You are on page 1of 29

SUPPLEMENT TO

CHAPTER 10

Acceptance Sampling

Supplement Outline Solved Problems, 14 Learning Objectives


Introduction and Sampling Plans, 2 Discussion and Review After completing this supplement,
Operating Characteristic Curve, 3 Questions,16 you should be able to:
Determining Single Sampling Internet Exercises, 16
LO1
Explain acceptance sampling,
Plans, 9 Problems, 16
and contrast single and
Average Quality of Inspected Lots Mini-case: CRYSTAL S.A., 19 multiple samplingplans.
and a Related Sampling Plan, 12 Mini-case: Gustave Roussy LO2
Construct and use an
Key Terms, 14 Institute,20 operating characteristic curve.
LO3
Determine single sampling
plans.
LO4
Determine the average
quality of inspected lots, and
determine a relatedsampling
plan.
2 PART FOUR Quality

LO1 Introduction and Sampling Plans


acceptance sampling Acceptance sampling is a form of inspection that is applied to lots or batches of items
A form of inspection applied either before or after a process. In the majority of cases, the lots represent incoming pur-
to lots or batches of items chased items or final products awaiting shipment to warehouses or customers. The pur-
before or after a process pose of acceptance sampling is to decide whether a lot satisfies predetermined standards
tojudge conformance with (specifications) for important characteristics of the item. Lots that satisfy these standards
predetermined standards are passed or accepted; those that do not are rejected. Rejected lots may be subjected to
(specifications). 100 percent inspection or, if purchased, returned to the supplier for credit or replacement
(especially if destructive testing is involved).
The alternatives to acceptance sampling are (a) 100% inspection and (b) no inspection.
The decision of which one to choose is mainly based on the costs. A measure used is the
breakeven point BEP1:
BEP=cost of inspection per item / cost of later repair due to a defective item
Let P=estimated proportion of defectives in the lot. The decision is:
If PBEP, use acceptance sampling
If P>BEP, use 100% inspection
If P<BEP but P is variable, use acceptance sampling
If P<BEP and P is stable, dont inspect.

Example S-1 A car assembly plant receives headlight assemblies from a new supplier in lots of 1,000
units. The cost of testing a headlight assembly at receiving is $.10. If a defective headlight
is assembled on a car, the cost of disruption to the assembly line to replace the headlight
is $100. It is expected that only .05% to .1% of the headlight assemblies will be defective.
Determine the extent of inspection for the headlight assemblies.

Solution BEP=$.1/$100=.001. Because .001.05% to .1%, perform acceptance sampling on


thelot.

Non-economic factors favouring acceptance sampling vs. 100% inspection include:


1. Time is short.
2. Destructive testing is required.
3. Fatigue or boredom caused by inspecting large numbers of items leads to inspection
errors.
Acceptance sampling can be applied to both attribute (counts) and variable (measure-
ments) inspection. However, inspection of attributes is more widely used, so the discussion
here focuses exclusively on attribute acceptance sampling.
A key element of acceptance sampling is the sampling plan. Given the lot size N, a
sampling plan A plan that sampling plan specifies the sample size, n; the number of samples to be taken; and the
specifies the sample size, the acceptance/rejection criteria. A variety of sampling plans can be used. Some plans call for
number of samples, and the selection of a single sample, and others call for two or more samples, depending on the
acceptance/rejection criteria. nature of the plan. The following briefly describes some of the different types of sampling
plans. In this supplement, our focus will be on single sampling, which is the simplest and
most commonly used.

Single-Sampling Plans
In this plan, one random sample is drawn from the lot, and every item in the sample is
examined and classified as either good or defective. If the sample contains less than
1
A. V. Feigenbaum, Total Quality Control, 3rd ed., New York: McGraw-Hill, 1991, pp. 50405.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 3

or equal to a specified number of defectives, c, the lot is accepted; otherwise (if it contains
more than c defectives) the lot is rejected.

Double-Sampling Plans
A double-sampling plan allows taking a second sample if the results of the initial sample is
inconclusive. Specifically, if the quality of the initial sample is high, the lot can be accepted
without the need for a second sample. If the quality in the initial sample is poor, the lot
is rejected (and there is also no need for a second sample). For results between those two
cases, a second sample is then taken and the items inspected, after which the lot is either
accepted or rejected on the basis of the evidence obtained from both samples. A double-
sampling plan specifies the size of the initial sample, the accept/reject criteria for the initial
sample, the size of the second sample, and a single overall acceptance number.
With a double-sampling plan, two values are specified for the number of defective items
in the first sample, a lower level, c1, and an upper level, r1. For instance, the lower level
might be two defectives and the upper level might be five defectives. If the number of
defective items in the first sample is less than or equal to the lower value (i.e., c1), the lot
is judged to be good and sampling is terminated. Conversely, if the number of defectives
in the first sample equals or exceeds the upper value (i.e., r1), the lot is rejected. If the
number of defectives in the first sample falls somewhere in between c1 and r1, a second
sample is taken and the total number of defectives in both samples is compared to a third
value, c2. For example, c2 might be six. If the combined number of defectives does not
exceed c2, the lot is accepted; otherwise, the lot is rejected.

Multiple-Sampling Plans
A multiple-sampling plan is similar to a double-sampling plan except that more than two
samples may be required. A multiple sampling plan will specify each sample size and
two limits for each sample. If, for any sample, the cumulative number of defectives found
(i.e., those in the present sample plus those found in all previous samples) is greater than
or equal to the upper limit specified for that sample, sampling is terminated and the lot
is rejected. If the cumulative number of defectives is less than or equal to the lower limit,
sampling is terminated and the lot is accepted. If the cumulative number of defectives
is between the two limits, another sample is taken. The process continues until the lot is
either accepted or rejected.

Choosing a Sampling Plan


The cost and time required for inspection often dictate the type of sampling plan used.
The two primary considerations are the number of samples needed and the total number
of observations required. Single-sampling plans involve only a single sample, but the
sample size is larger than the expected number of observations taken under double- or
multiple-sampling plans.This stems from the fact that a very good or very poor quality lot
will often be accepted or rejected early in a multiple-sampling plan, and sampling can be
terminated. Where the cost to obtain a sample is relatively high compared with the cost
to analyze the observations, a single-sampling plan is more desirable. Conversely, where
item inspection costs are relatively high, such as destructive testing, it may be better to use
double or multiple sampling because the average number of items inspected per lot will
be lower. Another advantage of a single sampling plan is that it is easy to understand and
use. For this reason, we will focus on single sampling in this supplement.

