10 Acceptance Sampling

Acceptance sampling is the process of

evaluating a portion of the
producumaterial in a lot for the purpose of accepting or
rejecting the lot as
either
conformingor not
oonforming to a quality specifications.
Inspection for acceptance purpose is carried out at many stages in
manufacturing. There are generally two ways in which inspection is carried out:
100% inspection. (i) Sampling inspection.
In 100%
inspection all the parts or products
are subjected to inspection,
in
whereas sampling inspection only a is
sample drawn from the lot and inspected.
A sample may be defined as the number of items drawn from a lot, batch
or population for inspection purposes.
Sampling inspection can be defined as a technique to determine the
acceptance or rejection of a lot or population on the basis of number of defective
parts found in a random sample drawn from the lot. If the number of defective
items does not exceed a predefined level, the lot is accepted, otherwise it is rejected.
Sampling inspection is not a new concept. In our daily life we use sampling
inspection in selecting certain consumable items. For example, while purchasing
our annual or monthly requirements of wheat, rice or such other food grains we
naturally take a handful of grains to judge its quality for taking purchasing decision.
If we are not satisfied we take another sample and after two or three samples from
the same or different sources we take purchasing decision. Or, let us take another
example, suppose we want to purchase mangoes we nommally take one or two
mangoes from the lot and taste its quality, if the samples taken are found good
we decide to purchase the required quantty.
Similarly in engineering sampling inspection is preferred because it is more
practical, quick and economical as comnpared to 100% inspection. The main purpose
of acceptance sampling is to distinguish between good lots and bad lots, and to
classify the lots according to their acceptability or non-acceptability.
The advantages of sampling inspection are as follows:
1. The items which are subjected to destructive test must be inspected by
sampling inspection only.
2. The cost and time required for sampling inspection is quite less as
compared to 100% inspection.
3. Problem of inspection fatigue which occurs in 100%% inspection is

eliminated.
4. Smaller inspection staff is necessary.
S. Less damage to products because only few items are subjected to handling8

during inspection.
321 -
322
ONTROL
eror introduced by 10m
6. The problem of monotony and
inspector 100%
inspection is minimised.
of sampling inspection is that it evo
8. The mostimportant advantage the rejection of entire
on quality improvement. Since
more effective pressure on qualty improvement
much stironger pressure
on the basis of sampling brings
individual articles.
than the ejection of
decisions
Limitations. (1) Risk of making wrong

since only a part is inspected, it is inevitable

However, in sampling inspection,
the exact picture obtaining in the lot
that the sample may not always represent
or risk of making wrong
decisions about the
and hence, there will be likelihood
in two ways. Firstly, a really good lot (that
lot. This wrong decision can be made
defectives than specified) may be rejected because
is, containing less proportion of
bad lot (that is a lot containing
the sample drawn may be bad. Secondly, really
a

may be accepted because the sample

greater proportion of defectives than specified) suffer a risk of his
has to
drawn may be good. In the former case, the producer
associated risk (chance) is called as the
good lots being rejected and hence the
runs the risk of accepting bad lots
producer's risk. In the later case, the consumer
consumer's risk.
and hence the associated risk is called as
about the product than
(2) The sample usually provides less information
100 per cent inspection.
(3) Some extra planning and documentation is necessáry.
and the
However, in scientific sampling plans, these risks are quantified
of the economic
sampling criteria are adjusted to balance these risks, in the light
factors involved.
The success of a sampling scheme depends upon the following factors
() Randomness of samples.
(i) Sample size.
(ii) Quality characteristic to be tested.
(iv) Acceptance criteria.
(v) Lot size.
Industrial uses of Acceptance Sampling:
1. To detemine the quality and acceptability of incoming raw matct

component parts, products etc.

the acceptability of semifinished products for further processing
10 0CIe
as it undergoes the operations from machine to machine or section to section witnn

he factory.
3. To determine the quality of outgoing products.
O nproving, maintaining and controlling the quality of the products
manufactured.
ACCEPTANCE SAMPLING 323

Lot Formation
A lot is a collection ofitcms from which a sample of two or more articles
is drawn and inspected to determine its acceptability.
The lot formation greatly influcnces the outgoing quality and inspection costs.
The following points should be taken into consideration as guidelines for the
formation of a lot.

1. The products from different sources (processes, production shifts, input

materials, etc.) should not be mixed together, unless there is an evidence that the
lot-10-lot variation is small enough to be ignored.
2. For lot formation the products should not be accumulated over extensive
period of time.

