.283 1.002 0.708 0.838
0.868 0.053
1.269 0.903 0.648 0.402 0.409
75
Table 3.3 Optimal Parameters of Unknown Sigma VQSS with Double Specification Limits (Symmetric Fraction Non-conforming)
p1
p2
Optimal Parameters
n kTS kNS 0.001
0.0025
0.005
0.010
0.002 0.003 0.004 0.006 0.008 0.010 0.015 0.020
0.005 0.010 0.015 0.020 0.025 0.030
0.050 0.010 0.015 0.020 0.030 0.040 0.050
0.10 0.02 0.03 0.04 0.05 0.10 0.15 0.20
200 72 42 23 15 12 8 6
147 30 16 11 8 6
4 112 39 22 12 8 6
3 83 28 16 11 4 3 2
3.084 2.952 2.848 2.712 2.608 2.528 2.383 2.264
2.788 2.548 2.401 2.278 2.193 2.122
1.892 2.562 2.407 2.301 2.129 2.012 1.911
1.569 2.312 2.149 2.022 1.918 1.576 1.358 1.188
2.859 2.627 2.473 2.289 2.052 2.029 1.903 1.772
2.568 2.147 1.949 1.823 1.649 1.338
1.419 2.319 2.059 1.863 1.714 1.548 1.432
1.188 2.062 1.767 1.612 1.489 1.088 1.067 0.849
76
Table 3.3 Contd….
p1
p2
Optimal Parameters
n kTS kNS 0.03
0.05
0.060 0.090 0.120 0.150 0.300
0.100 0.150 0.200 0.250 0.500
46 15 8 5 2
33 10 6 4 2
1.857 1.669 1.523 1.404 0.959
1.632 1.408 1.244 1.103 0.591
1.578 1.269 1.018 0.690 0.528
1.329 0.879 0.853 0.690 0.482
77
Table 3.4 Optimal Parameters of Unknown Sigma VQSS with Double Specification Limits (Asymmetric Fraction Non-conforming)
p1
p2
Optimal Parameters n k1TS k2TS k1NS k2NS
0.001
0.0025
0.005
0.01
0.03
0.002 0.003 0.004 0.006 0.008 0.010 0.015 0.020
0.005 0.010 0.015 0.020 0.025 0.030 0.050
0.010 0.015 0.020 0.030 0.040 0.050 0.10
0.02 0.03 0.04 0.05 0.10 0.15 0.20
0.060 0.090
228 80 44 24 15 12 8 6
162 30 16 10 8 7 4
126 43 22 12 9 6 3
90 31 16 12 5 3 2
48 16
3.262 3.129 3.028 2.892 2.793 2.712 2.549 2.438
2.989 2.738 2.584 2.469 2.384 2.303 2.084
2.768 2.609 2.521 2.334 2.219 2.102 1.769
2.533 2.362 2.229 2.128 1.789 1.572 1.388
2.113 1.913
2.969 2.848 2.758 2.628 2.528 2.449 2.229 2.187
2.678 2.452 2.291 2.193 2.108 2.029 1.809
2.442 2.229 2.188 2.034 1.913 1.819 1.477
2.192 2.028 1.912 1.808 1.488 1.268 1.113
1.722 1.532
3.139 2.937 2.793 2.572 2.442 2.318 2.073 1.909
2.864 2.469 2.242 2.098 1.958 1.828 1.469
2.623 2.393 2.214 1.979 1.822 1.687 1.109
2.369 2.108 1.923 1.779 1.359 0.978 1.018
1.919 1.603
2.878 2.698 2.564 2.369 2.242 2.129 1.903 1.753
2.583 2.229 2.028 1.943 1.768 1.648 1.313
2.332 2.116 1.968 1.762 1.613 1.489 0.958
2.069 1.839 1.669 1.553 1.149 0.812 0.818
1.594 1.308
78
Table 3.4 Contd….
p1
p2
Optimal Parameters n k1TS k2TS k1NS k2NS
0.03
0.05
0.120 0.150 0.300 0.100 0.150
0.200 0.250 0.500
8 5 2
35 12
6 4 2
1.749 1.618 1.179 1.849 1.659
1.478 1.332 0.786
1.388 1.268 0.859 1.473 1.268
1.112 0.969 0.468
1.378 1.182 0.752 1.679 1.348
1.132 0.957 0.658
1.119 0.948 0.532 1.329 1.049
0.848 0.702 0.369
79
Table 3.5. Average Sample Number of Variables Single Sampling Plans and VQSS with Double Specification Limits (Symmetric Fraction Non- conforming)
p1
p2
Average Sample Number Known Sigma Unknown Sigma
SSP VQSS SSP VQSS 0.001
0.0025
0.005
0.010
0.002 0.003 0.004 0.006 0.008 0.010 0.015 0.020
0.005 0.010 0.015 0.020 0.025 0.030 0.050
0.010 0.015 0.020 0.030 0.040 0.050 0.10
0.02 0.03 0.04 0.05 0.10 0.15 0.20
84 33 20 12 9 7 5 4
72 17 10 7 6 5 3
63 24 15 9 6 5 3
55 21 13 9 4 3 3
42 17 10 6 4 3 2 2
34 9 5 3 3 2 1
30 12 7 4 3 2 1
25 10 6 4 2 1 1
481 173 101 54 37 28 18 13
353 72 38 25 19 15 9
270 94 53 28 18 14 6
199 68 38 26 9 6 4
200 72 42 23 15 12 8 6
147 30 16 11 8 6 4
112 39 22 12 8 6 3
83 28 16 11 4 3 2
80
Table 3.5 Contd….
p1
p2
Average Sample Number Known Sigma Unknown Sigma
SSP VQSS SSP VQSS 0.03
0.05
0.060 0.090 0.120 0.150 0.300
0.100 0.150 0.200 0.250 0.500
41 15 9 7 3
34 13 8 6 3
21 7 4 3 1
16 6 4 1 1
110 36 20 13 4
79 25 13 9 3
46 15 8 5 2
33 10 6 4 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 didn’t 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.
4.2 CONDITIONS OF APPLICATION
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.
In addition, the usual conditions for the application of variables single sampling
plan with known or unknown standard deviation should also be valid (see for
further details Schilling (1982), Grant and Leavenworth (1996) and Montgomery
(2005)).
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 σ.
Step 1: Start with the tightened inspection level using the single sampling
variables plan with a sample size n1 and the acceptance criterion k.
Accept the lot if kv and reject the lot if kv , where
1XUv and
1
111
1 n
iiX
nX .
