Module 20:

Basic principles of Acceptance sampling

(Most of the material in these notes is based on the material given in Introduction to Statistical
Quality Control by Douglas C. Montogomery)
Acceptance sampling is a statistical method that enables us to base the acceptreject decision on the inspection of a sample of items from the lot. The items of interest
can be incoming shipments of raw materials or purchased parts as well as finished goods
from final assembly. Typical application of acceptance sampling is for lot disposition,
sometimes referred to as lot sentencing, for receiving inspection activities. Accepted lots
are put into production. Rejected lots may be returned to supplier or subjected to other
lot-disposition action. Sampling methods may also be used during various stages of
production.
Acceptance sampling is a form of quality testing used for incoming materials or
finished goods. e.g., purchased material & components. The procedure of acceptance
sampling consists of the following steps.

Take one or more samples at random from a lot (shipment) of items

Inspect each of the items in the sample

Decide whether to reject the whole lot based on the inspection results

Acceptance plan: It consists of procedures for inspecting incoming materials or finished

goods, identifies

Type of sample

Sample size (n)

Criteria (c) used to reject or accept a lot

An acceptance plan is negotiated and decided by Producer (supplier) & consumer (buyer)
must negotiate.
Example 1:
An inspector takes a sample of 20 items from a lot. His policy is to accept a ot if no more
than 2 defectives are found in the sample. Assuming that a lot is 5 % defective, what is
the probability that he will accept the lot or reject the lot?
In this example, we have

n=20, c =2 and p=0.05

P(acceptance of lot)=f(0) + f(1) + f(2)
f ( 0) =

20!
0.05 0 (1 0.05) ( 200) = 0.3585
0!(20 0)!

f (1) =

20!
0.051 (1 0.05) ( 201) = 0.3774
1!(20 1)!

f ( 2) =

20!
0.05 2 (1 0.05) ( 20 2) = 0.1887
2!(20 2)!

P( acceptance of lot)=0.3585 + 0.3774 = 0.1887 = 0.9246

Acceptance sampling has following advantages over 100% inspection including:

Usually less expensive

Less product damage due to less handling

Fewer inspectors required

Provides only approach possible if destructive testing must be used

The lot is disposed off in shorter time so that scheduling and delivery are improved

The problem of monotony and inspector error induced by 100 percent inspection is
minimized

Rejection (rather than sorting) of nonconforming lots tends to dramatize the quality
deficiencies and to urge the organization to look for preventive measures

Proper design of the sampling plan commonly requires study of the actual level of quality
required by the user. The resulting knowledge is a useful input to the overall quality
planning.
The acceptance sampling procedure can be represented as follows.

The three important aspects of acceptance sampling are:

It is the purpose of acceptance sampling to sentence the lots, not to estimate the lot
quality
Acceptance sampling plans do not provide any direct form of quality control. They only
accept or reject lots. Process controls are used to control and improve quality
Acceptance sampling is an audit tool to ensure that the output of a process conforms to
requirements
Type I and Type II errors:
If the null hypothesis H0 is that the lot quality is good, then the following situations and
the corresponding types of error can be considered.
Type I & Type II
Errors

State of the Lot

H0 True

Ho False

Good Quality Lot

Poor Quality Lot

Accept H0
Accept the lot

Correct Decision

Type II Error
Accepting a Poor
Quality Lot

Reject H0
Reject the lot

Type I Error
Rejecting a Good
Quality Lot

Correct Decision

Decision

Economics of inspection: Let

N=number of items in lot, n=number of items in sample, p=proportion defective in lot
A=damage cost incurred if a defective slips through inspection, I=inspection cost per item
Pa=probability that lot will be accepted by sampling plan
Assuming
1) No inspection errors
2) Cost to replace a defective found in inspection is borne by the producer or is small
compared to the damage or inconvenience caused by the defective
Alternative

Total cost

No inspection

NpA

Sampling

nI + (N-n)pAPa + (N-n)(1-Pa)I

100% inspection

NI

Economic comparison of inspection alternatives:

100% inspection is justified in case of equipment where the cost of finding and correcting
defective increases rapidly for each major stage that the product moves to from production to the
customer.
An electronic component may cost $1 at incoming inspection, $10 at the PCB stage, $100 at the
system level and $1000 in the field. In such a case it is worthwhile to spend more on inspection
at initial stage rather than spending much more at a later stage.
It also needs to be noted that Neither sampling nor 100% inspection can guarantee that every
defective item in a lot will be found.
Sampling involves a risk that the sample will not adequately reflect the conditions of the lot. The
Type I error discussed earlier is also called Producers risk or also as risk. The Type II error is
also known as Consumers risk or risk.