3. The extraneous information such as process capability, poor inspection

ete. should not be used in lot formation.
4. The lot should be as large as possible consistent with the above to take
advantage of low proportionate sampling costs.
Sampling methods
The sampling methods can be classified as
1. Simple Random sampling
2. Stratified sampling
3. Systematic sampling
4. Cluster sampling
5. Two stage sampling
Simple Random Sampling
Selecting a sample in such a way that each item in a lot has an equal chance
of being selected, is called random sampling. Since, a judgement about the lot
is to be made on the basis of only a sample it is very important that the sample
tnuely represents the universe from which it is drawn. This requires that the sample
size be large enough and the sampling procedure such as to avoid bias. Parts resting
on the bottom or in the middle of a group must be selected as well as those lying
convenienly on top.
According to this method, the sample of the requisite size n is drawn from
a lot of size N, in such a manner; that while selecting an item, the chance for
any item of the lot being included in the sample is the same. An item once drawn
IS not placed back in the lot.
Stratified Sampling
In large lots, the difficulies of random selection may be so great that it may
be advisable to adopt stratified (proportional) sampling. For stratified sampling
the following rules should be followed.
1. Draw proportional samples. Acconding to this nule, inspection lots should,
 STATISTICAL QUALTY CONTROL

324 homogenity (into ccertain

ertai

on the basis of
into sub-lots
whenever possible,
be divided
or strata).
homogcneous groups sub-lots of the inspection loe
lot.
of such
number
items from
all parts of
2. Draw sample bias).
(without any
items blind
3. Draw sample lot is divided a sub-samnle
the inspection
Prom cach sub-lot
into which should be
proportional to the size
sub-sample
sclected. The size
of the
shoud be
of the sub-lot. more e i f i c i e n t
than the simple
may be generally
This method of
sampling not always result in the
random sampling may
random sampling. as the simple s t r a t u m of
the lot, thereby affecting
the
items from such
seection of the
drawn.
representativencss of
the sample
a minimum
of two items are selected
advisable to e n s u r e that
It would be
fron cach sub-lo

Systematic Sampling manner, (such piles

as
When the items in a lot are presented in an orderly
considerably simplify
of mild stcel shocts or stacks cement
of bags) it is possible to
of the required size.
the sclection of a random sample random from the lot and
onë item is chosen at
In systematic sampling has been
selected regularly at predeternmined intervals. It
thereafter, the itcms are
establishod that this method of systematic
sampling is quite good approximation
there is no deliberate attempt to manipulate
to the simple random sampling provided
the lot is presented
the scquence of the items in the lot
in any desired manner while
for inspection.
Cluster Sampling
of clusters
When the lot submitted for inspection consists of certain groups
economical to select a few clusters of items
of items, it may be advantageous and
selected clusters.
and then examine all the items in the
cartons and it is either
For example, wbeu the lot consists of items packed in
for selecting sample items,
impracticable or cosuy to repack the cartons opened
in the first
then only few cartons are selected at random without replacement
instance and all the items in the selected cartons are inspected.
Two Stage Sampling
When the lot sub1nitted for inspection consists of larger number of packages
cach consisting of a number of items, it may not be economical to select few
packages and inspect all the items in these packages (as in case of cluster sampling).
In such cases, the sample is selected in two stages. In the first stage a desired
Dunber of packages (primary units) are seleced at random and in the second stage,
the required nunber of items are chosen at random from the selected primary unils.
Sanpling inspecion may be cavied out as () Sampling by atuributes
(i) Sampling by variables.
ACCEPTANCE SAMPLING 325

In sampling by attributes, the items are classified as non-defectives or

defectives according to one or more charactenistics. The decision on acceptance
or sejection is made depending upon the number of defectives found in the sample
(or in some cases in more than onc sample). AS the degree of defectiveness is
not taken into account, sampling by attributes can be applied to both measurable
o non-measurable characteristics. Go and no-go gauges can be used for the
or

inspection of measurable characteristics it is not necessary to register the

measurements, but to detect only the number of defective items found in the sample.
Ini sampling by variables, it is necessary to take measurements on the
cbaracteristics inspected. A statistical treatment will then be used to decide the
acceptance or rejection of the lot.
Symbols Used in Relation to Sampling Acceptance
Following symbols and terms are used in connection with acceptance
sampling
N number of pieces in a given lot (Lot size)
nnumber of pieces in a sample (Sample size)
M number of defective pieces in a given lot of size N.
m= number of defective pieces in a given sample of size n.
acceptance number, the maximum allowable number of defective
pieces in a sample of size n.
p fraction defective. In a given submitted lot p = MN in a given
sample p = m/n.