Step 2: If t lots in a row are accepted under tightened inspection, then switch to
normal inspection.
Step 3: During the normal inspection, inspect the lots using the single sampling
variables plan with a sample size n2 (<n1) and the acceptance criterion k.
83
Accept the lot if kv and reject the lot if kv , where
2XUv and
2
122
1 n
iiX
nX .
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
12)(PP
pPa (4.1)
84
where 1s ,)1)(1(
2
22
2
s
s
PPP
s t , )1(
1
11
1
PPP
t
t
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.
The fraction non-conforming in a lot will be determined as
)()(11 vvUp
(4.2)
where )( y is given by
y
dzzy2
exp21)(
2
, (4.3)
provided that the quality characteristic of interest is normally distributed with
mean µ and standard deviation σ, and the unit is classified as non-conforming if it
exceeds the upper specification limit U.
Then its probability of acceptance is written as
)()(
)( 12 wwpPa (4.4)
where
1s ,))(1))((1(
)(2
22
2
s
s
www
85
s t ,))(1()(
)(1
11
1
www
t
t
Here mnkvw 1 and nkvw 2 where mnn 1 and nn 2
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 producer’s
risk α and the consumer’s risk β are prescribed then the OC function can be
expressed as
1
)()( 1121 ww (4.5)
and
)()( 1222 ww (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,
mnkvw )( 111 , nkvw )( 121
mnkvw )( 212 and nkvw )( 222
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
mnnpASN )( , m>1 (4.7)
where 1s ,)1)(1(
2
22
2
s
s
PPP
s t ,)1(
1
11
1
PPP
t
t
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
1)( 1pPa
)( 2pPa
0,1 ,1 k mn (4.8)
where )( 1pPa and )( 2pPa 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 σ. In this case, the operation of the scheme is as
follows.
Step 1: Start with the tightened inspection level using the single sampling
variables plan with a sample size n1 and the acceptance criterion kS.
Accept the lot if Skv and reject the lot if Skv , where S
XUv 1 ,
Sn
ii
S
Xn
X1
111
1 and 1
)(
1
21
S
i
nXX
S .
Step 2: If t lots in a row are accepted under tightened inspection, then switch to
normal inspection.
Step 3: During the normal inspection, inspect the lots using the single sampling
variables plan with a sample size n2σ (<n1σ) and the acceptance criterion kS.
Accept the lot if Skv and reject the lot if Skv , where S
XUv 2 ,
Sn
ii
S
Xn
X2
122
1 and 1
)(
2
22
S
i
nXX
S .
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
nkns 21
2
(4.9)
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
)54)44(
s
sS n
nkk (4.10)
Also the ASN of unknown sigma TNT scheme namely ASNs can be determined as
21
2s
SkASNASN (4.11)
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 nσ, kσ 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
producer’s and consumer’s 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 n2σ=n=5 and the
tightened plan sample size is obtained as n1σ=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
Compute
XUv , where
25
1251
iiXX . Accept the lot if 598.1v and
reject the lot if 598.1v . 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
XUv , where
5
151
iiXX . Accept the lot if 598.1v and reject the
lot if 598.1v . 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) nσ = 26, kσ = 1.73 and ASN=26
(iii) n1σ = n2σ = 19, kaσ = 1.83, krσ = 1.67 and ASN=21.2
(iv) n = 13, kσ = 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 nσ = 12, m = 5.25 and kσ = 1.857 along with two
variables single sampling plans (nσ =63, kσ =1.857) and (nσ = 12, kσ =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-α = 0.95) and (p2 = 0.045, β = 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 producer’s 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 producer’s and consumer’s 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
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n m kσ ASNσ nS m kS ASNS 0.001
0.005
0.01
0.02
0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010
0.007 0.008 0.009 0.01
0.012 0.015 0.020 0.03
0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
0.03 0.04 0.05 0.06 0.07 0.08
97 36 22 16 13 12 9 8 8
294 153 94 67 42 27 16 9
56 22 13 9 7 6 5 5 4
140 46 26 18 13 10
7.75 9.0 9.25 9.0 8.0 5.25 9.5 9.25 6.0
10.0 8.75 9.75 10.0 8.5 7.25 7.75 8.0
10.0 7.75 8.5 9.5 9.5 8.0 8.75 5.0 7.75
9.5 9.25 8.0 7.0 8.0 9.5
2.925 2.819 2.742 2.683 2.638 2.619 2.548 2.515 2.512
2.481 2.444 2.408 2.376 2.325 2.262 2.169 2.032
2.108 1.979 1.873 1.784 1.712 1.661 1.599 1.598 1.512
1.916 1.813 1.734 1.669 1.602 1.537
123.244 47.538 29.121 21.100 16.605 14.008 12.071 10.647 9.558
400.85 200.36 127.01 90.486 54.707 33.683 20.327 11.487
75.757 27.914 16.809 12.067 9.399 7.668 6.557 5.798 5.076
187.378 60.920 33.213 22.251 16.626 13.324
512 179 105 74 58 53 38 33 33
1199 610 367 256 156 96 54 28
180 65 36 23 17 14 11 11 9
397 122 65 43 30 22
7.75 9.0
9.25 9.0 8.0
5.25 9.5
9.25 6.0
10.0 8.75 9.75 10.0 8.5
7.25 7.75 8.0
10.0 7.75 8.5 9.5 9.5 8.0
8.75 5.0
7.75
9.5 9.25 8.0 7.0 8.0 9.5
2.926 2.704 2.749 2.692 2.650 2.632 2.565 2.535 2.532
2.482 2.445 2.410 2.378 2.329 2.268 2.179 2.051
2.111 1.987 1.886 1.805 1.739 1.694 1.640 1.639 1.561
1.917 1.817 1.741 1.679 1.616 1.555
650.977 236.957 139.123 97.566 74.894 62.515 51.793 44.852 40.190
1635.09 797.959 495.754 346.408 203.051 120.313 68.599 35.649
244.553 83.005 46.720 31.713 21.613 18.663 15.374 13.584 11.258
531.761 161.455 83.536 53.613 38.332 29.444
95
Table 4.1. Contd….
