Introduction to Acceptance

Sampling
Acceptance Sampling
 Acceptance Sampling

● Accept/reject entire lot based on sample results

● Created by Dodge and Romig during WWII
● Not consistent with TQM of Zero Defects
● Does not estimate the quality of the lot

4
What is acceptance sampling?

Lot Acceptance Sampling

– A SQC technique, where a random sample is
taken from a lot, and upon the results of
appraising the sample, the lot will either be
rejected or accepted
– A procedure for sentencing incoming batches
or lots of items without doing 100% inspection
– The most widely used sampling plans are
given by Military Standard (MIL-STD-105E)
5
What is acceptance sampling?

• Purposes
– Determine the quality level of an incoming
shipment or at the end of production
– Judge whether the quality level is within the
level that has been predetermined

• But! Acceptance sampling gives you

no idea about the process that is
producing those items!

OPRE 6364 6
 Types of sampling plans

• Sampling by attributes vs. sampling by

variables
• Incoming vs. outgoing inspection
• Rectifying vs. non-rectifying inspection
– What is done with nonconforming items found
during inspection
– Defectives may be replaced by good items
• Single, double, multiple and sequential
plans
OPRE 6364 7
 How acceptance sampling
works
• Attributes(“go-no-go” inspection)
– Defectives-product acceptability across a
range
– Defects-number of defects per unit
• Variable (continuous measurement)
– Usually measured by the mean and
standard deviation

OPRE 6364 8
Why use acceptance sampling?

• Can do either 100% inspection, or inspect a

sample of a few items taken from the lot
• Complete inspection
– Inspecting each item produced to see if each
item meets the level desired
– Used when defective items would be very
detrimental in some way

OPRE 6364 9
 Why not 100% inspection?

Problems with 100% inspection

– Very expensive
– Can’t use when product must be destroyed to
test
– Handling by inspectors can induce defects
– Inspection must be very tedious so defective
items do not slip through inspection

OPRE 6364 10
 A Lot-by-Lot Sampling Plan

Count
N Accept or
n Number
(Lot) Conforming Reject Lot

• Specify the plan (n, c) given N

• For a lot size N, determine
– the sample size n, and
– the acceptance number c.
• Reject lot if number of defects > c
• Specify course of action if lot is rejected
OPRE 6364 11
 The Single Sampling Plan

• The most common and easiest plan to use but not

most efficient in terms of average number of samples
needed
• Single sampling plan
N = lot size
n = sample size (randomized)
c = acceptance number
d = number of defective items in sample
• Rule: If d  c, accept lot; else reject the lot

OPRE 6364 12
 Take a randomized
sample of size n
from the lot N
The Single
Sampling
Inspect all items in the
sample
procedure
Defectives found = d

Yes
dc? Accept lot

No
Reject lot

Return lot Do 100%

to supplier inspection

OPRE 6364 13
Acceptance Sampling Terms
• Acceptance quality level (AQL)
The smallest percentage of defectives that will
make the lot definitely acceptable. A quality
level that is the baseline requirement of the
customer
• RQL or Lot tolerance percent defective (LTPD)
The quality level that is unacceptable
to the customer
 How acceptance sampling
works
• Remember
– You are not measuring the quality of the lot,
but, you are to sentence the lot to either
reject or accept it
• Sampling involves risks:
– A good product may be rejected
– Bad products may be accepted
 Because we inspect only a sample, not the
whole lot!
OPRE 6364 15
Producer’s & Consumer’s Risks
There are two types of risks that can occur in acceptance
sampling:
1. The risk that a good quality product will be rejected
(Producer's Risk).
2. The risk that a product of unsatisfactory quality will be
accepted (Consumer's Risk)

• TYPE I ERROR = P(reject good lot)

 or Producer’s risk 5% is
common

• TYPE II ERROR = P(accept bad lot)

 or Consumer’s risk 10% is
a typical value
16
 Acceptance sampling contd.
• Producer’s risk
– Risk associated with a lot of acceptable quality
rejected
• Alpha 
= Prob (committing Type I error)
= P (rejecting lot at AQL quality level)
= producers risk

OPRE 6364 17
 Acceptance sampling contd.