p' true process average fraction defective of a product submitted for

inspection.
p average fraction defective in observed samples.
Pa probability of acceptance.
Pe consumer's risk, the probability of accepting product of some stated
quality. The probability of rejecting product of some stated quality
is referred to as producers risk, (1 -P). Po9s, Po.so, Po.15 etc.
fraction defective having a probability of acceptance of 0.95,0.50,
0.15 etc. under any given acceptance criteria.
The Operating Characteristic (OC) Curve
The operating characteristic curve for an attribute sampling plan is a graph
of fraction defective in a lot against the probability of acceptance. For any fraction
defective p' in a submitted lot, the OC curve shows the probability Pa that such
a lot will be accepted by the sampling plan. In a single sampling plan three
parameters are specified
N= lot size from which the samples are drawn
n = sample size
C acceptance number.
 Ifthe sampling piau lc=2 lot of 100,
if the sam
if the sample contains
from a ns
random
sample of5 accept By cha
the lot. By changing
t hthe
e
It means
take a
reject the lot
otherwise

can be obtained E rerent

2 defectives,
sampling
plans
more
than different an OC c u r v e
C construct

parameters
N, n and
will differ.
To
varying ner
lots with varying per ccent
the OCcurve of accepting
sampling plans probability Poisson's
distribution
mathematical
the table of
know the obtained
from
This can be
defectives.

in Appendix).

1:0 Producer's
risk

Accept-
able Consumer's
Indifferent
Quality risk
Region Quality region Objection
able Quality RegtoT
LTPD
1-0
AQL
Proportion of defectives (P)-
Fig. 10.1. OC curve.
Producer's Risk and Consumer's Risk
There are always two parties to an acceptance procedure the party submitting
the product for acceptance and the party for whom the decision is made regarding
shows an ideal
acceptance or rejection (i.e. producer and consumer). Fig. 10.2
OC curve for a case where it is desired to accept all lots of 2% defectives. All
lots greater than 2% defectives have a probability of acceptance of zero. Actualy,
however, no sampling plan exists that can discriminate perfectly, there always
remains some risk that a good lot will be rejected or that a bad lot will be accepted.
Producer's risk. ideal sampling plan which will satisty both n
Since,
Consumer and producer is not possible, some compromise has to be made and
they have to tolerate certain risk.
ACCEPTANCE SAMPLING
327

quality is good still from

If the
some lots are rejected the
sampling plan 1.0
suffer. The producer's
producer has to
is the probability of rejecting good
a
isk Acceptance
which otherwisc would have been ine
lot
should be
acccpted. So the producer
the rejection of Pa
protected against
relatively products. The producer
beter

can decrease his risk by producing

prouducts atquality level than
a better

the specified AQL (cxplained

latter) (0,0) 1 23 4

on other economical 100 P

depending
considerations.

Saying a producers risk « = 0.05 Fig. 10.2. Ideal 0C curve.

about 1 lot in 20 will be rejected provided

that the
means that in the long run
level.
lots are coming from a process controlled at AQL quality
Consumer's risk. f the is bad still from the sampling plan some
quality
lots are to be acceptcd the consumer will suffer. Consumer's risk is the probability
of defective lots being accepted which otherwise would have been rejected. Saying
that Po.10 2.5% mcans the consumer does not want a worse quality containing
more than 2.5% defectives and he would at the most accept 10% of lots containing
2.5% defectives.
have
imnpression, it appears that the producer and consumer should
At first
the
completely opposite viewpoints towards the selection of sampling plans (i.e.
intersects are conflicting). But, more critical consideration will show that, there
is a continuing relationship between producer and consumer ; substantial rejection
of good products in the effort to exclude bad products are not necessarily in the
interest of consumer. The consumer is interested in quality, he is also interested
in cost. In the long run the costs incident to the rejection of good products tend
to be passed on by the producer to the consumer. Secondly, any good product
that he rejects is not available for his immediate use.

Quality Indices for Acceptance Sampling Plans

1. Acceptable Qualiry Level (AQL). It represents the maximum proportion
of defectives which the consumer finds definitely acceptable. AQL can also be
defined as the maximum per cent defectives that for the purpose of sampling
inspection can be considered satisfactory as a process average. It is the fraction
defective that can be tolerated without any serious effect upon further processing
or an customer relations.

As an AQL is an acceptable quality level, the probability of acceptance for

an AQL lot should be high. In fact the producer's safe point is termed as AQL.
 STATISTICAL QUALITY CONTROL
328
IN is als0 called as Lot
Tolerance Per
Level (RQL).
2. Rejectable Quality
definition of unsatistactory quality. represen.
It
cent Defective
This is a
(LTPD).
which the consumer iinds definitely unaccentahta
the proportion of defectives
level, the probability of acceptance for an ROr
As RQL is unacceptable
an quality at RQL level represente
be low. The probability of accepting a lot
lot should
consumer's risk.
This is a quality somewhere level
3. Indifference Quality Level (IQL).
defined as the quality level having
between the AQL and RQL. It is frequently
a probability of of 0.50 for a given sampling plan.
acceptance
It the average % defective
Average Outgoing Quality (A0Q). represents
4.
all accepted and all rejected
in the outgoing products after inspection, including
defectives replaced by non-defectives.
lots which have been 100% inspected and
as a result of first sampling
So, for a given fraction defectives, the lot accepted
lots are subjected to 100%
inspection will have a fraction defective p', the rejected
inspection and rectification (defective articles are either replaced or corrected)
the AO0Q will therefore be less than p'.
Let, n= sample size,
N= lot size,
k number of lots submitted for acceptance,
p fraction defective,
Pa probability of acceptance
(1- Pa) = probability of rejection.
Then, the proportion of the lots accepted = Pak
Proportion of the lots rejected = k(1 - Pa which are subjected to 100%
inspection and rectification.
Number defects originally present in k los = k.N.P'
Number of defectives in the outgoing lots

