Acceptance Sampling
Techniques
by
Attributes
Assoc. Prof. Ir. Dr. Cheong Kuan Yew
School of Materials & Mineral Resources Engineering
Engineering Campus
Universiti Sains Malaysia
Topic Outcome:
At the end of this topic, students will be able to:
Discuss the properties of an operating-characteristic
(OC) curve.
Design, construct, and use an OC curve.
Describe the consumer-producer relationship from a
OC curve.
Determine and explain producer’s risk, consumer’s
risk, AQL, and LQ.
Apply a sampling plan.
Topic Outline:
Operating Characteristics (OC) Curve
Introduction.
Methods of calculating the probability of acceptance.
Construction of an OC curve using Poisson
Distribution.
Properties of an OC curve
Consumer-Producer Relationship
Producer’s Risk (α ) & AQL
Consumer’s Risk (β ) & LQ
Operating-Characteristic (OC)
Curve – An Introduction
An OC curve is a graph of Lot Nonconforming (or Percent
Nonconforming, 100p0) versus Probability that a sampling
plan would accept the lots,
lots Pa (or Percent of Lots
Accepted, 100Pa).
Material with 0 nonconforming
a
Accepted always
P001
Pa = 1.0
Material with 100% nonconforming
Rejected always 100p0
Pa = 0
What is the usage of OC curves?
It shows the chance of a lot being
accepted for a particular incoming process
quality.
quality
It shows the discriminatory power
of a sampling plan.
Ideal OC Curve
• All lots >5%
nonconforming have
[or Percent of Lots Accepted,100Pa]
a probability of
acceptance of 0.
Probability of Acceptance, Pa
1.0
• All lots <5%
nonconforming have
a probability of
Acceptance
Rejection
Region
Region
0.5 acceptance of 1.0.
0
5.0 10.0
Percent Nonconforming (100p0)
In actual practice, no sampling plan exists
that can be discriminate perfectly.
perfectly
0.5
0
1.0
Non-Ideal OC Curve
5.0 10.0
Percent Nonconforming (100p0)
Summary of common probability distributions
Probability Distributions
Discrete Continuous
Uniform Uniform
Binomial Normal
Geometric Gamma
Hypergeometric Erlang
Weibull
Possion
Methods of Calculating the Probability of Acceptance:
For attribute sampling, the following distributions are used to
calculate the probability of acceptance.
Distribution Formula Conditions
Hypergeometric 1) Population is
CdDCnN−−dD FINITE.
P( d ) = 2) Random sample is
N
Cn taken without
D!
•
( N − D )! replacement.
d !( D − d )! ( n − d )!( N − D − n + D )! 3) n/N ≤ 0.10 can
P( d ) =
N! be approx. by
n!( N − n )! binomial distribution.
Distribution Formula Conditions
P( c ) =
( np0 ) c − np
e 0
When these
c!
assumptions are
met, the Poisson
Distribution is
preferable because
of the ease of
calculation.
Steps:
1) Click icon of fx.
2) Function Category: Statistical.
3) Function name: Poisson.
4) Click OK
5) x(number of events) = 2
6) Mean (np0) = 1.8
7) Cumulative: Type in “TRUE” [note: “FALSE” non-
cumulative]
Answer = 0.731
Syntax:
POISSON(x,mean,cumulative)
Steps of Constructing an OC curve:
1) Assume p0 value
2) Calculate np0 value
3) Attain Pa values from Poisson Table using applicable c
and np0 values or from EXCEL program
4) Plot point (100p0 vs.100Pa)
5) Repeat steps 1 to 4 until a smooth curve is obtained.
100
90
It shows the chance of a
Percent of Lots Accepted (100Pa )
80
40
process quality.
30
20 e.g.:
10
• Incoming process
0
0 1.5 3 4.5 6
Percent Nonconforming (100p0)
quality = 2.3%
• 66% of the lots is
expected to be
accepted.
The above OC curve is unique to the single
sampling plan defined by N = 3000, n = 89, and c = 2.
1.0
without question.
Middle portion between shoulder and
tail
At exact centre of the curve, where the Pa
0.5 is 50%, product of corresponding quality
has 50-50 chance of being either rejected
or accepted.
Example:
A process capability is 99%. Customer is willing to accept 1%
nonconforming units. However, during inspection a certain
percentage of lots still rejected ever though the percent of
nonconforming is less than 1%.
How to estimate Producer’s Risk (α )
1) Plot OC curve for the sampling plan.
2) Find the percentage nonconforming (100p0) in the
process.
• Rough estimation can be done if no exact data is
available.
• However for a more accurate check, a process
capability study is preferable.
1) Use the process capability percentage to determine the
probability of acceptance (Pa) on the OC curve.
2) Producer’s Risk (α ) = 1 - Pa
Q & A:
90 Percentage of Lots
Percent of Lots Accepted (100Pa )
70
c=4 c 100Pa 100α
60 0 44.9% 55.1%
c=3
50
1 80.9% 19.1%
40
c=2
30 2 95.2% 4.8%
20
c=1
3 99.1% 0.9%
10
c=0 4 99.8% 0.2%
0
0 1 2 3 4 5 6
Percent Nonconforming (100p 0)
Acceptable Quality Level (AQL)
It is associated with producer’s risk.
It is a numerical definition of an acceptable lot.
It is the maximum percent nonconforming that can be
considered satisfactory for the purposes of acceptance
sampling.
It is a reference point on the OC curve and is not
meant to convey to the producer that any percent
nonconforming is acceptable.
It is a statistical term and is not meant to be used by
general public.
Producer only can guarantee an acceptable lot when 0%
nonconforming or the number nonconforming in the lot
less than or equal to acceptance number, c.
Producer’s quality goal is to meet or exceed the
specifications so that no nonconforming units are
present in the lot.
A sampling plan should have a low Producer’s risk for
quality that is equal to or better than the AQL.
Example:
30
• 7 out of 100 lots that
are 0.7% 20
100
90
Percent of Lots Accepted (100Pa )
80
70
c=4 c 100Pa=100β
60 0 4.1%
c=3
50
1 17.1%
40
c=2
30 2 38.0%
20
c=1
3 60.2%
10
c=0 4 78.1%
0
0 1 2 3 4 5 6
Percent Nonconforming (100p 0)
Limiting Quality (LQ)
It is associated with consumer’s risk.
It is a numerical definition of a nonconforming lot.
It is the percent nonconforming in a lot for which (for
acceptance purposes) the consumer wishes the
probability of acceptance to be low.
For previous example, LQ = 4% for 100β = 5% (for c=0)
lots that are 4% nonconforming will have a 5%
chance of being accepted. 1 out of 20 lots that are
4% nonconforming will be accepted by this sampling
plan.
Design of Sampling Plan
Design of Sampling Plan
“Rule of Thumb”
A sampling plan should not be adopted
without seeing the OC curve!!!!!
END