• Consumer’s risk
– Receive shipment, assume good quality, actually bad
quality

• Beta 
= Prob (committing Type II error)
= Prob (accepting a lot at RQL quality level)
= consumers risk
The OC curve for a sampling plan quantifies these
risks
OPRE 6364 18
Operating Characteristic (OC) Curve

• It is a graph of the % defective (p) in a lot or batch vs.

the probability that the sampling plan will accept the lot
• Shows probability of lot acceptance Pa as function of
lot quality level (p)
• It is based on the sampling plan
• Curve indicates discriminating power of the plan
• Aids in selection of plans that are effective in reducing
risk
• Helps to keep the high cost of inspection down

OPRE 6364 19
Operating Characteristic Curve
1.00
 = 0.05 {
Probability of acceptance, Pa

0.80

OC curve for n and c

0.60

0.40

0.20

 = 0.10 {
0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
Proportion defective p
AQL OPRE 6364 LTPD 19
 Types of OC Curves

• Type A
– Gives the probability of acceptance for an individual
lot coming from finite production
• Type B
– Give the probability of acceptance for lots coming
from a continuous process or infinite size lot

OPRE 6364 21
Type A OC Curves
 Type B OC Curves
To construct a type B OC curve, we assume that a stream of lots is
produced by the process and that the lot size is large (at least 10 times)
compared to the sample size. In this situation, a binomial distribution can be
used to find the probability of observing x nonconforming items in a sample
of size n. Assuming the lot proportion nonconforming is p, this probability
is given by

If the lot size is large and the probability of a nonconforming item is small, a
Poisson distribution can be used as an approximation to the binomial
distribution. The probability of x nonconforming items in the sample is
found from

where λ=np represents the average number of nonconforming items in the

sample.
Example: Construct an OC curve for a single sampling plan
where the lot size is 2000, the sample size is 50, and the
acceptance number is 2.

Solution:
1. We are given N= 2000, n= 50, and c= 2.
2. The probability of lot acceptance is equivalent to the
probability of obtaining two or fewer nonconforming items in
the sample.
3. The Poisson probability distribution in Appendix A-2 is used
to obtain the lot acceptance probability for different values of
the proportion nonconforming p.
4. Let’s suppose that p is 0.02 (i.e., the batch is 2% non-
conforming). Since np= (50) (0.02) = 1.0, the probability Pa of
accepting the lot is 0.920.
5. Table 10-1 shows values of Pa for various values of p.
6. In some instances, the probability values are linearly
interpolated from Appendix A-2.
7. A plot of these values is the OC curve.
1. The discriminating power of the sampling plan N= 2000, n= 50, c= 2
can be seen from the OC curve in Figure.
2. If a series of batches, each of which is 1% nonconforming, comes in
for inspection, then (using this plan) the probability of lot
acceptance is 0.986.
3. It means that, on average, the sampling plan will accept about 986
out of 1000 such batches.
4. On the other hand, if batches are 5% nonconforming, only about 544
out of 1000 batches will be accepted.
5. As the lot quality becomes poorer, the probability of lot acceptance
decreases, as it should.
6. The steeper the drop in the probability of lot acceptance as lot
quality worsens, the higher the discriminatory power of the sampling
plan.
7. Producer and consumer risk can also be demonstrated through the
OC curve.
8. Suppose that our numerical definition of good quality (AQL) is 0.01
and that of poor quality (LQL) is 0.11. From the OC curve, the
producer’s risk α is 1-0.986 = 0.014.
9. We consider batches that are 1% nonconforming to be good. If our
sampling plan is used, such batches will be rejected about 1.4% of
the time.
Effect of the Sample Size and the Acceptance Number
1. The parameters n and c of the sampling plan affect the
shape of the OC curve.
2. As long as the lot size N is significantly large compared to
the sample size n, the lot size does not have an
appreciable impact on the shape of the OC curve.
3. For fixed values of N and c, as the sample size becomes
larger, the slope of the OC curve becomes steeper,
implying a greater discriminatory power.
4. The figure below shows the OC curves for three sampling
plans.
5. Note that, for lots that are 2% nonconforming, the
sampling plan N= 2000, n= 50, c= 2 will accept such lots
about 92% of the time. However, for the same lots, the
sampling plan N= 2000, n= 200, c= 2 will accept them only
23.8% of the time. Changing the sample size from 50 to 200
causes a drop in the acceptance probability of 68.2%.
1. For fixed values of the lot size N and the sample size n, as the
acceptance number decreases, the slope of the OC curve becomes
steeper. The figure below shows the OC curves for four sampling plans.
2. The probability of acceptance decreases for a given lot quality as the
acceptance number c decreases.
3. For the acceptance number is zero, the OC curve starts dropping
drastically even as the proportion nonconforming deviates slightly from
zero.
4. This may not be desirable from a producer’s point of view. For example,
if lots that are 0.5% nonconforming are considered acceptable, the
sampling plan N= 2000, n= 50, c= 0 will reject such lots about 22% of the
time, implying a high value of the producer’s risk.
5. Sampling plans with c= 0 do not have the desirable inverted-S shape of
the ideal OC curve. They are, however, stringent and serve a need.
6. The chosen values of n and c should be such that they match the goals
of the user. Given some desirable producer’s risk and the associated
quality level of a good lot (AQL) and/or a desirable consumer’s risk and
an associated quality level of a poor lot (LQL), the combination of n and
c that produces an OC curve that matches these goals will provide an
acceptable sampling plan.
 The Ideal OC Curve
● Ideal curve would be perfectly perpendicular from 0 to
100% for a fraction defective = AQL
● It will accept every lot with p  AQL and reject every lot
with p > AQL