p'PakN -n)
Total number of defectives
=
p' Pa.k{N -

n) + k(n)0)
Pap' (N- n) + k{n)(0)
AOQ =
kN
= PaP

If the sample
size is much less as
compared to lot size
A0Q = PaPp
The calculation of average outgoing quality gives the expected quality m
the long run. Over a short
period, the outgoing quality may be better or woTSc
than the long-run average.
ACCEPTANCE SAMPLING
329

Average Outgoing Quality Limit (AOQL). For any given n and c there
maximum value ofA0Q beyond which the average fraction of defective passed
ward will not rise, no matter how bad the quality of the lots when they arive
forward

inspection Because, when incoming quality is perfect, outgoing quality must

at
ikewise be perfect. However, when incoming quality is very bad, outgoing quality
will also be perfect, because the sampling plan will cause all lots to beextreme-
rejected,
are subjected at
100% inspection and rectification. Thus at either
which
ipcoming quality very very bad-the outgoing quality will tend to be
g00d or

Between these extremes is the point at which the percent of defectives

very good.
wll reach its maximum. This point is known as the average
in the outgoing product
the maximum possible value of the
outgoing limit (AOQL). Therefore, AOQL is and
defectives in the outgoing products after inspection
average percentage
rectification.

but for rectification

For acceptacne/rejection scheme the OC curve is used
is used.
scheme the curve of AOQ plotted against p'
of the incoming
Any acceptance/rectification plan guarantees that regardless
will not be worse than the
submitted, the outgoing quality in the long run
quality
ACQL.
calculation to determine
Table 10.1, Shows a common type of approximate
the AOQL. This table refers to the plan, n 75 and C= 1 when Nis large in
=

quality (AOQ)
comparison with n. The right hand column gives average outgoing value of
for each assumed per defective in submitted lots. The maximum
cent
2.2% defective. This maximum
A0Q is 1.129% occuring when submitted lots are
value of A0Q is the AOQL.
Table 10.1. A0Q from n =75, c = 1 for Acceptance/Rectification plan.

Per cent defectives in

Probability of acceptance AOQ 100p'Pa
submitted lots
0.963 0.385
0.4
0.827 0.827
1.0
0.718 005
1.4
1.6 0.663 1.061
0.610 1.098
1.8
0.558 1.116
2.0
0.509 1.120
2.2
2.5 0.441 1.102
0.343 1.029
3.0
3.5 0.262 0.917

4.0 0.199 0.796

 330
STATISTICAL QUALITY CONTRO
The above computed figures are plotted in Fig. 10.3. The curve
known as AOQ curve. plotted is

2-0 Curve without1

rectification

5F
5 Max AOQ
o 0
A0QL:1.12

0-5+

1-0
Percent
15
2-0 25 3-0 35 4
Defective (100 p')
Fig. 10.3. Average Outgoing Quality for
The line A0Q
Acceptance Rectification plan.
p' represents what would
=

rectification. For any given values of 'n' and 'c', the

happen if there were no
line A0Q A0Q curve falls below the
=p. Asp' increases, the proportion of rectified
the A0Q curve falls lots increases and bence
below the line AOQ p'. =