p1
p2
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n m kσ ASNσ nS m kS ASNS 0.02 0.03
0.04
0.05
0.09 0.1 0.11 0.12
0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11 0.12 0.15
0.05 0.06 0.07 0.08 0.09 0.1 0.11 0.12 0.13 0.14 0.15
0.06 0.07 0.08 0.09 0.10 0.11 0.12 0.13
9 7 7 6
252 75 39 26 19 15 12 10 9 6
370 110 56 38 25 19 16 13 11 10 9
511 148 72 46 33 27 19 16
6.25 9.5 5.25 6.0 8.75 9.5 10.0 8.75 8.5 7.75 8.5 8.5 7.0 8.5
10.0 9.25 9.0 6.5 9.5 9.75 7.5 8.25 8.5 7.0 6.25
9.75 9.0 10.0 9.0 7.75 5.5 9.25 8.25
1.512 1.439 1.438 1.389
1.778 1.693 1.620 1.561 1.506 1.460 1.409 1.366 1.337 1.216
1.666 1.595 1.533 1.487 1.424 1.376 1.344 1.299 1.259 1.234 1.208
1.573 1.511 1.453 1.404 1.362 1.332 1.272 1.238
10.906 9.390 8.184 7.197
328.964 100.604 52.979 33.970 24.577 19.018 15.515 12.996 11.131 7.794
502.286 145.605 73.817 46.417 33.323 25.517 20.170 16.755 14.262 12.347 10.886
691.102 195.102 97.839 60.393 41.982 31.890 25.316 20.639
19 14 14 12
650 182 90 58 41 31 24 33 17 10
883 250 122 80 50 37 30 24 20 18 16
1143 317 148 91 64 51 34 28
6.25 9.5
5.25 6.0
8.75 9.5
10.0 8.75 8.5
7.75 8.5 8.5 7.0 8.5
10.0 9.25 9.0 6.5 9.5
9.75 7.5
8.25 8.5 7.0
6.25
9.75 9.0
10.0 9.0
7.75 5.5
9.25 8.25
1.533 1.467 1.466 1.421
1.779 1.695 1.625 1.568 1.515 1.472 1.424 1.385 1.358 1.251
1.666 1.597 1.536 1.492 1.431 1.386 1.356 1.313 1.276 1.252 1.228
1.573 1.512 1.455 1.408 1.367 1.339 1.282 1.250
23.726 19.496 16.981 14.467
849.342 245.182 122.889 75.723 52.798 39.630 31.255 25.464 21.398 13.891
1199.73 331.187 160.914
98.06 67.455 50.011 38.704 31.203 25.868 22.03
19.101
1546.47 418.172 201.468 120.248 81.231 60.464 46.109 36.751
96
Table 4.1. Contd….
p1
p2
MinASN(p1) Known Sigma
MinASN(p1) Unknown Sigma
n m kσ ASNσ nS m kS ASNS 0.05
0.06
0.07
0.14 0.15
0.08 0.09 0.10 0.11 0.12 0.13 0.14 0.15
0.09 0.10 0.11 0.12 0.13 0.14 0.15 0.20
13 12
184 91 61 41 30 24 19 16
229 112 75 46 34 29 22 11
10.0 7.25
10.0 9.75 6.5
7.25 8.0 7.5
8.75 8.25
9.75 9.0 6.0
10.0 9.75 6.5
8.75 7.5
1.193 1.174
1.435 1.384 1.346 1.301 1.258 1.222 1.180 1.148
1.368 1.322 1.287 1.235 1.197 1.174 1.129 0.983
17.650 14.961
250.769 122.634 74.263 51.333 38.452 30.177 24.763 20.645
308.023 147.810 89.608 62.231 45.974 35.430 28.840 13.792
22 20
373 178 116 76 54 42 32 27
443 210 137 81 58 49 36 16
10.0 7.25
10.0 9.75 6.5 7.25 8.0 7.5 8.75 8.25
9.75 9.0 6.0 10.0 9.75 6.5 8.75 7.5
1.207 1.190
1.436 1.386 1.349 1.305 1.264 1.229 1.896 1.159
1.369 1.324 1.289 1.239 1.202 1.180 1.137 0.9996
30.514 25.548
509.305 240.413 141.827 95.065 69.167 52.985 42.284 34.514
596.562 277.280 164.092 109.986 79.200 60.101 47.484 20.682
97
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Fraction Nonconforming, p
Prob
abili
ty o
f Acc
epta
nce,
Pa(
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)
Variables Normal SSP
Variables TNT Scheme
Variables Tightened SSP
98
Table 4.2: ASN Values of the Known Sigma Variables Single Sampling Plan, Variables Double Sampling Plan and Variables TNT Scheme
p1
p2
ASN Variables
SSP Variables
DSP Variables*
TNT Variables **
TNT 0.001 0.001 0.005 0.005 0.03 0.03 0.04 0.04 0.05 0.05
0.002 0.003 0.010 0.012 0.04 0.06 0.07 0.08 0.07 0.08
191 74 138 85 506 81 114 72 300 149
154.9 59.4
112.0 69.5
434.6 127.7 180.6 58.4
246.7 122.3
179.792 68.673 126.310 76.809 471.546 72.746 108.116 68.675 271.860 137.421
123.244 47.538 90.486 54.707 328.964 52.979 73.817 46.417 195.102 97.839
* ASN given in Senthilkumar and Muthuraj (2010) ** ASN given in Table 4.1 Table 4.3: ASN Values of the Unknown Sigma Variables Single Sampling Plan, Variables Double Sampling Plan and Variables TNT Scheme
p1
p2
ASN Variables
SSP Variables
DSP Variables*
TNT Variables **
TNT 0.001 0.001 0.005 0.005 0.03 0.03 0.04 0.04 0.05 0.05
0.002 0.003 0.010 0.012 0.04 0.06 0.07 0.08 0.07 0.08
1032 381 547 327
1333 197 258 159 660 319
829.1 302.4 437.1 263.0 1138.7 156.5 205.0 125.8 535.4 258.0
971.031 355.665 500.564 296.943
1238.397 176.451 245.252 151.353 598.751 294.171
650.977 236.957 346.408 203.051 849.342 122.889 160.914
98.06 418.172 201.468
* ASN given in Senthilkumar and Muthuraj (2010) ** ASN given in Table 4.1
99
Table 4.4: Parameters of Known Sigma Variables TNT Scheme for different AQL and LQL Values
p1
p2
*Parameters of Variables TNT
**Parameters of Variables TNT
n m kσ ASN n m kσ ASN 0.001 0.005 0.01 0.02 0.03 0.04 0.05 0.06 0.07
0.002 0.007 0.020 0.03 0.04 0.06 0.07 0.08 0.10
146 451 83
224 387 163 224 328 163
2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
2.955 2.503 2.150 1.941 1.796 1.621 1.541 1.460 1.351
151.596 469.248 86.288 233.170 401.699 169.272 233.170 341.510 169.301
97 294 56 140 252 110 148 184 112
7.75 10.0 10.0 9.5 8.75 9.25 9.0 10.0 9.0
2.925 2.481 2.108 1.916 1.778 1.595 1.511 1.435 1.322
123.244 400.85 75.757 187.378 328.964 145.605 195.102 250.769 147.810
*Parameters from Senthilkumar and Muthuraj (2010) **Parameters from Table 4.1 Table 4.5: Parameters of Unknown Sigma Variables TNT Scheme for different AQL and LQL Values
p1
p2
*Parameters of Variables TNT
**Parameters of Variables TNT
nS m kS ASN nS m kS ASN 0.001 0.005 0.01 0.02 0.03 0.04 0.05 0.06 0.07
0.002 0.007 0.020 0.03 0.04 0.06 0.07 0.08 0.10
728 265 83
224 387 163 224 328 163
2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0 2.0
2.951 2.503 2.150 1.941 1.796 1.621 1.541 1.460 1.351
771.730 469.248 86.288 233.170 401.699 169.272 233.170 341.510 169.301
97 294 56 140 252 110 148 184 112
7.75 10.0 10.0 9.5 8.75 9.25 9.0 10.0 9.0
2.925 2.481 2.108 1.916 1.778 1.595 1.511 1.435 1.322
123.244 400.85 75.757 187.378 328.964 145.605 195.102 250.769 147.810
*Parameters from Senthilkumar and Muthuraj (2010) **Parameters from Table 4.1
100
CHAPTER 5
OPTIMAL DESIGNING OF
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
5.2 CONDITIONS OF APPLICATION
The following assumptions should be valid for the application of the
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
plans with known or unknown standard deviation should also be valid.