Operating Characteristic Curve


LO2

An important feature of a sampling plan is how it discriminates between lots of high and
low quality. The ability of a sampling plan to discriminate between lots of high and low operating characteristic
quality is described by its operating characteristic (OC) curve. A typical OC curve for (OC) curve Curve that shows
a single-sampling plan is shown in Figure 10S-1. The curve shows the probabilities of the probabilities of accepting
accepting lots with various qualities (proportion defectives). For example, it shows that a lots with various quality
lot with 3 percent defectives (a proportion of defectives of .03) would have a probability (proportion defective).
4 PART FOUR Quality

FIGURE 10S-1 1.00

A typical OC curve .90

.80

Probability of accepting the lot


.70

.60

.50

.40

.30

.20

.10

.00
0 .05 .10 .15 .20 .25
3% Lot quality (proportion defective)

of .80 of being accepted (or a probability of 1.00.80=.20 of being rejected). Note the
downward relationship: as lot quality decreases, the probability of lot acceptance decreases,
although the relationship is not linear.
A sampling plan cannot provide perfect discrimination between good and bad lots; some
low-quality lots will inevitably be accepted, and some high-quality lots will inevitably be rejected.
Even lots containing more than 20 percent defectives still have some probability of being accepted,
whereas lots with as few as 3 percent defectives have some chance of being rejected.
The degree to which a sampling plan discriminates between good and bad lots is a func-
tion of the steepness of its OC curve: the steeper the curve, the more discriminating the
sampling plan (see Figure 10S-2.) Principles of sampling imply that the larger the sample
size n, the steeper the curve. Note the curve for an ideal plan (i.e., one that can discriminate
perfectly between good and bad lots). To achieve that, you need to inspect 100 percent of
each lot. Obviously, if you are going to do that, theoretically all of the defectives can be
eliminated (although errors due to boredom might result in a few defectives remaining).
However, the cost of additional discrimination may be larger than the cost of additional
inspection (i.e., it may not be cost-effective).

FIGURE 10S-2 1.00


The steeper the OC curve,
Ideal
the more discriminating the
sampling plan
Probability of accepting the lot

Not very
discriminating

Better

0
Good Bad
Lot quality (proportion defective)
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 5

Buyers (or consumers) are generally willing to accept lots that contain small percent- acceptable quality level
ages of defectives as good, especially if the cost related to a few defects is low. Often (AQL) The percentage of
this percentage is in the range of .01 to 2 percent. This figure is known as the acceptable defects at which a consumer
quality level (AQL). AQL should be set based on the criticality of the characteristic that (buyer) is willing to accept lots
is being inspectedthe more critical the characteristic, the smaller the AQL should be. as good.
For example, for spoons, the defect of being cracked may have AQL of .5% whereas the
defect of being scratched may have AQL of 2%. Also, AQL should be set so that the good
incoming lots have quality equal to AQL. Otherwise the supplier will be overwhelmed
with rejected lots and there may not be enough accepted lots for the buyer to continue
production. For example, if the incoming lots generally have 1 to 4% defectives, then AQL
should be set to 1%.
Because of the inability of random sampling to identify all lots that contain more than
AQL percentage of defectives, consumers (buyers) recognize that some lots that actually
contain more defectives than AQL will be accepted. However, there is usually an upper
limit on the percentage of defective items that a consumer is willing to tolerate in accepted
lots. The percentage just larger than this is known as the lot tolerance percent defective lot tolerance percent
(LTPD). Thus, consumers want quality equal to or better than the AQL, and are willing to defective (LTPD) The
live with poorer quality, but they prefer not to accept any lots with a defective percent- percentage just larger than the
age greater or equal to the LTPD. LTPD should also be set based on the criticality of the upper limit of the percentage
characteristic that is being inspectedthe more critical the characteristic, the smaller the of defectives of a lot that a
LTPD. Also, LTPD should be set so that the bad incoming lots have quality equal to LTPD. consumer is willing to accept.
Otherwise the seller may receive more rejected lots than expected or the buyer may lose
consumers risk The
the opportunity to demand higher quality for bad lots. For example, if the incoming lots
probability that a bad lot
generally have 1to 4% defectives, then LTPD should be set to 4%.
containing defects equal to the
As mentioned above, sampling plans are not perfect in discriminating between good and
LTPD will be accepted on the
bad lots, i.e., mistakes will be made in acceptance of bad lots and rejection of good lots.
basis of sample data.
The probability that a bad lot containing defectives equal to the LTPD will be accepted is
known as the consumers risk, or beta ( ), or the probability of making a Type II error. On
producers risk The
the other hand, the probability that a good lot containing defectives equal to the AQL will
probability that a good lot
be rejected is known as the producers risk, or alpha (), or the probability of making a
containing defects equal to the
Type I error. Many sampling plans are designed to have a producers risk of 5 percent and
AQL will be rejected on the
a consumers risk of 10 percent, although other combinations are also used. Figure 10S-3
basis of sample data.
illustrates an OC curve with the AQL, LTPD, producers risk, and consumers risk. Note
that the probability of accepting a lot with AQL quality is 1 .
A certain amount of insight is gained by constructing an OC curve. The probability of
observing up to and including c defectives in a sample of size n from a lot with proportion
of defectives P is given by cumulative hyper-geometric formula:

NP N NP NP ! ( N N P )!
c x n x c
x !( NP x )! ( n x )!( N NP n + x )!
P( x c ) =
x =0
N
=
x =0
N!
n n !( N n )! (10S-1)

N
where x=number of defectives in the sample, and represents the number of combina-
n
tions (different ways) of choosing a sample of n from the lot N, etc. Note that the expected
number of defectives in the lot is NP. Intuitively, probability of observing x defectives in
a sample of size n from a lot of size N equals the number of ways to choose xdefectives
from all the defectives in the lot (NP) times numbers of ways to pick n x non-defectives
from all the non-defectives in the lot (N NP) over the number of ways to choose a sample
of n from the population N. For example, suppose a short quiz will contain twoques-
tions. Each question is from a different topic. There are four topics. You know twoout of
the four topics. There will be six different pairs of topics in the quiz. One pair you know
both questions, and another you dont know either question. The remaining pairs contain
6 PART FOUR Quality

FIGURE 10S-3 1.00


.95 = .05
An OC curve with the AQL,
.90
LTPD, producers risk , and
consumers risk
.80

.70

Probability of accepting the lot


.60

.50

.40

.30

.20

.10
= .10
.00
0 .05 .10 .15 .20 .25
AQL LTPD proportion defective
Indifferent in the lot
Good Bad

one question you know and one you dont. There are four of these because there are two
ways to get the question you know multiplied by two ways to get the question you dont
know. The probably that you will know both questions is 1/6, the probably that you will
not know either question is 1/6, and the probably that you will know exactly 1 of the two
questions is 4/6.
Because there are no hyper-geometric tables, if the lot size N is large relative to samplesize
n
n (so that 0.1) we can approximate the hyper-geometric probabilities by binomial
N
probabilities. The difference is that binomial probabilities assume that a sampled item is
put back in a lot after being inspected before the next item is selected from the lot. The
probability of observing up to and including c defectives in a sample (with replacement)
of size n from a lot with proportion defective P is given by cumulative binomial formula:
c

x !(n x )! P
n!
P( x c ) = x
(1 P) n x (10S-2)
x =0

where x=number of defectives in the sample.