Steps in the Design of an

Aceptance Plan
The protection provided to the consumer and the
described by the OC curve of the producer is completely
plan. When
the acceptance
it calls for plan is designed,
designing a plan whose OC curve will
points agreed upon by the consumer and pass through two stipulated
acceptable quality level (AQL) producer. The first of these is the
which the consumer finds representing maximum
the
of defectives proportion
Percent Defective (LTPD) definitely acceptable. The second, is the Lot Tolerance
which represents the
the consumer finds
definitely unacceptable of defectives proportion which
described earlier.
as
Accordingly,
in Fig. 10.1.
an OC curve can be
divided into three egions,
(zones) as shown
1.
Acceptable Quality Region.
2. Indifferent Quality
3. Region.
Objectionable Quality Region.
The consumer is
willing to
accept all lots which have a
represented by AQL fraction
reject all lots which have a fraction defective
or less and
represented by LTPD or more. So defecuves
AQL and LTPD are far, as the lots with a fraction
said to be concerned, in quality control detective between
"indifferent"
available by
to them. But
to take
terminology, the consumer
advantage of the
acceptance sampling, the consumer economies mauc
must be willing toassume soue
ACCEPTANCE SAMPLING
331
riskof accepting some lots which he finds definitely unacceptable, that is, whose
tion of defec
oportion of defectives equals or excecds the LTPD. Also, the produccr must
e villing to assume some risk ot having some of definitcly acceptable lots rejected
be
that is some
that is
some of
of the ots whose proportion of defectives equals or is less than the
the lots
AQL. Usually the consumer's and producer's risks are agrecd upon and cxplicitly
stated in quantutauve erms. For example, the producer may agree to accept the
risk that 10 per cent of the lots with a fraction defective of say 0.01 (AQL) will
he rejected, and the consumer may agree to accept the risk that 5 per cent of the
lats with a fraction defective of 0.25 (LTPD) will be accepted.
Let us discuss the hypothetical case we have presented in light of Fig. 104.
In our case, the consumer defined the AQL as being 0.0I and the LTPD as being
0.25. The producer is willing to accept a 10 per cent risk that lots with a proportion
of defectives of 0.01 would be rejected end the consumer is willing to accept a
These
5% risk that lots with a proportion of defectives of 0.25 would be accepted.
conditions and the symbols commonly used to represent them are
P1 AQL 0.01 =

P2 LTPD = 0.25
a = producer's risk = 0.10

B consumer's risk =0.05.

to be 0.10, this means that the
When we say that the producer's risk is
cent defectives should be 1
of accepting a lot which contains 1 per
probability there should be
0.10 0.90. Also, a consumer's risk of 0.05 means that
minus or
cent defectives.
of 0.05 of accepting a lot which contains 25 per
probability

1.00- a : 0-10
0-90F

. _ B 0 . 0 5

0-05
0-025
0-01
(P2)
(P1) deiectives (p)
Proportion of

0.10, B = 0.05).
10.4. OC curve (a =

Flg.
 332 STATISTICAL QUALITY CONTROL

The OC curve for this set of conditions would have the general form
in Fig. 10.4.
orm sho.
shown
In effect, what the consumer and producer agree upon is a set of
two non
for the OC curve; these points are (P1, 1 a) and B). In other
-

P2,
agree that any acceptance plan whose OC curve passes through these pre-selerH
words, the
points will be satisfactory. A trial and error approach must be employed
in thia
design. To illustrate this, the question may be; given a lot size of 100, what shoula
be the sample size and acceptance number ? The only
way in which this canbe
determined is by arbitrary selecting a sample size and an
acceptance number
calculating the respective probabilities of acceptance for the AQL and LTPD,
hoping that they coincide with (l a) and B. Chances are and
-

that they will no

Another combination of sample size and
acceptance number must then be selected
and the procedure repeatd until
they coincide.
1.0

0-9

0-8
N 50
n 5
0-7H c0

0.6

0-5 N-100
n10
le:0
04

N=200
0-3 n:20
c 0

0-2
N 1000
n 100
0-1 lc 0

6 8 10 14
Percent defective (100 p')
Fig. 10.5. Comparison of OC curve.
ACCEPTANCE S A M P L I N G

333

The problem
of design is further
complicated by the fact that even if p,
a and B remain fixed, a different plan would be required for each lot size
becaus the lot size enters into the probability computations.
P2

OC Curve
Characteristics of
1. The Operating Characteristic (OC) Curve of an Acceptance Sarnpling
chows
plan the ability of he plan to distinguish between good lots and bad lots.
shows th
2. Sampling acceptance plans with same per cent samples gives very
fierent quality protection. For example, the curves (Fig. 10.5) shows that lots

1 0

0-9

0-8

0.7

0-6

N 1000
n 20
0-5
c:0

0-4 N 200
20
c: 0
0.3
N 100
n: 20
c: 0
0-2
N 50
n: 20

01
c 0

0 14
0 6 8
2 defective (100 p')
Percent

for four sampling

plan.
curve
of OC
Fig. 10.6. Comparion
 KOL

which are 4% defective will be accepted 81% of the ume using a 10% samnl
from a lot of 100, 42% of the time using a 10% sample from a lot of 200, and ample
less than 2% of the time by a 10% sample from a lotor lO00, assuming an
acceptance number of zero in all cases.

Obviously a producer making a product 46 defective would have a strong

motive for trying to have his product inspected in lots of 50 rather than in lots
of 100.
3. Fixed sample size tends towards constant quality protection. It is the
absolute size of the sample rather than its relative size that detennines the quality
protection given by an acceptance sampling plan.
Fig. 10.6, illustrates this point.
4. The OC curves of plans with acceptance numbers greater than zero are
superior to those of comparable plans with acceptance. number of zero.
With fixed value of N, and coen, larger the value of n, the better is the ability
of the plan to discriminate between good and bad lots. Fig. 10.7, illustrates this
point.