5.3 OPERATING PROCEDURES OF VARIABLES ChSP 5.3.1 KNOWN SIGMA CASE
Suppose that the quality characteristic of interest has the upper
specification limit U and follows a normal distribution with unknown mean μ and
known standard deviation σ. 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
nXXX ..., 21 and compute
XUv
, where
n
iiX
nX
1
1 .
Step 2: Accept the lot if kv and reject the lot if 'kv . If kvk ' , then
accept the current lot provided that the preceding i lots were accepted on
the condition that kv but reject, otherwise. (Note: 'kk ).
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 nXXX ..., 21 and
compute
LXv , where
n
iiX
nX
1
1 .
Step 2: Accept the lot if kv and reject the lot if 'kv . If kvk ' , then
accept the current lot provided that the preceding i lots were accepted on
the condition that kv but reject, otherwise.
Thus, the proposed variables chain sampling plan is characterized by
four parameters, namely kin ,, and 'k . If 'kk , 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 σ. In this case, the plan operates as follows.
103
Step 1: From each submitted lot, take a random sample of size Sn , say
SnXXX ..., 21 and compute
SXUv
, where
Sn
ii
S
Xn
X1
1 and
1
)( 2
S
i
nXX
S .
Step 2: Accept the lot if Skv and reject the lot if SkV ' . If SS kvk ' , then
accept the current lot provided that the preceding Si lots were accepted
on the condition that Skv .
Thus, the proposed unknown sigma variables chain sampling plan is
characterized by four parameters namely SSS kin ,, and Sk ' . If SS kk ' , 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 )1,( 1 p and ),( 2 p , where 1p is called the AQL, 2p is
the LQL, is the producer’s risk and is the consumer’s 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
Up 1 (5.1)
where )(Y is the cumulative distribution function of standard normal
distribution and is given by
Y
dZZY2
exp21)(
2
, (5.2)
Here the quality characteristic of interest is normally distributed with
mean µ and standard deviation σ, and the unit is classified as non-conforming if it
exceeds the upper specification limit U. So, the unknown mean μ can be
determined if p is specified. Let us define the standardized quality characteristic
corresponding to the fraction conforming as
)1(1 pZ 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)
ia kvkvkvkvpP )Pr( )Pr()'Pr()Pr()( (5.4)
105
where kv Pr is the probability of accepting a lot based on a single sample with
parameters ),( kn and 'Pr kv is the probability of rejecting a lot based on a
single sample with parameters )',( kn . 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
ia wwwwpP )( )()()()( 2212 (5.5)
where nkZw p '1 and nkZw p 2
The OC function given in (5.5) under the specified AQL and LQL conditions can
be written as
1)( )()()( 21211121
iwwww
and iwwww )( )()()( 22221222
Here 11w is the value of 1w at 1pp , 21w is the value of 2w at 1pp , 12w is
the value of 1w at 2pp and 22w is the value of 2w at 2pp . That is,
nkZw p )'(111 , nkZw p )(
121
nkZw p )'(212 and nkZw p )(
222
where 1pZ is the value of pZ at AQL and
2pZ is the value of pZ at LQL.
106
The parameters of the known sigma variables chain sampling plan are
denoted by ( kkn ,', ).The following optimization problem is considered to
determine the optimal parameters of known sigma variables sampling plan such
as kin ,, and 'k .
Minimize npASN )( 1
Subject to 1)( 1pPa
)( 2pPa
,0,0' ,1i ,2 k kn n ЄN, i ЄN (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 ( SSS kkn ,', ) of unknown sigma
variables chain sampling plan is explained as follows. It is known that for large
samples, SkX S is approximately normally distributed with mean )(SEkS
and variance )(2
SVarkn S (see Duncan (1986), Balamurali and Jun (2007)).
That is,
nk
nkNSkX SSS 2
,~2
22
Therefore, the probability of accepting a lot at each repetition is given by
pSkUXPkvP SS )(
pUSkXP S
107
21)/(
2S
S
S
kn
kU
21
)( 2S
SSp k
nkZ (5.7)
If we let,
21
)( 22S
SSpS k
nkZw then )()Pr( 2SS wkv . (5.8)
Similarly if we let,
2'
1)'( 21
S
SSpS k
nkZw then we have
)(1)'Pr( 1SS wkv (5.9)
Hence the lot acceptance probability for sigma unknown case is given by
SiSSSSa wwwwpP )( )()()()( 2212 (5.10)
where SSpS nkZw '1 and SSpS nkZw 2
If (AQL, 1 ) and (LQL, ) are prescribed then the OC function can be written
as
1)( )()()( 21211121Si
SSSS wwww
and SiSSSS wwww )( )()()( 22221222 (5.11)
108
Here Sw11 is the value of Sw1 at 1pp , Sw21 is the value of Sw2 at 1pp , Sw12 is
the value of Sw1 at 2pp and Sw22 is the value of Sw2 at 2pp .