Example S-2 Draw the OC curve for a situation in which a sample of n=10 items is drawn from a lot
containing N=2,000 items, and the lot is accepted if no more than c=1 defect isfound
and rejected if 2 or more defects are found in the sample.

Solution Because the sample size is small relative to the lot size (10/2000=.005<.1), it is reason-
able to use the binomial distribution to obtain the probabilities that a lot will be accepted
for various lot qualities. Although we can use formula 10S-2 to calculate the probabilities,
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 7

it is easier to use a binomial table. A portion of the cumulative binomial table found at the
end of this supplement is reproduced here to facilitate the discussion.

PROPORTION DEFECTIVE, p
n x .05 .10 .15 .20 .25 .30 .35 .40 .45 .50 .55 .60
10 0 .5987 .3487 .1969 .1074 .0563 .0282 .0135 .0060 .0025 .0010 .0003 .0001
c=1 1 .9139 .7361 .5443 .3758 .2440 .1493 .0860 .0464 .0233 .0107 .0045 .0017
2 .9885 .9298 .8202 .6778 .5256 .3828 .2616 .1673 .0996 .0547 .0274 .0123
3 .9990 .9872 .9500 .8791 .7759 .6496 .5138 .3823 .2660 .1719 .1020 .0548

To use the table, select various lot qualities (values of p listed across the top of the table),
beginning with .05, and find the probability that a lot with that percentage of defects would
be accepted (i.e., the probability of finding zero or one defect in this case). For p=.05,
the probability of one or no defects is .9139. For a lot with 10 percent defective (i.e., a
proportion defective of .10), the probability of one or fewer defects drops to .7361, and
for 15 percent defective, the probability of acceptance is .5443. In effect, you simply read
the probabilities across the row for c=1. By plotting these points (e.g., .05 and .9139,
.10 and .7361) on a graph and connecting them, you obtain the OC curve illustrated in
Figure 10S-4.
If the lot size N is large relative to sample size n (so that n/N0.1) and np<5, we can
approximate the binomial probabilities by the Poisson probabilities. The probability of
observing up to and including c defectives in a sample of size n from a lot with proportion
defectives P is given by cumulative Poisson formula:
e nP ( nP )x
c

P( x c ) =
x =0
x!
(10S-3)

1.00
FIGURE 10S-4
.90 .9139 QC curve for single sampling
plan n=10, c=1
.80
.7361
.70
Probability of acceptance

.60
.5443
.50

.40
.3758

.30
.2440
.20
.1493 .0233
.0107
.10 .0860 .0045
.0464 .0017

.00
0 .10 .20 .30 .40 .50 .60
Proportion defective in lot
8 PART FOUR Quality

where x=number of defectives in the sample. The Poisson approximation involves


treating the mean of the binomial distribution (i.e., np) as the mean of the Poisson (i.e.,
): =np. As with the binomial distribution, you select various values of lot quality, p,
and then determine the probability of accepting a lot (e.g., finding up to and including
c defects) by either using formula 10S-3 or referring to the cumulative Poisson table
at the end of the textbook. Values of p in increments of .01 are often used in this
regard. Example S-3 illustrates the use of the Poisson table in constructing an
OC curve.

Example S-3 Use the cumulative Poisson table at the end of the textbook to construct an OC curve for
the following single sampling plan:
N=5,000, n=80, c=2

Solution Note that 80/5000=.016<.1 and np<5 for most means below. Therefore, we can use
the Poisson distribution.

Selected Values of p =np Pac = [P (x2) from


Appendix B Table C]
.01 80(.01)=0.8 .953
.02 80(.02)=1.6 .783
.03 80(.03)=2.4 .570
.04 80(.04)=3.2 .380
.05 80(.05)=4.0 .238
.06 80(.06)=4.8 .143
.07 80(.07)=5.6 .082
.08 80(.08)=6.4 .046

The probabilities of acceptance, Pac, are drawn against lot proportion defectives P to
construct the OC curve:

Pac
1.00

N = 5,000
.80 n = 80
c = 2
.60

.40

.20

.00
.01 .02 .03 .04 .05 .06 .07 .08
Proportion defective

As mentioned above, as long as lot size N is large enough relative to n, it will not
have a significant effect on the OC curve. However, the parameters n (sample size) and
c (acceptance number) of a single sampling plan do affect the shape of the OC curve. To
illustrate this, given c=2, the OC curves for various n values are shown in the following
graph. As expected, the OC curve becomes steeper (will have more discriminatory power)
as n increases.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 9

Effect of sample size n on the shape of the OC curve

c=2
1.00
Probability of acceptance of the lot

.90
n = 80
.80
n = 50
.70
n = 125
.60
n = 200
.50
.40
.30
.20
.10
.00
.01 .02 .03 .04 .05 .06 .07 .08 .09 .1 .11 .12 .13
Proportion defective

On the other hand, for fixed sample size (n=80), the OC curve becomes steeper (will
have more discriminatory power) as acceptance number c decreases (see the graph below).
Perhaps it is easier to understand this by considering very large c values: in this case, most
lots will be accepted, i.e., no discrimination.

Effect of acceptance number c on the shape of the OC curve

n = 80
1.00
.90 c=0
Probability of acceptance

.80 c=1
.70
c=2
.60
c=3
.50
.40
.30
.20
.10
.00
.01 .02 .03 .04 .05 .06 .07 .08 .09 .1 .11 .12 .13
Proportion defective

LO3 Determining Single Sampling Plans


A sampling plan and its operating characteristic (OC) curve have a one-to-one relation-
ship. Therefore, determining the sample size n and acceptance number c is equivalent to
determining the sampling plans OC curve. One way to determine an OC curve is to specify
two points on it, for example, the two points (AQL, 1 ) and (LTPD, ). There are other
approaches to determine the sampling plan as well. Two of the important methods are
Dodge-Romig and MIL-STD-105E. Dodge and Romig in AT&T used the point (LTPD,
) and expected lot proportion defective P to minimize the expected total items inspected
(assuming that a rejected lot is completely inspected). Another approach, MIL-STD-105E,
created by the U.S. Armed Forces, uses the point (AQL, 1 ), lot size N, and a chosen
inspection level to determine the sampling plan. We will illustrate these methods below.

Using (AQL, 1 ) and (LTPD, )


Substituting (AQL, 1 ) in the cumulative binomial formula 10S-2:
c

x !(n x )! (A QL ) (1 A QL )
n! nx
1 = x
(10S-4)
x =0
10 PART FOUR Quality

Substituting (LTPD, ) in the cumulative binomial formula 10S-2:


c

x !(n x )! (LTPD ) (1 LTPD )


n! nx
= x
(10S-5)
x =0

We can try to solve equations (10S-4) and (10S-5) simultaneously to determine the two
unknown quantities n and c. However, this is not easy because these equations are non-
linear. Fortunately, Larson has determined a nomograph (a graphical calculating device)
that will provide the solution (see page 24 of this supplement).