1 0f

0.8
n:75
C1

0-6

n 150
c:2
0 4
n 750
C 10

0-2

0.5 1.0 1.5 2-0 2.5 3.0 3-5

Percent Detective (100 p)

Fig. 10.7.
5. The larger the sample size and acceptance number, the
of the OC curve. Fig. 10.7 shows that the
steeper the Siope
larger sample size which the protects
consumer against the
acceptance of rclaively bad lots also gives better protection
to the producer against rejection of relatively good lots.
ACCEPTANCE SAMPLING
Sampling Plans 335
Sampling plans may be grouped into
1. Single sampling plan. three categories.
2. Double sampling plan.
3. Multiple sampling plan.
Single Sampling Plan
When a decision on
of only one sample, the acceptance or rejection of the lot is made on
In a single
acceptance plan is known as a single the basis
sampling plan three numbers are sampling plan.
N= lot size, from specified
which the sample is drawn.
n sample size.
C acceptance number.
If the
sampling plan is
N 50
n= 5
C 1
It means, take a
sample of 5 items from a lot of 50, if the
sample contains
more than 1 defective reject
the lot, otherwise accept the lot.
Inspect a sample of n pieces

If number of defectives

Does not exceed 'c' Exceeds c

Accept the lot Reject the lot

Single sampling Plan
Double Sampling Plan
In double sampling plan the decision on acceptance or rejection of the lot
is based on two samples. A lot may be accepted at once if the first
sample is good
enough or rejected at once if the first sample is bad enough. If the first sample
is neither good enough nor bad enough, the decision is based on the evidence
of first and second sample combined.
Parameters. n = number of pieces in the first sample.
c1 = acceptance number for the first sample,
i.e. the maximum number of defectives that will permit the acceptance of the lot
on the basis of the first sample.
n2 number of pieces in the second sample.
n 2= number of pieces in the two samples combined.
 336 STATISTICAL QUALTY CONTROL
c2 acceptance number for the two samples
combined.
i.e. the maximum number of defectives that will permit the acceptance of the lot
on the basis of the first and second sample combined.

Parameters Inspect n pieces

N
I f the no. of defectives
C1
n2
C2

Does not exceed ci | C def. c2 Exceeds cC2

Accept the lot Take second sample Reject the lot

of n2 pieces

No. of defectives in the first

and second samples combined,
ie. in (n1+n2)

Does not exceed C2 Exceeds c2

Accept the lot Reject the lot

Double Sampling Plan

The following example illustrates the use of a double sampling plan. Let,
N 500, nj = 35
C1 =1, n2 =50
and C24
This may be interpreted as follows
1. Take a first sample of 35 items from a lot of 500 and inspect.
2. Accept the lot on the basis of first sample, if it contains 0 or 1 defecüve.
3. Rcject the lot on the basis of first sample if it contains more than 4 defectives.
4. Take a second sample of 50 items if the first sample contains 2, 3 or
defectives.
ACCEPTANCE SAMPLING
337
Accept the lot on the basis of first and second
combined sample of 85 items contains 4 or sample combined, if the
less defectives.
A Reject the lot on the basis of combined sample if the
contains more than 4 defectives. combined sample
The lot thus may be accepted in the
following ways:
(a) 0 or 1 defective in the first sample (without
taking second sample).
(b) defectives in the first sample followed by 0, 1 or 2 defectives in the second
2
sample.
(C) 3 defectives in the first sample followed by 0 or 1 defective in the second
sample.
(d) 4 defectives in the first sample followed by 0 defective in the second
sample.
The probability of accepting the lot is the sum of the probabilities of these
different ways in wbich it may be accepted.
Multiple Sampling Plan
The phrase multüple sampling is generally used when three or more
samples
of stated size are permitted and when the decision on acceptance or rejection must
be reached after a stated number of samples.
The phrase sequential sampling is generally used when a decision is possibBe
after each item has been inspected and when there is no specified limit on the
total number of units to be inspected. However, some writers use the two
phrases
interchangeably.
A multiple sampling procedure can be represented on a table such as the
following:
Combined Samples

Sample
Sample Size Acceptance Rejection
size umber number

First n C1
Second n2 ni + n2 C2 2
Third n3 ni +n2 +n3 c3 3
Fourth n4 ni+n2+n3 + n4 C4 T4
Fifth ns n+n2+ n3 +n4+ns C5 C5+1