33
 TYPES OF SAMPLING PLANS
1. There are three types of attribute sampling plans: single, double, and
multiple.
2. In a single sampling plan, the information obtained from one sample is
used to make a decision to accept or reject a lot.
3. There are two parameters in this sampling plan: the sample size n and
the acceptance number c.
4. The plan operates as follows. A random sample of size n is selected
from the batch. The number of nonconforming items or non-
conformities in the sample is found and compared to the acceptance
number c. If the observed number is less than or equal to the
acceptance number, the lot is accepted. If more than c nonconforming
items or nonconformities are found in the sample, the lot is rejected.

5. A double sampling plan involves making a decision to accept the lot,

reject the lot, or take a second sample. If the inference from the first
sample is that the lot quality is quite good, the lot is accepted. If the
inference is poor lot quality, the lot is rejected. If the first sample gives
an inference of neither good nor poor quality, a second sample is taken.
Thereafter, based on the combined number of nonconforming items or
nonconformities in both samples, a decision is made to accept or reject
the lot.
6. The parameters of a double sampling plan are as follows:
n1:size of the first sample,
c1: acceptance number for the first sample,
r1: rejection number for the first sample,
n2: size of the second sample,
c2: acceptance number for the second sample,
r2:rejection number for the second sample
 Multiple sampling plans
1. Multiple sampling plans are an extension of double sampling plans.
2. Three, four, five, or as many samples as desired may be needed to
make a decision regarding the lot.
3. The sampling plan can be terminated at any stage once the
acceptance or rejection criteria have been met.
4. The sample sizes in multiple sampling plans are usually less than
those for an equivalent double sampling plan, which in turn are
usually less than those for an equivalent single sampling plan.
5. The ultimate extension of the multiple sampling plans is the
sequential sampling plan, which is an item-by-item inspection plan.
After each item is inspected, a decision is made to accept the lot,
reject the lot, or choose another item for inspection, depending on
whether the observed cumulative number of nonconforming items
is less than or equal to the acceptance number, greater than or equal
to the rejection number, or in between the two.
 Advantages and Disadvantages
1. We can design single, double, or multiple sampling plans that are equivalent in
the sense that they have the same probability of lot acceptance for batches of a
given quality.
2. As far as simplicity is concerned, the single sampling plan is the best, followed by
double and then multiple sampling plans.
3. Administrative costs for record-keeping, training and inspection are the least for
single and the highest for multiple sampling plans.
4. On average, for equivalent plans, the number of items inspected to make a
decision regarding the lot is usually more for a single sampling plan. This is
because double and multiple sampling plans use fewer items in their samples, so
if the lots are of very good or poor quality, a decision to accept or reject them is
made quickly.
5. Inspection costs will therefore be the most for single and the least for multiple
sampling plans.
6. The information content of the samples is a function of the sample size; the more
samples we inspect, the more information we have about the product and
consequently the process. Single sampling plans provide the most information
and multiple sampling plans the least.
EVALUATING SAMPLING PLANS
1. The OC curve is one measure of the performance
of a sampling plan.
2. The following are other measures to evaluate the
goodness of a sampling plan.
1. The average quality level of batches leaving the
inspection station,
2. The average number of items inspected before making a
decision on the lot
3. The average amount of inspection per lot if a rejected
lot goes through 100% inspection.
3. We discuss single sampling plans here, but the
concepts apply to all three plans.
 Average Outgoing Quality
1. Rectifying inspection: The lots that are rejected through sampling plans. go
through 100% inspection, known as screening, where nonconforming items are
replaced with conforming ones. Such a procedure is known as rectification
inspection.
2. The average outgoing quality (AOQ) is the average quality level of a series of
batches that leave the inspection station, assuming rectifying inspection, after
coming in for inspection at a certain quality level p.
3. AOQ is the expected outgoing average percent defective after inspecting
incoming materials or products
4. The AOQ is not the quality level of a single batch that leaves the inspection
station. For instance, a batch with incoming quality level p will leave the
inspection station with about the same quality level if accepted by the sampling
plan.
5. We assume the sample is a small enough proportion of the lot such that if
nonconforming items are found in the sample and replaced with conforming
ones, the quality level of the lot is not significantly affected.
6. Similarly, another batch with the same incoming quality p that is rejected by the
sampling plan will be screened and so will leave the inspection station with no
nonconforming items. (This assumes that screening detects all nonconforming
items.)
7. The AOQ measures the average quality level of a large number of such batches
of incoming quality p, the proportion of nonconforming in the lots, assuming
rectification.
1. Consider, N as the lot size, n as the sample size, p as the incoming
lot quality, and Pa as the probability of accepting the lot.
2. The average outgoing quality is given by:

3. During acceptance sampling, n items in the sample will have no

nonconforming items after they have been inspected.
4. If the lot is rejected by the sampling plan, the N –n items left in the
lot go through screening so that no nonconforming items are in
the outgoing product.
5. Only if the lot is accepted by the sampling plan, the N-n items left
in the lot will leave the inspection station with p(N-n)
nonconforming items.
6. However, the probability that the batch will be accepted by the
sampling plan is Pa. So, Pa p(N -n) is the number of nonconforming
items per lot expected to leave the inspection station.
Average Outgoing Quality Limit (AOQL)
1. The average outgoing quality limit (AOQL) is
the maximum value, or peak, of the AOQ
curve.
2. It represents the worst average quality that
would leave the inspection station, assuming
rectification, regardless of the incoming lot
quality.
3. The AOQL value is also a measure of the
goodness of a sampling plan. Note that the
protection offered by the sampling plan, in
terms of the AOQL value, does not apply to
individual lots. It holds for the average quality
of a series of batches.
1. In the above AOQ curve, the AOQL value is approximately 0.0265, or 2.65%.
2. This means that for the sampling plan above, N= 2000, n= 50, c= 2, we have some
protection against the worst quality for a series of batches that leave the inspection
program.
3. The average quality level will not be poorer than 2.65% non-conforming.
4. However, it is possible for an individual lot to have an outgoing quality level of more
than 2.65% non-conforming.
5. The AOQL value and the shape of the AOQ curve depend on the particular sampling
plan.
6. Sampling plans are designed such that their AOQL does not exceed a certain specified
value.
Average Total Inspection
1. If rectifying inspection is conducted for lots rejected by the sampling
plan, another evaluation measure is the average total inspection
(ATI).
2. The ATI represents the average number of items inspected per lot.
3. If a lot has no nonconforming items, it will obviously be accepted by
the chosen sampling plan, and only n items (the sample size) will be
inspected for a lot.
4. At the other extreme, if the lot has 100% nonconforming items, the
number inspected per lot will be N (the lot size) assuming that
rejected lots are screened.
5. For a lot of quality between these extremes, the average amount
inspected per lot will vary between these two values.
6. For single sampling plans, the average total inspection per lot for
lots with an incoming quality level of p is given by

7. Here Pa represents the probability of accepting a lot that has an

incoming quality level of p. A plot of the average total inspection
versus p is an ATI curve. Note that for an individual lot the amount
inspected is either n or N.
 Average Sample Number
1. The average number of items inspected for a series of lots with a
given incoming lot quality in order to make a decision is known as
the average sample number (ASN).
2. Assume that inspection is not curtailed for a single sampling plan
when making a decision. For example, if 3 nonconforming items are
found by the twentieth unit when using a single sampling plan N =
800, n = 60, c = 2, even though a decision can be made after the
twentieth unit to reject the lot, inspection continues for all 60 items
in the sample. Under this assumption, the average sample number
for a single sampling plan is equal to the sample size n.
3.