Example S-4 Suppose AQL=.02 with =5% and LTPD=.08 with =10%. Use Larsons nomograph
at the end of this supplement to determine n and c.

Solution Larsons nomograph on page 24 can be used as follows: the vertical line on the left-hand
side is for lot percentage defectives such as AQL and LTPD. The vertical line on the right-
hand side is for the probability of lot acceptance such as (1 ) and . Connect AQL with
(1 ) and LTPD with with straight lines. The intersection of these two lines gives the
sample size n and acceptance number c. In this case, n=90 and c=3 (see page 24).

Using (LTPD,
) Lot size N, and Expected Lot Proportion
Defective P to Minimize Expected Total Items Inspected
(The Dodge-Romig Approach)
Let I=Expected total items inspected. If a lot is rejected, we assume that it is 100%
inspected. In this case, I=n+Probability of rejecting the lot(N n). Dodge and Romig
used the Poisson approximation to binomial to calculate the probability of accepting a lot.
Therefore, we have:
c
e nP ( nP ) x
I = n + 1
x =0
x! (N n)

(10S-6)

Let Pt=LTPD. We also use Poisson probabilities to ensure that the probability of accept-
ing a lot with proportion of defectives=Pt is . Dodge-Romig used =.10. Therefore,
we have:

e nP ( nPt )x
c

= .1 (10S-7)
x!
x =0

The Dodge-Romig approach is to determine n and c so that I, given in (10S-6), is mini-


mized subject to satisfying the equation (10S-7). Again, this problem is difficult to solve
because of the non-linear functions involved. Fortunately, Dodge and Romig have provided
some curves to do this: Figures DR1 to DR3 at the end of this chapter supplement.
Example S-5
Suppose that the expected lot proportion defective P = .02, LTPD=.08 with =10%, and
lot size N=200. Determine n and c using Dodge-Romig approach.

P .02
Solution First, we need to f ind the ratio = = .25 . Next, we need to f ind
LTPD .08
LTPDN=.08(200)=16. Then, in Figure DR1 on page 25 at the end of this supplement,
we draw a vertical line at x-axis value .25 and a horizontal line at y-axis value 16. They
intersect in an area associated with c=2. Next, in Figure DR2 on page 26, we draw a
vertical line at x-axis value 16 and see at what horizontal line it intersects the c=2 curve.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 11

They intersect at approximately 5. This is nLTPD. Therefore, n=5 / .08=62.5 or 63.


To determine the minimum expected total items inspected Imin, in Figure DR3 on page 27,
we draw a vertical line at x-axis value .25 and a horizontal line at y-axis value 16. They
intersect close to curve numbered 6. This is LTPDImin. Therefore, Imin=6 / .08=75.

Using MIL-STD-105E
The MIL-STD-105E uses the point (AQL, .95), the lot size N, the level of inspection
desired, and the type of inspection required (depending on the past history of the supplier)
to determine n and c.
The inspection level determines the relationship between the lot size and sample size.
There are three general inspection levels (I, II, and III) and four special inspection levels
(S-1 to S-4). Level II is designated as normal. Level I requires about half the amount of
inspection as level II, and is used when reduced sampling is required and a lower level of
discrimination can be tolerated. Level III requires about twice the amount of inspection as
level II, and is used when more discrimination is needed. The four special inspection levels
S-1,S-2,S-3,S-4 use very small samples, and should be used when small sample sizes are
necessary, and when large sampling risks can be tolerated.
There are three types of inspection (other than discontinue inspection): Normal inspec-
tion, reduced inspection, and tightened inspection. Reduced inspection results in smaller
n (less discrimination after a good history) and tightened inspection results in larger n
(more discrimination after a bad history). The discontinue inspection requires corrective
action by the supplier before any new lots are purchased. The switching rules between the
4 inspection types are displayed below:2

5 consecutive 5 consecutive
batches not batches not
rejected rejected

Reduced Normal Tightened Discontinue


inspection inspection inspection inspection
1 batch not 2 of 5 10 consecutive
accepted consecutive batches remain
batches rejected on tightened
inspection

The following Web site will provide n and c according to the MIL-STD-105E: http://
www.sqconline.com/mil-std-105.html. Alternatively, a copy of the MIL-STD-105E docu-
ment, which includes results tables, can be found online, e.g., at http://www.dianyuan.com/
bbs/u/39/1142140688.pdf.

Suppose lot size N is 2,000 and AQL is 1 percent. We would like to use normal type of Example S-6
inspection at general inspection level 2. Determine the sample size n and the acceptance
number c.

In www.sqconline.com/mil-std-105.html we choose the right range for N and check to see Solution
that default values for AQL, inspection level, and type of inspection are right. We click
on Submit.

2
http://www.sqconline.com/switching_rules_enter.php4
12 PART FOUR Quality

The values for n and c for single sampling appear: n=125 and c=3 (see the top left
box below). As an added bonus, the results for double sampling plan, which has a similar
http://www.sqconline.com
OC curve, and the OC curve are also provided:
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 13

LO4 Average Quality of Inspected Lots


anda Related Sampling Plan
It is fair to expect that acceptance sampling would reduce the proportion of defective items
accepted. Indeed, this is the case, provided that the rejected lots are not re-submitted for
acceptance sampling without improvement in quality. If they are, they will eventually be
accepted and the whole point of acceptance sampling is lost (i.e., the quality of accepted
items will be the same as the quality of incoming or rejected items). This fact can be seen by
considering the following example: suppose that the probability of a bad lot being accepted
is .1. If this lot is rejected and is returned to the supplier but the supplier sends it back to the
customer unchanged, and this process is repeated 40 times, the probability of acceptance
within 40 tries is .1+.9(.1)+.92(.1)+. . .+.939(.1)=.985. Therefore, the buyer has to either
require the supplier to perform 100% inspection of rejected lots or do it himself.
The average outgoing quality (AOQ) of the inspected lots is average percentage defec- average outgoing quality
tive of accepted lots assuming that rejected lots are 100 percent inspected and defective (AOQ) Average percentage
items in those lots are replaced with good items. AOQ can be calculated using the fol- defective of accepted lots
lowing formula: assuming that rejected lots
are 100 percent inspected and
N n defective items in those lots are
AOQ = Pac p (10S-8)
N replaced with good items.

where
ac =Probability of accepting the lot
P
p =Lot proportion defective
N =Lot size
n =Sample size
In practice, the last term in (10S-8) is often omitted because it is usually close to 1.0
and therefore has little effect on the resulting values. The formula then becomes

AOQ = Pac p (10S-9)

Construct the AOQ curve for N=500, n=10, and c=1 using formula 10S-9: Example S-7