Multiple Sampling Plan

A first sample of n is drawn, the lot is accepted if here are no more than
CI defectives, the lot is rejected if there are more than ri defectives. Otherwise
second sample of n2 is drawn the lot is accepted if there are no more than c2
defecivesin the combined sample of ni +n2. The lot is rejected if there are more
than r2 defectives in the combined sample of n1 + n2. The procedure is conunucd
n accordance with the above table. If by the end of the fourth sample, the lot
1S neither accepted nor rejected, a sample ns is drawn. The lot is accepted if
338
STATISTICAL QUALITY CONTROL
the number of defectives in the combined sample ot n1 +
n2 +
n3 +n4 + ns does
not exceed C5. Otherwise the lot
is rejected.
Note that Ci < C2<... <C5 and c;< r
for all i.
Amultiple sampling plan will generally involve less total inspection, han
the comesponding single or double sampling
plan guaranteeing the same protection,
But they usually require bigher administrative costs, and
bigher calibre inspection
personnel may be necessary to guarantee proper use of the plans.
Design of Item by Item Sequential Sampling Plans
A certain product is subjected to lot by lot
acceptance or rejection on the
basis of a destructive test applied to a sample. The conditions of
the test are
considerably more severe than it is expected in praçtuce. All items tested are
damaged to such an extent that they are of no further use. In order to keep the
number of items tested to a minimum consistent with the desired
quality protection,
it is necessary to design an item by item sequential
sampling plan.

RE JECT

Sn+h
2
d2
l i n e

Rejection
CONTINUE
INSPECTION Sn*h1

di:

Acceptance
l i n e

ACCEPT

5 10 15 20 25
No. of items inspected,n

Fig. 10.8. Graphical representation of item by item an sequential plan.

Such a plan may be designed, so that OC curve passes through any two
points desired.
Fig. 10.8 gives a graphical representation of an item by item sequential plan.
The plan is fully defined by the equation of the rejection line, d2 = Sn + hz i
the acceptance line, d = S - h1. To compute S, the slope of these lines, and hi
and h2, the intercepts, certain auxiliary symbols, 81. 82 a and b, are used.
The symbol a represents (1 - Pa) for a lot quality P1.

B represents Pa for a lot quality P2

 computations are as follows:
The necessary
82
S
81 = log
p 81+82
1-P1 h
b - and
g2 1o
1- P2 81+82
a
a log 21+ 82
1- a
b log
reduce the average
of adopting item-by-item plan is to
such an
The purpose sampling
the amõunt that would obtain with a multiple
below
mount of inspection same quality protection.
about the
alan giving Acceptance Plan
a Good
haracteristics of characteristics
should have these
An acceptance sampling plan should reflect the
to define 'quality'
The index (AQL, AOQL, etc.) used primaily for statistical
1. and not be chosen
and producer
needs of the consumer

terms (the OC
curve).
convenience. quantitative
known in
sampling risks should be the rejection of good
lots;
2. The protection against
should have adequate bad lots.
acceptance of
The producer
be protected
against the
should inspection.
the consumer
total cost of
should
minimize the
reflect changes in lot size, quality
3. The plan enough
flexible
to
should be factors.
4. The plan other pertinent information
useful
submitted, and any should provide
of product required by plan
m e a s u r e m e n t s

quality.
5. The and long rnun Plans
individual lot
quality
and Multiple Sampling
in estimating Double
between Single, Multiple
Comparison Double sampling
Single samp- sampling plan
plan
ling plan Lowest
and
between single
In
Largest
number of multiple plans
1. Average and Largest
lot. between single
inspected per In
pieces Lowest
administration multiple plans Least
2. Cost of singleand
between
LargestIn
3. Information available multiple plans

regarding prevailing Indecisionis

acceptable
quality level.
| Most continued
Less (gives
4. Acceptability to only one
for a long8

producers. chance of term

passing the

lot)
 STATISTICAL QUALITY cONTROL
340
four aspects double
sampling nian
Compared with regards to all the choose muluple sampling pDlan.
a

Where the cost of inspection is high,

we umpling plans.) best.
If the
plan the lot has not to wait, hut .
wait, thernultiple
decision is taken on single samnpling
continued
and lot has
fora long period to wait tiple
has to
sampling plan indecision is decision ore
il the decision is
store the items until
space is necessary to
more storage
Attribute Plans Having
a Stated Val.
Value of
Designing Single Sampling
Po.soThe quality Po.50, for any lot-by-lot acceptace sampling plan is the lot o
or
of acceplance or 0.0. Po.so is called as
process quality that has probability
a the
indifference quality.
The single acceptance criteria is designed for any desired Poso using the
following approximate formula
C+0.67
n Po.50
Suppose that lots 2.5% defective or better are considered to be acceptable
but that it is desired to reject lots that are any worse. With this quality standard
we can find a set of single sampling plans for which a lot 2.5% defective wil
have a Pa of 0.50. Assuming value of acceptance number c from 0 up to any
desired number, and solving for sample size n, using Po50 as 0.025. The resulting
family of acceptance plan will be as follows

27 187 8 347

67 227 9 387
2 107 267 10 427
3 147 307 11 467

The OC curves of these 12 single sampling plans will pass through the poInt
Pa =0.50, p = 0025, but the plans having the larger sample size gives beter
discrinination between lots that are somewhat better than 2.5% defective and 1ots

that are somewhat worse.