We obtain SSSS wwww 22122111 ,,, respectively by
2'
1)'( 2111
S
SSpS k
nkZw ,
21
)( 2121S
SSpS k
nkZw
2'
1)'( 2212
S
SSpS k
nkZw and
21
)( 2222S
SSpS k
nkZw (5.12)
where 1pZ is the value of pZ at AQL and
2pZ is the value of pZ at LQL.
For given AQL and LQL, the parametric values of the unknown sigma
variables chain sampling plan namely SSS kin ,, and Sk ' 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 SnpASN )( 1
Subject to 1)( 1pPa
)( 2pPa
,0,0' ,1 ,2 SSSS k kin Sn ЄN, Si ЄN (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
producer’s and consumer’s 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 ( kin ,, and
'k ) for known sigma plan and the parameters ( SSS kin ,, and Sk ' ) 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 1p = 2%, 2p = 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
XUv , where
18
1181
iiXX . Accept the lot if 779.1v and reject the lot if
544.1v . If 779.1544.1 v , then accept the current lot provided that the
preceding 3 lots were accepted on the condition that 779.1v 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
SSS kin ,, and Sk ' are used in the place of kin ,, 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 Sn = 47, Si = 1, Sk ' =
1.925 and Sk = 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.
Step 2: For this data, we calculate 708915.011
Sn
iiX
nX and
017583.01
)( 2
nXX
S i .
Step3: Calculate 4741.3017583.0
)708915.077.0()(
SXUv
Step 4: Since 190.24741.3 Skv , 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 925.1' Skv .
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 producer’s risk (α) and consumer’s risk (β). 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
α=1% and β=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 )1,( 1 p and ),( 2 p . Suppose that for given values of
1p =0.5%, =5%, 2p =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, ak = 2.41 and rk = 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 consumer’s 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 producer’s 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
)(1)()( 1 pPpPLf aL
aG (5.17)
Its mean and variance are respectively given by
)(11)(
pPLEARL
a (5.18)
2)(1
)()(
pPpPLVar
a
a
(5.19)
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 iσ values, when iσ 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%
1p
2p
Known Sigma Unknown Sigma
n i 'k k Sn Si Sk ' Sk 0.001
0.0025
0.005
0.0020 0.0025 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010 0.012 0.015
0.004 0.005 0.006 0.0075 0.010 0.012 0.015 0.020 0.025 0.030 0.035
0.0075 0.010 0.012 0.015 0.020 0.025 0.030 0.035 0.040 0.050
119 68 46 31 21 19 15 13 12 12 9 7
230 105 64 39 26 19 16 12 10 8 7
259 86 53 33 24 16 14 12 10 8
1 2 1 2 2 4 3 3 3 6 4 3
2 2 2 2 3 2 4 4 6 5 6
1 1 1 1 4 3 6 6 4 5
2.883 2.810 2.776 2.740 2.610 2.597 2.534 2.495 2.521 2.517 2.331 2.237
2.658 2.598 2.530 2.441 2.379 2.335 2.296 2.260 2.206 2.114 2.307
2.434 2.326 2.277 2.211 2.161 2.046 1.989 2.002 1.976 1.893
3.013 2.965 2.961 2.885 2.860 2.807 2.789 2.765 2.736 2.697 2.686 2.657
2.738 2.703 2.675 2.641 2.579 2.555 2.491 2.425 2.366 2.334 2.297
2.524 2.486 2.457 2.421 2.316 2.281 2.224 2.182 2.156 2.098
650 361 242 147 103 84 72 62 51 45 40 34
1131 474 282 171 106 76 55 38 32 29 25
1113 347 211 128 75 55 43 36 31 25
1 1 1 1 1 2 3 3 2 2 3 4
2 2 1 1 1 1 1 1 2 5 5
2 1 1 1 2 2 3 3 3 4
2.888 2.834 2.774 2.716 2.637 2.599 2.532 2.518 2.470 2.425 2.414 2.387
2.676 2.585 2.533 2.464 2.430 2.321 2.237 2.139 2.163 2.107 2.092
2.442 2.348 2.299 2.233 2.100 2.083 1.974 1.979 1.964 1.925
3.013 2.984 2.964 2.921 2.897 2.829 2.792 2.768 2.765 2.750 2.694 2.637
2.736 2.705 2.698 2.669 2.610 2.601 2.572 2.529 2.413 2.342 2.302
2.512 2.483 2.454 2.418 2.345 2.293 2.249 2.209 2.174 2.105
117
Table 5.1. Contd….
1p
2p
Known Sigma Unknown Sigma
n i 'k k Sn Si Sk ' Sk 0.0075
0.010
0.02
0.010 0.012 0.015 0.020 0.025 0.030 0.035 0.040
0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050 0.060 0.070
0.030 0.035 0.040 0.045 0.050 0.060 0.070 0.080 0.090 0.10 0.11 0.12 0.13
505 173 78 43 29 19 17 13
248 74 42 27 21 18 15 13 11 9
187 96 59 45 34 22 18 14 13 10 10 7 8
2 1 1 2 5 2 5 3 1 1 1 1 2 3 3 3 5 5 2 2 1 2 2 2 3 3 4 2 4 1 5
2.339 2.259 2.185 2.146 2.008 1.968 1.888 1.837
2.212 2.090 2.027 1.906 1.846 1.804 1.762 1.731 1.662 1.573
1.896 1.835 1.779 1.749 1.693 1.563 1.544 1.434 1.502 1.436 1.457 1.235 1.360
2.384 2.369 2.335 2.251 2.198 2.178 2.123 2.107
2.257 2.220 2.177 2.161 2.096 2.054 2.027 2.001 2.942 1.903
1.976 1.945 1.939 1.889 1.868 1.833 1.779 1.749 1.697 1.691 1.632 1.725 1.580
1817 690 286 133 83 61 47 41
786 247 135 93 66 53 46 39 30 26
528 272 169 117 88 63 44 36 27 22 23 17 17
1 2 1 1 1 1 1 3
1 1 1 2 1 2 3 2 3 4
1 2 2 1 1 2 2 3 2 1 4 1 4
2.329 2.278 2.182 2.083 2.003 1.984 1.925 1.851
2.187 2.071 2.011 1.954 1.900 1.846 1.811 1.824 1.730 1.728
1.886 1.831 1.770 1.734 1.683 1.670 1.577 1.525 1.423 1.386 1.476 1.379 1.356
2.394 2.353 2.337 2.293 2.258 2.214 2.190 2.111
2.267 2.226 2.186 2.124 2.125 2.066 2.026 2.004 1.955 1.898
1.991 1.946 1.920 1.919 1.903 1.820 1.792 1.745 1.737 1.766 1.636 1.699 1.596
118
Table 5.1. Contd….