Let values of p vary from .05 to .40 in steps of .05. You can read the probabilities of Solution
acceptance, Pac, from the binomial table at the end of this supplement.
AOQ = Pac p

P Pac AOQ
(Outgoing proportion defective)

Approximate AOQL = .082


.08
.05 .9139 .046
.10 .7361 .074 .06
AOQ

.15 .5443 .082


.04
.20 .3758 .075
.25 .2440 .061
.02
.30 .1493 .045
.35 .0860 .030 .00
.10 .20 .30 .40
.40 .0464 .019 Incoming proportion defective
14 PART FOUR Quality

Note that the average outgoing quality is best for either very good lots or very bad lots
(i.e., the outgoing proportion defective is least for lots with either very low or very high
incoming proportion defective). The reason very bad lots also will have high outgoing
quality is that they will likely be rejected and then 100% inspected and rectified.
A quantity of interest to the buyer is the maximum outgoing proportion defective, also
average outgoing quality called Average Outgoing Quality Limit (AOQL). AOQL is the worst quality of outgoing
limit (AOQL) The worst (i.e., accepted) items. In Example 7, The AOQL is approximately 8 percent, associated
quality of outgoing (i.e., with incoming lots of approximately 15% defective.
.4
accepted) items. An approximate value for AOQL can be obtained using the formula (1.25c + 1).3 For
.4 n
Example 7, AOQL (1.25 1 + 1) = .09, or 9%.
10

An AOQL based Sampling Plan


Dodge and Romig have provided a sampling plan involving AOQL similar to that given
by them in the previous section, but instead of using the (LTPD, ), a desired value for the
AOQL is assumed. Recall that P =expected lot proportion defective and I=Expected total
items inspected. As for the case of (LTPD, ), I is determined from equation (10S-6).
Let PL=the desired value of AOQL, i.e., a limit on the worst outgoing quality possible,
and P the proportion defective of an incoming lot. Using equation (10S-8), we get:

N n c e nP ( nP ) x
Max P
N x =0 x! = PL (10S-10)

Taking the derivative of the term inside the curly brackets in (10S-10) with respect to
P and setting it equal to 0, after some work we get

e nP ( nP ) x e nP ( nP )c + 1
c


x =0
x!
=
c! (10S-11)

Let Pm be the solution to equation (10S-11) for P, given values for n and c. Pm is the
x-axis value of lot proportion defective that is associated with PL. Substituting Pm for P
in(10S-10) and omitting the Max [because Pm maximizes the left term in formula 10S-10],
we get:

N n e nP ( nPm ) x
c

Pm = PL (10S-12)
N x!
x =0

The Dodge-Romig approach is to determine n and c so that I in (10S-6) is minimized


subject to determining Pm from (10S-11) and satisfying equation (10S-12) for AOQL. Again,
this problem is difficult to solve because of the non-linear functions involved. Fortunately,
Dodge and Romig have provided Figure DR4 (on page 28) and Table DR1 (on page 29)
at the end of this supplement that can be used to determine n and c.

Example S-8
Suppose that the expected lot proportion defective P = .02, AOQL=.05, and lot size
N=200. Determine n and c.
P .02
Solution First, we need to find the ratio = = .4. Next, we need to find P N = .02( 200 ) = 4.
A OQL .05
Then, in Figure DR4 on page 28 at the end of this supplement we draw a vertical line
at x-axis value .4 and a horizontal line at y-axis value 4. These lines intersect in an area
associated with c=1. Next, in Table DR1 on page 29, for c=1, we pick up x=1.62 and
y=.84. Now we can calculate n=yN/(AOQLN+y)=.84(200)/(.05200+.84)=15.5,
round to 16. AOQL occurs at Pm=x / n=1.62 / 16=.10.

3
J. M. Juran and F. M. Gryna, Quality Planning and Analysis, 3rd ed., 1993, New York: McGraw-Hill, p. 25.13.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 15

acceptable quality level (AQL), 5 lot tolerance percent defective (LTPD), 5 Key Terms
acceptance sampling, 2 operating characteristic (OC) curve, 3
average outgoing quality (AOQ), 12 producers risk, 5
average outgoing quality limit (AOQL), 13 sampling plans, 2
consumers risk, 5

Solved Problems
A process for manufacturing shock absorbers for light trucks produces .5 percent defectives. In-
spection cost per shock is $.40. Currently 100 percent inspection is performed, which is assumed Problem 1
to catch all the defectives. If a defective shock absorber were to be installed on a truck, it must
eventually be replaced at a cost of $120 per shock. Is 100 percent inspection justified?

The BEP=.4/120=.0033<.005, therefore 100 percent inspection is justified. However, if the Solution
percentage defective is not stable, acceptance sampling should be used.

Shipments of 300 boxes of glassware are received at a warehouse. Random samples of five Problem 2
boxes are checked, and the lot is rejected if more than one box contains a breakage. Construct
the OCcurve for this sampling plan.

Because n/N = 5/300=.0167<.1, binomial distribution can be used to obtain the probabilities Solution
of acceptance, Pac, for various lot percentage defectives. A portion of the cumulative binomial
table is shown below. Note that n=5 and c=1.

CUMULATIVE BINOMIAL PROBABILITIES


P=PROPORTION DEFECTIVE
n x .05 .10 .15 .20 .25 .30
5 0 .7738 .5905 .4437 .3277 .2373 .1681
c=1 1 .9974 .9185 .8352 .7373 .6328 .5282
2 .9988 .9914 .9734 .9421 .8965 .8369
3 1.0000 .9995 .9978 .9933 .9844 .9692
4 1.0000 1.0000 .9999 .9997 .9990 .9976
5 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

.35 .40 .45 .50 .55 .60 .65 .70 .75 .80
.1160 .0778 .0503 .0313 .0185 .0102 .0053 .0024 .0010 .0003
.4284 .3370 .2562 .1875 .1312 .0870 .0540 .0308 .0156 .0067
.7648 .6826 .5931 .5000 .4069 .3174 .2352 .1631 .1035 .0579
.9460 .9130 .8688 .8125 .7438 .6630 .5716 .4718 .3672 .2627
.9947 .9898 .9815 .9688 .9497 .9222 .8840 .8319 .7627 .6723
1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000

The table indicates that Pac=.9974 when lot quality is 5 percent defective, .9185 when lot
quality is 10 percent defective, .8352 when 15 percent, and so on. The resulting operating char-
acteristic (OC) curve is:
Pac
1.00

.80 N = 300
n = 5
c = 1
.60

.40

.20

0 .10 .20 .30 .40 .50 .60 .70 .80 .90 1.00
Lot proportion defective
16 PART FOUR Quality

Problem 3 Develop the AOQ curve for the previous problem using formula 10S-9.