A choice among the various plans requires balancing the extra inspectiou
Costs of larger samples against the advantage of better assurance that a lot meemg
the quality standard will be accepted and one failing to meet it will be rejecteu
The Average Total Inspection Curve (ATI Curve) [Fig. 10.9)
I00%
The average tolal amount ofinspection called under a programme ot
iued.
inspection of rejected lots will depend on the quality of the material suohe
the
If the material contains no defective
items, there will be no rejections, u will
amount of inspection per lot will be n. If the items are all defective every loty
lot
be subjeced to 100% inspection, in which case the amount of inspecnou
 the size otthe lot. If material
willbe N, is 341
the average
amount
of between zero
ero
inspection per lot will bedefective and 100%
defective,

1Ifthe product is of 1009%

quality p and the between n and N.
inspectionprobability
then, on the average, the amount of of lot
(N-n). per lot will be, ATIacceptance is Pa
=
n + (1
where. ATI stands for
-

P
Average Total Inspection.
example, if N 10,000, n 100,
Eor =
=
nteial Submitted for 2 and if the c =
materia
inspection
and (1- Pa) = 0.88. Hence, (p') 0.05 then, for all
= fraction defective of
the 0C
curves, P, =
0.12
ATI
Fig. 10.9 shows the
=

100+(0.88)(10,000
100) = 8,812
Average Total Inspection curves
n 100, c = 2, lots of 1,000, for the
5,000 and 10,000. sampling plan
10,000
N:10,000
9,000

8,000

7,000

S 6,000

5,000 N=5,000

4,000 F

3,000

2,000

N=1,000
1,000

0 0-01 0-02 0-03 0-04 005 0-06 0-07 008

Product Quality -

' g 10.9. Average total inspection curve for the sampling plans n = 100, c = 2.
Lots of 1,000, 5,000 and 10,000.
 342 STATISTICAL QUALITY CONTROL
The
Average Total Inspection (ATI) for Double Samnl.
Plans. Average total inspection under a double sampling plans in which reie
ected
Sampling
lots are inspected 100 per cent is given by the formula
ATI = n P a + (n1 + n2)Pan + N1 - Pa)

in which Pas is the probability of acceptance on he irst sample, Pa» is

the
probability of acceptance on the second sample. Pa is the probability of final
acceptance and (1 P) is the probability of final rejecions Since Pa = Pa + P
the above formula can be written as, ATI n1 + na(1 Pai) +
(N -
=
n1-n
-

(1 - Pa).

The reasoning underlying the above formula is as follows

1. n items only will be inspected if the lot is accepted on the first sample
and the chance of this is
Pa
2. (n + n) items will be inspected if the lot is accepted on the second
sample, and the chance of this is Pn
3. N items will be inspected if the lot is rejected, and the chance of this
is (1 -

P).
To illustrate the use of the above formula for
computing average total
inspection, consider the computation of average total inspection for the plan
n = 50, n2 = 100, C1 =2 and c2 = 6 and lot size N = 1600. Take p' = 0.06. Then
from Molina's tables we find that the probability of
acceptance on the first sample
(i.e. probability of 2 or less defective items out of 50) is 0.423, and the
of rejection on the first sample (i.e.
probability
probability of more than 6 defective items
out of 50) is 0.0034.. The probability of acceptance on the combined samples is
the sum of
1. The probability of3 defective units in the first
of 3 or less defective units in the second. For a lot
sample times the probability
fraction defective of 0.06, this
is equal to 0.22 x 0.15 =0.033.
2. The probability of 4 defective units in the first sample times the
of 2 or less defective units in the second probability
sample 0.17 x 0.06
=
0.010. =

3. The probability of 5 defective units in the

first sample times the probabilhy
of 1 or less defective units in the second
sample
0.010x 0.02 =0.002.
4. The
probability of6 defective units in the first sample times the probabiy
of zero defecive units in the
second sample
0.05 x 0.002 = 0.0001.
=

The total of these four

probabilities 0.045. =

nerefore, the total probability of acceptance for a lot fraction defecuve

0.06 0.423 0.045 =
0.468 =
Pa
Hence, for p' = 0.06
ATI = 50
+ 100 (1 -

0423) + (1600 150 1 0.40

-

= 50+ 58 + 772 880,