1p
2p
Known Sigma Unknown Sigma
n i 'k k Sn Si Sk ' Sk 0.030
0.04
0.050
0.040 0.045 0.050 0.060 0.070 0.080 0.090 0.100 0.110 0.120 0.130 0.150
0.060 0.070 0.080 0.090 0.100 0.110 0.120 0.130 0.140 0.150 0.170
0.070 0.080 0.090 0.100 0.110 0.120 0.130 0.140 0.150 0.160 0.170
341 157 96 52 33 26 23 17 13 11 11 10
145 73 50 34 27 24 18 14 14 11 10
188 93 60 46 33 26 20 19 17 13 12
2 1 1 2 1 3 5 3 2 1 4 6 2 1 2 2 3 5 3 2 4 2 4 1 1 2 3 2 1 1 3 4 2 2
1.771 1.711 1.650 1.566 1.508 1.427 1.449 1.378 1.268 1.270 1.204 1.273
1.568 1.513 1.478 1.394 1.324 1.309 1.233 1.130 1.158 1.108 1.070
1.487 1.408 1.359 1.331 1.288 1.270 1.157 1.160 1.108 1.011 1.045
1.821 1.811 1.795 1.736 1.728 1.657 1.609 1.593 1.588 1.605 1.514 1.443
1.663 1.643 1.588 1.564 1.529 1.489 1.478 1.475 1.423 1.428 1.360
1.582 1.558 1.509 1.471 1.453 1.450 1.452 1.375 1.348 1.356 1.330
1027 416 253 126 81 62 44 36 29 26 23 18
339 166 102 71 53 42 34 31 27 23 17
427 203 127 86 63 49 45 34 30 25 24
5 2 1 1 1 2 2 2 2 3 3 3 2 1 1 1 1 1 1 2 3 3 2 2 1 2 1 1 1 3 2 2 1 3
1.771 1.704 1.679 1.585 1.539 1.514 1.401 1.384 1.305 1.274 1.289 1.230
1.564 1.507 1.425 1.374 1.315 1.294 1.234 1.271 1.193 1.130 1.061
1.485 1.422 1.365 1.331 1.261 1.213 1.235 1.148 1.165 1.134 1.112
1.816 1.799 1.789 1.760 1.724 1.659 1.641 1.609 1.590 1.549 1.519 1.475
1.664 1.647 1.625 1.600 1.580 1.554 1.539 1.461 1.433 1.415 1.396
1.570 1.557 1.510 1.506 1.491 1.473 1.395 1.393 1.365 1.384 1.312
119
Table 5.2:Variables Chain Sampling Plans Indexed by AQL and LQL for =1% and =1%
1p
2p
Variables Chain Sampling Plan (Known Sigma) n i 'k k )( 1pPa )( 2pPa ASN
0.001 0.0025 0.005 0.0075
0.01 0.02 0.03 0.04 0.05
0.005 0.01 0.02 0.03 0.04 0.05 0.06 0.08 0.10
79 91 77 69 64
124 193 172 156
1 1 1 1 1 1 1 1 1
2.82801 2.56203 2.31004 2.15205 2.03504 1.84403 1.71300 1.57299 1.45798
2.83801 2.57203 2.32004 2.16205 2.04504 1.85403 1.73200 1.58299 1.46798
0.99007 0.99024 0.99012 0.99003 0.99007 0.99018 0.99007 0.9906 0.99017
0.00991 0.00956 0.00974 0.00976 0.00929 0.00997 0.00977 0.00986 0.00999
79 91 77 69 64
124 193 172 156
Table 5.3. Variables Single Sampling Plans Indexed by AQL and LQL for =1% and =1%
1p
2p
Variables Single Sampling Plan (Known Sigma)
n k )( 1pPa )( 2pPa ASN 0.001 0.0025 0.005 0.0075
0.01 0.02 0.03 0.04 0.05
0.005 0.01 0.02 0.03 0.04 0.05 0.06 0.08 0.10
82 94 80 72 66 130 205 182 165
2.83305 2.56703 2.31505 2.15806 2.03907 1.84907 1.71806 1.57805 1.46305
0.99006 0.99001 0.99016 0.99004 0.99015 0.99018 0.99009 0.99007 0.99025
0.00992 0.00980 0.00970 0.00932 0.00957 0.00996 0.00971 0.00982 0.00984
82 94 80 72 66 130 205 182 165
120
Table 5.4: ASN Values of the Variables SSP, DSP and Variables Chain Sampling Plans
1p
2p
ASN Known Sigma Unknown Sigma
SSP DSP ChSP SSP DSP ChSP 0.001 0.0025 0.005 0.0075
0.01 0.02 0.03 0.04 0.05
0.003 0.0075 0.015 0.025 0.05 0.08 0.09 0.10 0.12
74 62 53 39 19 21 30 39 39
59.4 50.1 43.0 31.3 14.9 16.9 24.0 31.3 32.0
46 39 33 29 13 14 23 27 26
381 267 196 129 54 50 66 82 76
302.4 214.2 157.6 102.1 41.5 39.6 52.5 65.1 60.9
242 171 128 83 39 36 44 53 49
Table 5.5. ARL of Variables Chain Sampling Plan for Different i Values
p
ARL i = 1 i = 2 i = 3 i = 4 i =
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05 0.055 0.06 0.065 0.07 0.075 0.08 0.085 0.09 0.10
88768.34
868.88 81.89 20.07 8.12 4.41 2.88 2.14 1.73 1.48 1.33 1.23 1.16 1.11 1.08 1.06 1.04 1.03 1.02
71697.5 554.66 50.25 12.66 5.42 3.14 2.19 1.73 1.47 1.32 1.22 1.15 1.11 1.08 1.06 1.04 1.03 1.02 1.01
60133.4 410.18 37.21 9.77 4.41 2.70 1.98 1.62 1.41 1.28 1.20 1.15 1.11 1.08 1.06 1.04 1.03 1.02 1.01
51781.5 327.19 30.13 8.25 3.92 2.50 1.89 1.58 1.39 1.28 1.20 1.14 1.11 1.08 1.06 1.04 1.03 1.02 1.01
607.80 38.29 10.64 5.06 3.15 2.29 1.83 1.56 1..39 1.28 1.20 1.14 1.11 1.08 1.06 1.04 1.03 1.02 1.01
121
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Fraction Non-conforming, p
Prob
abili
ty o
f Acc
epta
nce,
Pa(
p)
Figure 5.1: OC Curves of a Variables Chain Sampling Plan for Different iσ Values
iσ =1
iσ =2
iσ =3
iσ =4
iσ =∞
122
CHAPTER 6
OPTIMAL DESIGNING OF
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.