Types of Sampling Plans:

Single sampling plan

Sample size n (randomly selected)
Acceptable number c

1.
2.
3.

Double-sampling plans
One of the following decisions, based on the information in the initial sample, is
made
Accept the lot
Reject the lot
Take a second sample
If a second sample is taken, information in both the samples is used to
decide acceptance or rejection

Multiple sampling-plans

Extension of Double-sampling plan, more than two samples may be required

Ultimate extension of sampling is Sequential sampling

Module 21

Single sampling plans for attributes

A lot of size N has been submitted for inspection
Sample size n
Acceptance number of defective c
Lot sentencing is based on one sample of size n
Operating characteristic (OC) curve
Curve plots the probability of accepting the lot (Pa) versus the lot fraction defective (p)

Example 2:
N=10000 , n = 89 and c=2
The probability of acceptance for different values of p, fraction defective are shown in the
following table.

The OC curve is shown in the figure below.

The OC curve shows discriminatory power of the sampling plan.

In the plan n=89, c=2, if the lots are 2% defective, Pa is approximately 0.74. This means in 100
lots 74 are expected to be accepted and 26 to be rejected
The ideal OC curve which has perfect discriminatory power is as shown below.

Comparison of two different OC curves with regard to their discriminatory power is done in the
following figure.

The

and

risks referred to earlier are shown on the OC curve as below.

The following figure shows the effect of sample size on the OC curves. It is noted that the
discriminatory power of the sampling plan increases with sample size.

The following important observations can be made from the above.

The ideal OC curve cannot be obtained in practice
The ideal OC curve can be approached by increasing the sample size.
This is seen from the previous slide in which acceptance number c is kept proportional to
n.
The greater the slope of the OC curve, the better is its discriminatory power
Effect of c on the OC curve: This is shown in the following figure

Changing c does not dramatically change the slope of the OC curve

As c is decreased the OC curve shifts to the left.
Plans with lower values of c provide discrimination at lower levels of fraction defective
Quality Indices for Acceptance Sampling Plans:
Acceptable Quality Level (AQL) represents the poorest level of quality for the vendors
process that the consumer would consider to be acceptable as a process average.
Lot Tolerance Percent Defective (LTPD) is the poorest level of quality that the consumer
is willing to accept in an individual lot. LTPD is also called Rejectable quality Level
(RQL) or Limiting Quality Level (LQL)
AQL and LTPD are not characteristics of sampling plan. AQL is a property of the
vendors manufacturing process. LTPD is a level of the lot quality specified by the
consumer.

Indifference Quality Level (IQL) is a quality level somewhere between the AQL and
LQL. It is frequently defined as the quality level having a probability of acceptance of 0.5
for a given sampling plan
Average Outgoing Quality (AOQ) is the average quality of rejected and accepted lots
Type-A and Type-B OC Curves:
Type-B OC curve is constructed with the assumption that the sample is taken from a
large lot or from a stream of lots selected at random from a process.
Binomial distribution is the exact probability distribution for calculating the probability
of acceptance.
Type-A OC curve is used to calculate probability of acceptance for an isolated lot of
finite size.
The exact sampling distribution of the number of defective items in the sample is the
hypergeometric distribution.
As the size of the lot increases, the lot size has a decreasing impact on the OC curve.
If the lot size is at least 10 times the sample size, the two types of curves are almost same.
Type-A OC curve always lies below the Type-B OC curve.

OC curve for single-sampling with c=0: The curves are shown in the figure below

Sampling plans with c=0 have OC curves that are complex through out their range.
Probability of acceptance drops very rapidly even for small values of lot fraction
defective.
This is hard on the vendor and some times uneconomical for the consumer.
From slide 14, for n=89, AQL=1%, Pa=0.78 for c=1 and Pa=0.41 for c=0
Designing a Single-Sampling Plan with a Specified OC Curve

Suppose we wish to construct a sampling plan such that

Pa=1- for lots with fraction defective p1
and Pa= for lots with fraction defective p2
Then ,n and c are solution to the equations shown in below

The equations are non-linear and not easy to solve.