AOQ = Pac p
Solution
(Values of probability of acceptance Pac, can be taken from the top portion of the binomial table
shown on the previous page)

Max = .158
.16

(Outgoing proportion defective)


.12
AOQ

.08

.04

0 .1 .2 .3 .4 .5 .6 .7 .8
Incoming proportion defective, p

p Pac AOQ p Pac AOQ


.05 .9974 .050 .45 .2562 .115
.10 .9185 .092 .50 .1875 .094
.15 .8352 .125 .55 .1312 .072
.20 .7373 .147 .60 .0870 .052
.25 .6328 .158 .65 .0540 .035
.30 .5258 .158 .70 .0380 .027
.35 .4284 .150 .75 .0156 .012
.40 .3370 .135 .80 .0067 .005

Discussion and 1. What is acceptance sampling and what is its purpose? (LO1)
Review Questions
2. How does acceptance sampling differ from process control using control charts? (LO1)
3. When should a buyer use sampling inspection vs. 100% inspection vs. no inspection?
(LO1)
4. What general factors govern the choice between single-sampling and multiple-sampling plans?
(LO1)
5. What is an operating characteristic curve, and how is it useful in acceptance sampling?
(LO2)
6. Briefly explain or define each of these terms. (LO2)
a. AQL.
b. LTPD.
c. Producers risk.
d. Consumers risk.
7. When can each of the following distributions be used in calculating the probability of accep-
tance of a lot? (LO2)
a. Hyper-geometric.
b. Binomial.
c. Poisson.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 17

8. When would you use each of the four methods given in this supplement for determining single
sampling plans? (LO3)
9. Explain or define each of the following: (LO4)
a. AOQ.
b. AOQL.

1. Visit http://www.itl.nist.gov/div898/handbook/pmc/section2/pmc22.htm, read about (a) Internet Exercises


sequential, and (b) skip lot sampling plans, and define or explain them. (LO1)
2. Visit http://www.astm.org/SNEWS/JF_2010/datapoints_jf10.html, and summarize why we
still need Acceptance Sampling (given the existence of control charts). (LO1)

1. An assembly operation for the trigger mechanism of a semiautomatic spray gun produces a Problems
small percentage of defective mechanisms. Management must decide whether to continue the
current practice of 100 percent inspection, perform acceptance sampling, or replace defective
mechanisms after final assembly when all guns are inspected. Replacement at final assembly
costs $30 each; inspection during trigger assembly costs $12 per hour for labour and overhead.
The inspection rate is one trigger per minute. (LO1)
a. Would 100 percent inspection during trigger assembly be justified if there are (1) 4 percent
defective? (2) 1 percent defective?
b. At what point would management prefer acceptance sampling?
2. Random samples of n=20 circuit breakers are tested for damage caused by shipment in
each lot of 4,000 received. Lots with more than one defective are pulled and subjected to 100
percent inspection. (LO2 & 4)
a. Construct the OC curve for this sampling plan.
b. Construct the AOQ curve for this plan using formula 10S-9, assuming defectives found
during 100 percent inspection are replaced with good parts. What is the approximate
AOQL?
3. Auditors use a technique called discovery sampling in which a random sample of items is
inspected. If any defects are found, the entire lot is subjected to 100 percent inspection. (LO2
& 4)
a. Draw an OC curve for the case where a sample of 15 credit accounts will be inspected out
of a total of 8,000 accounts.
b. Draw an OC curve for the case where 150 accounts out of 8,000 accounts will be examined.
(Hint: Use p=.001, .002, .003, . . .)
c. Draw the AOQ curve for the preceding case (part b), and determine the approximate
AOQL.
4. Random samples of lots of textbooks are inspected for defective books just prior to shipment
to the publishers warehouse. Each lot contains 3,000 books. (LO2)
a. On a single graph, construct OC curves for n=100 and (1) c=0, (2) c=1, and (3) c=2.
(Hint: Use p=.001, .002, .003, . . .)
b. On a single graph, construct OC curves for c=2 and (1) n=5, (2) n=20, and (3)
n=120.
5. A manufacturer receives shipments of several thousand parts from a supplier every week. The
manufacturer has the option of inspection before accepting the parts. Inspection cost is $1 per
unit. If parts are not inspected, defectives become apparent during a later assembly operation,
at which time replacement cost is $6.25 per unit. (LO1 & 2)
a. At what proportion defective would the manufacturer prefer acceptance sampling?
b. For the sample size n=15, what acceptance number c would result in probability of accep-
tance close to .95 for AQL=2%?
18 PART FOUR Quality

c. If the shipment actually contains 1 percent defective items and AQL =2 percent:
i. What is the correct decision?
ii. What is the probability that the lot would be accepted if acceptable number c=0?
iii. What is the probability that it would be rejected if c=0?
iv. Answer the questions in part c for a shipment that contains 3 percent defective
items.
6. Suppose there are two defective units in a sample. (LO2 & 4)
a. If the acceptance number is c=1, what decision should be made? What type of error is
possible?
b. If the acceptance number is c=3, what decision should be made? What type of error is
possible?
c. Use formula 10S-9 to determine the average outgoing quality for each of the following
percent defectives if c=1 and n =15.
i. 5 percent.
ii. 10 percent.
iii. 15 percent.
iv. 20 percent.
7. Suppose lot size N is 432 and acceptable quality level AQL is .65 percent. We would like to use
normal inspection at general inspection level II. Determine the sample size n and acceptance
number c using MIL-STD-105E.4 (LO3)
8. A manufacturer of colour TV picture tubes is wondering if its current sampling procedure can
be improved.5 Currently, the defects are classified into critical (C: e.g., contaminated anode),
major (B: e.g., bent pins) and minor (A: e.g., wrong label). The company has also grouped its
customers into three groups and has one plan for each group: (LO3)

Existing Sampling Plans

Plan1 Plan2 Plan3


Nonconformity class C&B A C&B A C&B A
Lot size (N) 48 48 48 48 48 48
Documented AQL 1.0 1.0 2.5 2.5 2.5 2.5
Documented inspection severity Normal Reduced Reduced
Sample size (n) 8 8 5 5 3 3
Acceptance number (c) 0 1 1 1 0 0

For each plan and defect category, determine the sample size n and acceptance number c
using MIL-STD-105E. Compare your results with the current plans above.
9. A single sampling plan uses sample size n=100 and the inspector accepts the lot if there are
2 or fewer defectives in the sample.6 You may use the Poisson approximation to answer the
following questions. (LO2 & 4)
a. What is the probability of accepting a lot with proportion defective p=.01?
b. What protection does the buyer have against accepting lots with proportion defective
p=.05?
c. What is the average outgoing quality AOQ for p=.01? Use formula 10S-9.
d. What is the average outgoing quality AOQ for p=.05? Use formula 10S-9.