6.2 CONDITIONS OF APPLICATION
The following assumptions should be valid for the application of the
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.
In addition, the usual conditions for the application of variables single sampling
plans with known or unknown standard deviation should also be valid.
123
6.3 OPERATING PROCEDURE OF THE KNOWN SIGMA VARIABLES RESAMPLING SCHEME
Suppose that the quality characteristic of interest has the upper
specification limit U and follows a normal distribution with unknown mean μ and
known standard deviation σ. 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
n, say nXXX ..., 21 and compute
XUv
, where
n
iiX
nX
1
1 .
Step 2: Accept the lot if kv and reject the lot if kv . 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.
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
n, say nXXX ..., 21 and compute
LXv
, where
n
iiX
nX
1
1 .
Step 2: Accept the lot if kv and reject the lot if kv . 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.
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-α) and (p2, β), where p1 is called the
acceptable quality level (AQL), p2 is the limiting quality level (LQL), α is the
producer’s risk and β is the consumer’s risk. A well designed sampling plan must
provide at least (1- α) probability of acceptance of a lot when the process fraction
nonconforming is at AQL and the sampling plan must also provide not more than
β 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-α) and (LQL, β). The proposed variables resampling
scheme is designed based on the two points on the OC curve in the following
manner.
The fraction non-conforming in a lot will be determined as
Up 1 (6.1)
where )(y is given by
y
dzzy2
exp21)(
2
(6.2)
provided that the unit is classified as non-conforming if it exceeds the upper
specification limit U. So, the unknown mean μ can be determined if p is specified.
125
Let us define the standardized quality characteristic corresponding to
the fraction conforming as
)1(1 pz 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
)()(1....)()(1)()(1)()( 12 pPpPpPpPpPpPpPpL am
aaaaaa
ma pP )(11 (6.4)
where kvpPa Pr)( 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
npPnpPnpPnpASN maaa
12 )(1....)(1)(1)(
)()(11 pPpPn am
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,
)(2)()( pPpPpL aa (6.6)
)(2)( pPnpASN a (6.7)
The above measures of the resampling scheme under variables inspection can also
be written as
)(2)()( wwpL (6.8)
)(2)( wnpASN (6.9)
where nkzw p
If (AQL, 1-α) and (LQL, β) are prescribed then we have
1)(2)( 11 ww
and )(2)( 22 ww (6.10)
Here w1 is the value of w at p=p1 and w2 is the value of w at p=p2. That is,
nkzw p )(11 and nkzw p )(
22 (6.11)
where 1pz is the value of pz at AQL and
2pz is the value of pz at LQL.
For given AQL and LQL, the parametric values of the variables
resampling scheme namely nσ, k, and m are determined by satisfying the required
127
producer’s and consumer’s 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 )()(2 222 wwnpASN
Subject to 1)( 1pL
)( 2pL
0 ,1m ,1 kn (6.12)
and Minimize
)(
)(2)(
)(2)()(
2
2
1
121 w
ww
wnpASNpASN
Subject to 1)( 1pL
)( 2pL
0 ,1,1 kmn (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 σ. In this case, the proposed scheme operates as follows.
Step 1: Perform original inspection by using variables single sampling plan as the
reference plan, i.e From each submitted lot, take a random sample of size
128
nS, say SnXXX ..., 21 and compute
SXUv
, where
Sn
ii
S
Xn
X1
1 and
1
)( 2
S
i
nXX
S .
Step 2: Accept the lot if Skv and reject the lot if Skv . 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 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 SkX S is
approximately normally distributed with mean )(SEkS and variance
)(2
SVarkn S (see Duncan (1986)). That is,
nk
nkNSkX SSS 2
,~2
22
Therefore, the probability of accepting a lot at each repetition is given by
pSkUXPkvP SS )(
pUSkXP S
129
21)/(
2S
S
S
kn
kU
21
)( 2S
SSp k
nkz (6.13)
If we let,
21
)( 2S
SSpS k
nkzw then )()Pr( SS wkv . (6.14)
Hence the lot acceptance probability of the resampling scheme under sigma
unknown case is given by
)(2)()( SS wwpL
We obtain SS ww 21 and corresponding to AQL and LQL respectively by
21
)( 211S
SSpS k
nkzw and
21
)( 222S
SSpS k
nkzw (6.15)
In this case, the optimization problem becomes,
Minimize )()(2 222 wwnpASN S
Subject to 1)( 1pL
)( 2pL
0 ,1,1 SS kmn (6.16)
and
Minimize
)(
)(2)(
)(2)()(
2
2
1
121 w
ww
wnpASNpASN S
130
Subject to 1)( 1pL
)( 2pL
0 ,1,1 SS kmn (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 producer’s and consumer’s 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 nS = 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
LXv , where
35
1351
iiXX .