The nomograph shown below can be used for this purpose.

The illustration is for the case p1=0.01, =0.05,p2=0.06 and =0.1

OC curve can also be drawn using Poisson Approximation.

The Poisson distribution yields a good approximation when n is at least 16, N at least 10
times n and p is less than 0.1.
The Poisson distribution as applied to acceptance sampling is
np ( np ) r
P (r ) = e
r!
where, P(r) is the probability of exactly r defectives in sample of n.

Module 22

Rectifying inspection
Corrective action required after rejection of lots.
Generally, rejected lots are 100% inspected, with defective items removed for rework or
replaced by good items.
Such sampling programs are called Rectifying Inspection Programs

Average Outgoing Quality (AOQ) is the quality in the lot that results from the application
of rectifying inspection.
If the lots are from a process with p fraction defective, then In lots of size N, ,
1. n items contain no defectives as defectives are
replaced
2. N-n items contain no defectives if the lot is rejected
3. N-n items contain p(N-n) defectives if the lot is accepted
Expected number of defectives is Pa(N-n)
Therefore, AOQ = Pap(N-n)/N
If N is large relative to n, Approximately, AOQ= Pap
The curve that plots AOQ against incoming quality, p is called AOQ curve
From the AOQ curve, we note that average outgoing quality is good both for good and
very bad incoming qualities.

The maximum ordinate on the curve represents the worst possible average quality that
results from the rectifying plan.
This point is called the Average Outgoing Quality Limit (AOQL)
AOQL is an average level of quality, across a large stream of lots and it does not give
assurance that an isolated lot will have quality no worse than AOQL.
If the lot has no defectives, the amount of inspection per lot =n
If all are defectives, the amount of inspection per lot = N
If the lot quality is 0<p<1, then the Average Total Inspection (ATI) per lot will be
ATI = n + (1 Pa)(N n)

Double Sampling Plan

Decision of acceptance or rejection based on two samples
A lot may be accepted if the sample is good enough or may be rejected if the sample is
bad enough
If the first sample is neither good enough nor bad enough, decision is based on evidence
of first and second sample combined.
This is shown in the following .

OC Curve of a double sampling plan:

Example 3:
Draw the OC curve of the single-sampling given with sample size 300 and acceptance number 5.
We first fix tentative probabilities of acceptance (Pa)
Consider the table shown below
n

np

Pa

300

0.98

300

0.95

300

0.70

300

0.50

300

0.20

300

0.05

300

0.02

Pap

Search the cumulative Poisson distribution table under c=5 column for a value close to the
required Pa value and note the corresponding np value . For this np calculate the corresponding
p. Complete the table and draw the required curve.

np

Pa

Pap
=AOQ

300

2.0

0.0067

0.983

0.0065

300

2.6

0.0087

0.951

0.00827

300

4.4

0.0147

0.70

0.0106

300

5.6

0.0187

0.512

0.00957

300

7.8

0.025

0.210

0.00526

300

10.5

0.035

0.05

0.00175

300

12.0

0.04

0.02

0.0008

From the above table the OC curve can be constructed which is given in the following

Example 4:
In a double sampling plan, N=5000, n1=100, c1=0, n2=100, c2=1
(a) Use Poissons table to compute the probability of acceptance of a 1% defective lot.
(b) assume that a lot rejected by this sampling plan will be 100% inspected. What will be the
AOQ if the submitted product is 1% defective? Considering the inspection of both
rejected lots, what will be the avarage number of articles inspected per lot if the
submitted product is 1% defective?
No of defective articles = 5000x0.01=50 No of non-defective articles= 5000-50=4950
(a) n1p=100x0.01=1
From Poissons table, PI(0)=0.368
Accept the lot if
I)
Zero defective in the first sample
II)
One defective in the first sample and zero defective in the second sample
PI(1)= 0.736-0.368=0.368
For the second sample, N=5000-100=4900,
No of defectives=49
P=49/4900=0.01,
n2p=100x0.01=1
PII(0)=0.368
Therefore, probability of acceptance,
Pa=PI(0) + PI(1)xPII(0)
= 0.368+0.368x0.368=0.5034
(b) AOQ=Paxp=0.5034x0.01=0.005034 or 0.5034 %
Average no of articles inspected =PI(0)n1+PI(1)(n1+n2) + (1-Pa)N
i.e. 0.368 x100+0.368x0.368x200+ (1-0.5034)5000=2547
Designing a single sampling plan with stated value of P0.5
The quality P0.5 is the lot or process quality that has a probability of acceptance of 0.5 and is
called as the indifference quality.
The single acceptance plan is designed by using the following approximate formula
n = (c + 0.67) / P0.5
A vendor selects P0.5 = 0.02 and acceptance number 1. How many lots are likely to accepted if he
has 1000 lots each of quality 2.5% and 1.5%?
n=(1+0.67)/0.02 = 84.
1) np = 84(0.015)=1.26. From cumulative Poissons table (c=1), Pa=0.641
2) np= 84(0.025) = 2.1, From table, Pa = 0.38.
No. of lots likely to be accepted = 0.641(1000) + 0.38(1000)
= 1021