4
 . F. Bauer, A Move from Attribute to Variables Acceptance Sampling in an ISO-Certified Manufacturing
E
Plant, M.S. thesis, California State University, Dominguez Hills, 2000.
5
E. Gamino, Improvement to the Acceptance Control System of a Manufacturer of Color Picture Tubes, M.S.
thesis, California State University, Dominguez Hills, 2005.
6
P. W. M. John, Statistical Methods in Engineering and Quality Assurance, New York: Wiley, 1990, p. 188.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 19

e. What is the average outgoing quality limit AOQL for this plan? You may use the approxi-
.4
mation (1.25c + 1) .
n
*10. A manufacturer wishes to sample a purchased component used in its assembly operation. 7
The company wishes to reject lots that are 5% defective. The components are received in
lots of 1,000 units, which average 2.2% defective. The supplier has agreed to perform 100%
inspection of all rejected lots. Find a sampling plan method that meets these conditions and
determine nand c. (LO3)
*11. You are the quality manager for a company receiving large quantities of material from a sup-
plier in lots of 1,000 units.8 The cost of inspecting the items is $.76 per unit. The cost of repair
if bad material is introduced into your product is $15.20 per unit. A single sampling plan of
75 units with acceptance number of 2 has been suggested by one of your quality inspectors.
In the past, lots submitted by this supplier have averaged 3.4% defective. (LO14)
a. Is acceptance sampling economically justified?
b. If you want to accept only lots of 4% defective or better, what do you think of the sampling
plan of the inspector?
c. Suppose that rejected lots are 100% inspected. If a supplier submits many 4% defective
lots, what will be the average outgoing quality of these lots? Use formula 10S-8.
12. a. Determine a sampling plan that will have AQL=1% with producer risk =.05 and
LTPD=5% with consumer risk =.10.9 (LO3)
b. Suppose a sample is taken according to the sampling plan derived in part a and two non-
conforming units are found. What action should be taken?
13. A housing development company buys heavy-duty nails in lots of 10,000 nails. A destruc-
tive test is performed to determine the strength of the nails.10 AQL is 1% and LTPD is 10%.
Asingle sample of n=100 and c=2 is used. Determine and . (LO3)
14. A manufacturer inspects all of its shipments to its customers prior to delivery using a double
sampling MIL-STD-105E standard plan, Inspection level II, normal type of inspection, and
an AQL of 1%.11 The lot is 500 units. (LO3)
a. Find the sampling plan used and explain it in words.
b. Upon delivery of the product, the customer also inspects the lot using a single sampling
MIL-STD-105E standard plan, AQL=1.5%, inspection level II, and normal type of inspec-
tion. Find the customers sampling plan.
c. If a lot that is 10% defective is produced, calculate the probability that it will pass the
manufacturers inspection.
d. Calculate the probability that a 10% defective lot will pass the customers inspection.
e. What is the probability that a 10% defective lot will pass both inspections?
15. Find the Dodge-Romig single sampling plan for AOQL=4%, lot size of 125, and average lot
percentage defective=1%.12 (LO4)
16. Binder clips are packaged 12 to a box and 12 boxes to a carton.13 You have received a lot
consisting of four cartons of binder clips. (LO3)
a. Use MIL-STD-105E to determine a single sampling plan to decide whether to accept or
reject the lot. Use Inspection Level II, normal type of inspection, and an AQL of 2.5%.
b. If in inspecting your sample you find three defective binder clips, what would you do?

7
J . M. Juran and F. M. Gryna, Quality Planning and Analysis, 3rd ed., 1993, New York: McGraw-Hill, p.486.
8
J. M. Juran and F. M. Gryna, Quality Planning and Analysis, 3rd ed, 1993, New York: McGraw-Hill, p. 487.
9
J. M. Juran and F. M. Gryna, Jurans Quality Control Handbook, 4th ed., 1988, New York: McGraw-Hill,
pp.25.2625.27.
10
S. Nahmias, Production and Operations Analysis, 3rd ed., 1997, Chicago: Irwin.
11
E. I. Grant and R. S. Leavenworth, Statistical Quality Control, 4th ed., 1972, New York: McGraw-Hill,
p.445.
12
E. G. Schilling, Acceptance Sampling in Quality Control, 1982, New York: Marcel Dekker, p. 398.
13
http://www.shsu.edu/~mgt_ves/mgt481/lesson9/lesson9.htm.
20 PART FOUR Quality

Mini-Case

the fridge off, insert the nogs in the pallet, and put the fridge
CRYSTAL S.A. back on the pallet. The pallets are bought in lot sizes of 800. The
percentage of defective pallets in the past has ranged between
CRYSTAL S.A. is a Greek commercial fridge manufacturer.14
5 and 8%.
During packaging, CRYSTAL uses a wooden platform (pallet)
under the fridge. The pallets use some wooden pegs (nogs). Questions
However, some nogs could be missing. The cost of inspecting a. Should the company perform 100% inspection, acceptance
a pallet is $.119. The average cost of putting nogs in pallets sampling, or no inspection?
missing them is $.535 per pallet. If the defect is not identified b. If acceptance sampling is best, what sampling plan should
before the pallet is used in packaging, it will cost $2.3 to take the company use? Justify your choice.

Mini-Case

Group 1 (17 drugs) were each issued less than 400 times a
Gustave Roussy Institute year. For this group it was decided to continue 100% inspection
because the small number of issues during a semester may not
Acceptance sampling is used in the Department of Clinical
be enough to be able to control their quality. Group 3 consisted
Pharmacy of Gustave Roussy Institute in Villejuif, France.15
of over 20% of production but the percentage defective of each
Most of the over 10,000 custom-made units of 40 or so che-
(3 to 9%) was considered high enough to continue 100% inspec-
motherapy drugs produced during a semester (half year) for
tion. Group 2 (6 drugs) were considered to be candidates for
5semesters starting in 2001 were tested (at the cost of $1.50
acceptance sampling because their average percentage defective
each) to make sure that their content, dosage, and concentration
of 2.2% was stable and acceptable. For this group, AQL is 2.2%
were acceptable (i.e., within 10% of the specification). Causes
with =.05 and LTPD is 5% with =.05. Being cautious, the
of defects were investigated and fixed so that percentage defec-
transition from 100% inspection into acceptance sampling is
tive was reduced from an average of 8.9% to 2.2%. Based on
planned in stages. MIL-STD-105E tables are being used to deter-
50,000 test results, the test administrators classified the drugs
mine the sampling plans. The plans for one of the drugs are:
into three classes:
n c AQL LTPD
Category of Drugs Drugs Concerned
Fluorouracil
Group 1 (17 drugs) <400 Irinotecan, oxaliplatin,
preparations a year gemcitabine, daunorubincin, Observed 2484 67 2.22 3.16
idarubicin, fludarabine, Proposed sampling plans 2000 55 2.22 3.28
melphalan, mitoxantrone, 1500 43 2.25 3.50
vindesine, vinorelbine, 1000 30 2.25 3.81
vinblastine, vincristine,
500 16 2.17 4.46
dacarbazine, thiothepa,
carboplatin, paclitaxel, docetaxel.
a. Calculate and comment on the assumed cost of a defective
Group 2 (6 drugs) Acceptance Fluorouracil, ifosfamide, drug in Group 2 to a patient.
sampling plan cisplatin, epirubicin, doxorubicin,
b. Determine the sample plan that should be used for Fluorou-
cyclophosphamide
racil and compare your plan with the proposed plans.
Group 3 (3 drugs) At risk to be Methotrexate, etoposide,
outside the specification cyarabine