4. Where p1 is
1. For single sampling plans, ASN is constant and is represented
by a horizontal line.
2. Typically, for equivalent plans, the ASN for a double sampling
plan is below that for a single sampling plan.
3. It sometimes happens that the middle segment of the ASN
curve for a double sampling plan is above the ASN for a single
sampling plan.
4. Management must then use historical information to obtain an
idea of the process quality. If the quality level p happens to fall
in the segment where the ASN is greater for a double sampling
plan, the choice of a single sampling plan may be justified to
cut down on inspection time and costs.
5. On the other hand, estimates of very high or low levels of
process quality may justify the use of a double sampling plan
that will yield smaller values of ASN.
6. If curtailed inspection is used on the second sample of a double
sampling plan, the ASN curve will be lowered further and will
become even more attractive than a single sampling plan.
 MIL-STD-105E
1. MIL-STD-105E is a standard sampling methodology
that has been established for the acceptance or
rejection of lots based on the evaluation of item
attributes.
2. The batch or lot is either accepted or rejected
based on one or more samples taken from a batch
or lot containing N units.
3. It is similar to the ANSI Z1.4 and ISO 2859
standards.
4. It includes the specification of sample sizes,
acceptance numbers, and rejection numbers.
5. It also contains rules for switching between normal
inspection, tightened inspection, and reduced
inspection, based on the results of recent samples.
 ACCEPTANCE SAMPLING Based on AQL
1. AQL based acceptance sampling accept or reject a
production lot based on a representative sample size.
2. AQL is basically defined as the lowest or worst quality
level, or limit, that is acceptable in an order of goods.
This is generally measured in quality defects found, or
pieces found with quality defects, in the inspected
sample size.
3. Two potential problems can result from using the wrong
sample size for inspection:
1. Inspecting too few pieces to accurately assess the total order.
You risk discovering only after receiving the total shipment at
your warehouse that there are more defects than you
anticipated based on the inspection result.
2. Inspecting too many pieces than necessary to accurately
assess the total order. This can raise your QC costs and delay
production or shipping.
 ANSI-ASQ Z.14
1. ANSI-ASQ Z.14 was developed by the American
National Standards Institute (ANSI) and the
American Society for Quality (ASQ).
2. It is an AQL QC sampling
3. The ANSI-ASQ Z.14 helps to determine:
1. The appropriate sample size you need to inspect,
determined by your lot size
2. The maximum number of defects you will accept in the
order, sometimes known as an “accept point”, as
determined by your defect tolerance, or AQL
4. The AQL table outlines several different sample
sizes, which vary slightly in size and risk. While all
the sample sizes are statistically valid, some will
provide more insight into order quality than others.
It’s up to you to select your desired sample size
based on your appetite for risk.
WHAT ARE THE THREE AQL GENERAL INSPECTION LEVELS?

1. There are three general AQL inspection levels: “GI”, “GII” and “GIII”.
2. Each indicate sample sizes for a non-destructive inspection.
3. The inspector generally inspects the entire sample size for quality
issues related to performance, function and visual appearance. Any
quality defects found during inspection contribute to the overall
“pass” or “fail” result.
4. GI, GII and GIII sample sizes for a lot size is given in the AQL table.
5. GI inspection level, also called Reduced Inspection, offers the
smallest sample sizes of the three general inspection levels.
Considered as the “budget option”, GI may be your best option if
you’re particularly short on time or money.
6. GII inspection level, also called normal inspection level, is often
especially fitting for the first inspection of a lot. Seeing these first
results can help you determine if GII is still suitable going forward,
or if you should raise or lower your inspection level.
7. The GIII inspection level, also called Tighten Inspection, provides the
largest sample size for your lot out of the three AQL inspection
levels. GIII minimizes your risk of unknowingly accepting a shipment
with a higher portion of defective pieces than found during
inspection.
 Special Inspection levels
S-1. Select this level when the act of inspection damages or destroys the
inspection part. Use this inspection level only if the inspection costs are
extremely severe to your company. It provides the lowest protection to your
customer.
S-2. Select this level when the act of inspection damages or destroys the
inspection part. Useful for features of minor importance. A minor defect will
not prevent the parts usability for its intended purpose. It provides less
protection to the customer. You will incur the least inspection costs with this
plan.
S-3. Select this level when the act of inspection damages or destroys the
inspection part. Useful for inspection of features with major importance. A
major defect is likely to result in failure when the part is used but the failure
does not cause hazardous or unsafe condition. When using this plan you will
encounter lower inspection cost.
S-4. Select this level when the act of inspection damages or destroys the
inspection part. Use this plan when inspecting features that could cause
critical defects. Critical causes a hazardous or unsafe condition. When using
this plan you will encounter lower inspection costs.
 Normal vs. Reduced vs. Tighten
Inspection Reduced Inspection

1. Start with normal inspection level.

2. Switch to reduce inspection, if 10 or more lots
consecutively pass inspection. Further, the
total number of defects from those 10 lots must
be less than specified in a table in the
standard. In addition the production must be at
a steady rate.
3. If sampling rejects 2 out of 5 lots, switch to
tighten inspection. To return back to normal
inspection, 5 consecutive lots with tighten
inspection sample size must pass inspection.
How Does the Standard Work?
1. First you determine the lot size
2. Then you determine the inspection level
3. Then you review the table(s) in Mil Std 105e or
ANSI-ASQ Z.14
4. The table provides your sample size and
accept and reject criteria
5. You then select your sample from the lot
6. Inspect the parts
7. If you have less defects than the reject number
you accept the lot. Else you reject the lot