Module 23

Multiple Sampling Plans

A multiple sampling plan is an extension of double sampling plan more than two
samples are required
If at any stage the number of defective items is less than or equal to the corresponding
acceptance number, the lot is accepted.
If at any stage the number of defective items is greater than or equal to the corresponding
rejection number, the lot is rejected.
These plans generally involve less total inspection compared to single or double sampling plans
guaranteeing the same protection, but usually require higher administrative costs and higher
caliber inspection personnel
Sample

Sample
size

Combined samples
Size

Acc. No.

Rej. No.

First

n1

n1

c1

r1

Second

n2

n1+n2

c2

r2

Third

n3

n1+n2+n3

c3

r3

Fourth

n4

n1+n2+n3+n4

c4

r4

Fifth

n5

n1+n2+n3+n4+n5

c5

c5+1

Sequential Sampling Plans

Extension of double and multiple sampling plans
Theoretically can continue till 100% inspection is completed, but usually truncated after
the number inspected is equal to three times the number that would be inspected using a
corresponding single sampling plan.
If the sample size at each stage is greater than one, the process is usually called Group
Sequential Sampling.
If it is one at each stage, the procedure is called Item-By-Item Sequential Sampling.
Item-by-item sequential plans:
The cumulative observed number of defectives is plotted on the chart.
If the plotted points stay within the boundaries of acceptance and rejection lines, another
sample must be drawn.
If a point lies on or above the upper line, the lot is rejected.
When a sample point falls on or below the lower line, the lot is accepted.

Rejection

Acceptance

The equations for the two limit lines for specified values of p1, 1- , p2, and
(Acceptance line)
where

are

(Rejection line)

And

Numerical example:
Design a sequential sampling plan for the following specifications:
= 0.05,
= 0.2,

p1 = 0.1
p2 = 0.3

XA = -1.154 + 0.186 n and

XB = 2.054 + 0.186 n
Instead of using a graph to determine the lot disposition , the sequential sampling plan
can be displayed in the form of a table.
The entries in the table are found by substituting values of n into the above equations.
Acceptance and rejection numbers must be integers. Acceptance no. is the next integer
less than or equal to XA.
Rejection no. is the next integer greater than or equal to XB.
Characteristics of a good Acceptance Plan
he indices such as AQL, AOQL, etc should reflect the needs of the consumer and
producer.
The sampling risks (OC curve) should be known. Both producer and consumer should
have adequate protection.
Minimum total cost
Flexible enough to take into account changes in N,p, etc.
Should be useful in estimating individual lot quality and long run quality

Comparison of plans
In general double sampling is preferred.
Multiple and sequential sampling are used where cost of inspection is high.
In a single sampling plan the lot does not have to wait.
In multiple sampling plans indecision may continue for a long time more storage space
is necessary.
Published sampling plans
The sampling plans discussed so far are individual sampling plans.
A sampling scheme is an overall strategy specifying the way in which sampling plans are
to be used.
Military standard 105 E:
Developed during world war II
A derivative civilian standard also available ANSI / ASQC Z1.4
The primary focal point is AQL
Provides for single, double and multiple samplings
For each type provision is made for normal, tightened and reduced inspection
Dodge Romig sampling plan:
Dodge and Romig developed these set of inspection tables.
Two types of sampling plans are presented
LTPD protection
Specified AOQL
For both approaches, single and double sampling tables are available
For critical components and parts these plans are preferred to AQL oriented plans.