Step 2: Accept the lot if 924.1v and reject the lot if 924.1v . 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 nσ, kσ
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-α) and (p2, β). 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) nσ = 138 and kσ = 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 kσ = 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 producer’s 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
p1
p2
MinASN(p2) Known Sigma Unknown Sigma
n kσ ASN(p2) nS kS ASN(p2) 0.001
0.005
0.01
0.02
0.03
0.04
0.05
0.0020 0.004 0.006 0.008
0.006 0.008 0.01
0.012
0.02 0.025 0.03
0.035
0.03 0.035 0.04 0.05
0.04
0.045 0.05 0.06
0.05 0.06 0.07 0.08
0.06 0.07 0.08 0.09 0.10
129 30 18 13
1445 208 94 57
78 43 29 22
194 99 63 35
342 168 104 54
512 151 76 49
711 201 101 63 44
3.023 2.951 2.902 2.862
2.555 2.523 2.497 2.475
2.240 2.210 2.184 2.163
1.999 1.977 1.957 1.924
1.839 1.822 1.805 1.777
1.717 1.688 1.663 1.639
1.616 1.591 1.568 1.547 1.528
251.552 58.477 35.118 25.335
2815.899 405.631 183.401 111.153
152.108 83.827 56.514 42.905
378.350 193.036 122.805 68.274
666.519 327.547 202.676 105.235
997.757 294.335 148.102 95.514
1385.524 391.712 196.871 122.797 85.753
713 160 90 64
6164 868 381 230
272 148 99 73
581 291 183 99
919 446 273 140
1266 364 182 113
1640 455 224 137 95
3.023 2.951 2.906 2.873
2.555 2.523 2.496 2.474
2.239 2.209 2.183 2.161
1.998 1.977 1.957 1.922
1.839 1.822 1.806 1.777
1.717 1.688 1.663 1.641
1.616 1.591 1.568 1.548 1.529
1389.883 311.820 175.410 124.766
12011.75 1692.537 742.489 448.209
530.048 288.427 192.950 142.264
1132.236 567.281 356.665 192.942
1790.912 869.483 532.261 272.895
2467.042 709.339 354.743 220.263
3195.917 886.678 436.525 267.022 185.177
136
Table 6.2.Variables Resampling Schemes (with m=2) Involving Minimum Sum of ASN Indexed by AQL and LQL
p1
p2
Min[ASN(p1)+ASN(p2)] Known Sigma Unknown Sigma
n
kσ
ASN(p1)+ ASN(p2)
nS
kS
ASN(p1)+ ASN(p2)
0.001
0.005
0.01
0.02
0.03
0.04
0.05
0.0020 0.004 0.006 0.008
0.006 0.008 0.01
0.012
0.02 0.025 0.03
0.035
0.03 0.035 0.04 0.05
0.04
0.045 0.05 0.06
0.05 0.06 0.07 0.08
0.06 0.07 0.08 0.09 0.10
129 30 18 13
1445 208 94 57
78 43 29 22
194 99 63 35
342 168 104 54
512 151 76 49
711 201 101 63 44
3.023 2.951 2.902 2.862
2.555 2.523 2.497 2.475
2.240 2.210 2.184 2.163
1.999 1.977 1.957 1.924
1.839 1.822 1.805 1.777
1.717 1.688 1.663 1.639
1.616 1.591 1.568 1.547 1.528
409.290 95.166 56.940 41.005
4571.700 660.100 298.320 180.887
247.511 136.414 91.948 69.788
615.683 314.102 199.763 111.029
1083.797 533.050 329.550 171.273
1624.070 478.664 241.006 155.159
2253.381 637.426 320.076 199.569 139.392
713 160 90 64
6164 868 381 230
272 148 99 73
581 291 183 99
919 443 273 140
1266 364 182 113
1640 455 224 137 95
3.023 2.951 2.906 2.873
2.555 2.523 2.496 2.474
2.239 2.209 2.183 2.161
1.998 1.977 1.957 1.922
1.839 1.821 1.806 1.777
1.717 1.688 1.663 1.641
1.616 1.591 1.568 1.548 1.529
2262.286 507.560 285.412 202.938
19500.28 2754.619 1207.84 729.240
862.196 469.126 313.654 231.305
1840.637 923.313 580.282 313.605
2912.349 1403.939 866.205 444.022
4015.786 1153.925 577.120 358.503
5197.656 1442.941 709.906 434.460 301.225
137
Table 6.3: ASN Values of the Variables Single, Double Sampling Plans and Resampling Schemes
p1
p2
ASN at AQL Known Sigma
Single Double Resampling Scheme 0.001
0.0025 0.005
0.0075 0.01 0.02 0.03 0.04 0.05
0.004 0.0075 0.015 0.030 0.03 0.05 0.09 0.13 0.11
45 62 53 29 44 52 30 22 49
36.8 50.1 43.0 23.2 35.0 42.3 24.0 18.3 40.0
36.69 50.15 42.81 23.22 35.48 42.82 24.45 18.35 40.35
Table 6.4: Sample Size of the Attributes and Variables Resampling Schemes
p1
p2
Sample size Attributes Resampling
Scheme Variables Resampling
Scheme 0.001 0.001 0.001
0.0025 0.0025 0.0025 0.004 0.004 0.004
0.0065 0.0065 0.0065
0.01 0.02 0.03 0.01 0.02 0.03 0.01 0.02 0.03 0.01 0.02 0.03
470 148 98 624 234 98
1177 311 156
* 249 207
10 6 4
25 11 7
231 65 38 927 117 58
* sample size greater than 5000
138
Table 6.5: ASN of the Variables Resampling Schemes for Different m values
p1
p2
m
Variables Resampling Scheme
Sample Size
ASN at LQL
0.001
0.002
0.01
0.004
0.005
0.02
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
226 160 136 121 112
439 306 255 228 209
388 272 227 201 186
226.0 311.822 394.123 465.480 537.145
439.0
596.375 739.072 877.211 1002.380
388.0
530.048 657.840 773.324 892.258
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-α) and (LQL, β). 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 consumer’s 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
Statistics, Vol.11, No.3, pp.137-147.
Bender, A. J.(1975),”Sampling by Variables to Control the Fraction Defective:
Part II”, Journal of Quality Technology, Vol.7, pp.139-143.
Bravo, P. C. and Wetherill, G. B. (1980),”The Matching of Sampling Plans and the
Design of Double Sampling Plans”, Journal of the Royal Statistical
Society A, Vol.143, pp.49-67.
Bruhn-Suhr, M. and Krumbholz, W. (1990),”A New Variable Sampling Plan for
Normally Distributed Lots with Unknown Standard Deviation and
Double Specification Limits”, Statistical Papers, Vol.31, pp.195-207.
Calvin, T.W.(1977):TNT Zero Acceptance Number Sampling, American Society
for Quality Control Annual Technical Conference Transactions,
Philadelphia, PA, pp.35-39.
Cameron, J. M. (1952),”Tables for Constructing and for Computing the Operating
Characteristics of Single Sampling Plans”, Industrial Quality Control,
Vol.9, No.1, pp.37-39.
Chen, C.H. (2005),”Economic Design of Dodge-Romig AOQL Single Sampling
Plans by Variables with the Quadratic Loss Function”, Tamkang
Journal of Science and Engineering, Vol.8, No.4, pp.313-318.
Chen, J. and Lam, Y. (1999),” Bayesian Variable Sampling Plan for Weibull
Distribution with Type I Censoring”, Mathematicae Applicatae Sinica,
Vol.15, No.3,pp.269-280.
150
Chen, J., Li, K., and Lam, Y., (2007), “Bayesian Single and Double Variable
Sampling Plans for the Weibull Distribution with Censoring,”
European Journal of Operational Research, Vol. 177, No. 2,
pp.1062–1073.
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.169–176.
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.234–238.
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.105–109.
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.28–37.
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.457–516.
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.417–423.
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
Statistics, Vol.19, No.1, pp.141-153.
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.1379–1391.
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 Cpmk 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.