14
 . Nikolaidis and G. Nenes, Economic Evaluation of ISO 2859 Acceptance Sampling Plans Used with
Y
Rectifying Inspection of Rejected Lots, Quality Engineering 21 (1), 2009, pp. 1023.
15
I. Borget et al, Application of an Acceptance Sampling Plan for Post-production Quality Control of
Chemotherapeutic Batches in a Hospital Pharmacy, European Journal of Pharmaceutics and Biopharmaceutics,
64 (2006), pp. 9298.
c
n x nx
P( x c ) = x P (1 P) 0 1 2 3 4 x
x =0 c=1

Cumulative binomial probabilities


P
n x .05 .10 .15 .20 .25 .30 .35 .40 .45 .50 .55 .60 .65 .70 .75 .80 .85 .90

1. . . . 0 .9500 .9000 .8500 .8000 .7500 .7000 .6500 .6000 .5500 .5000 .4500 .4000 .3500 .3000 .2500 .2000 .1500 .1000
1 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
2. . . . 0 .9025 .8100 .7225 .6400 .5625 .4900 .4225 .3600 .3025 .2500 .2025 .1600 .1225 .0900 .0625 .0400 .0225 .0100
1 .9975 .9900 .9775 .9600 .9375 .9100 .8775 .8400 .7975 .7500 .6975 .6400 .5775 .5100 .4375 .3600 .2775 .1900
2 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
3. . . . 0 .8574 .7290 .6141 .5120 .4219 .3430 .2746 .2160 .1664 .1250 .0911 .0640 .0429 .0270 .0156 .0080 .0034 .0010
1 .9928 .9720 .9393 .8960 .8438 .7840 .7183 .6480 .5748 .5000 .4253 .3520 .2818 .2160 .1563 .1040 .0608 .0280
2 .9999 .9990 .9966 .9920 .9844 .9730 .9571 .9360 .9089 .8750 .8336 .7840 .7254 .6570 .5781 .4880 .3859 .2710
3 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
4. . . . 0 .8145 .6561 .5220 .4096 .3164 .2401 .1785 .1296 .0915 .0625 .0410 .0256 .0150 .0081 .0039 .0016 .0005 .0001
1 .9860 .9477 .8905 .8192 .7383 .6517 .5630 .4752 .3910 .3125 .2415 .1792 .1265 .0837 .0508 .0272 .0120 .0037
2 .9995 .9963 .9880 .9728 .9492 .9163 .8735 .8208 .7585 .6875 .6090 .5248 .4370 .3483 .2617 .1808 .1095 .0523
3 1.0000 .9999 .9995 .9984 .9961 .9919 .9850 .9744 .9590 .9375 .9085 .8704 .8215 .7599 .6836 .5904 .4780 .3439
4 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
5. . . . 0 .7738 .5905 .4437 .3277 .2373 .1681 .1160 .0778 .0503 .0313 .0185 .0102 .0053 .0024 .0010 .0003 .0001 .0000
1 .9974 .9185 .8352 .7373 .6328 .5282 .4284 .3370 .2562 .1875 .1312 .0870 .0540 .0308 .0156 .0067 .0022 .0005
2 .9988 .9914 .9734 .9421 .8965 .8369 .7648 .6826 .5931 .5000 .4069 .3174 .2352 .1631 .1035 .0579 .0266 .0086
3 1.0000 .9995 .9978 .9933 .9844 .9692 .9460 .9130 .8688 .8125 .7438 .6630 .5716 .4718 .3672 .2627 .1648 .0815
4 1.0000 1.0000 .9999 .9997 .9990 .9976 .9947 .9898 .9815 .9688 .9497 .9222 .8840 .8319 .7627 .6723 .5563 .4095
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling

5 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
6. . . . 0 .7351 .5314 .3771 .2621 .1780 .1176 .0754 .0467 .0277 .0156 .0083 .0041 .0018 .0007 .0002 .0001 .0000 .0000
1 .9672 .8857 .7765 .6554 .5339 .4202 .3191 .2333 .1636 .1094 .0692 .0410 .0223 .0109 .0046 .0016 .0004 .0001
2 .9978 .9842 .9527 .9011 .8306 .7443 .6471 .5443 .4415 .3438 .2553 .1792 .1174 .0705 .0376 .0170 .0059 .0013
3 .9999 .9987 .9941 .9830 .9624 .9295 .8826 .8208 .7447 .6563 .5585 .4557 .3529 .2557 .1694 .0989 .0473 .0159
4 1.0000 .9999 .9996 .9984 .9954 .9891 .9777 .9590 .9308 .8906 .8364 .7667 .6809 .5798 .4661 .3446 .2235 .1143
5 1.0000 1.0000 1.0000 .9999 .9998 .9993 .9982 .9959 .9917 .9844 .9723 .9533 .9246 .8824 .8220 .7379 .6229 .4686
6 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
7. . . . 0 .6983 .4783 .3206 .2097 .1335 .0824 .0490 .0280 .0152 .0078 .0037 .0016 .0006 .0002 .0001 .0000 .0000 .0000
1 .9556 .8503 .7166 .5767 .4449 .3294 .2338 .1586 .1024 .0625 .0357 .0188 .0090 .0038 .0013 .0004 .0001 .0000
21
22
P
n x .05 .10 .15 .20 .25 .30 .35 .40 .45 .50 .55 .60 .65 .70 .75 .80 .85 .90

2 .9962 .9743 .9262 .8520 .7564 .6471 .5323 .4199 .3164 .2266 .1529 .0963 .0556 .0288 .0129 .0047 .0012 .0002
3 .9998 .9973 .9879 .9667 .9294 .8740 .8002 .7102 .6083 .5000 .3917 .2898 .1998 .1260 .0706 .0333 .0121 .0027
4 1.0000 .9998 .9988 .9953 .9871 .9712 .9444 .9037 .8471 .7734 .6836 .5801 .4677 .3529 .2436 .1480 .0738 .0257
5 1.0000 1.0000 .9999 .9996 .9987 .9962 .9910 .9812 .9643 .9375 .8976 .8414 .7662 .6706 .5551 .4233 .2834 .1497
6 1.0000 1.0000 1.0000 1.0000 .9999 .9998 .9994 .9984 .9963 .9922 .9848 .9720 .9510 .9176 .8665 .7903 .6794 .5217
7 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
8. . . . 0 .6634 .4305 .2725 .1678 .1001 .0576 .0319 .0168 .0084 .0039 .0017 .0007 .0002 .0001 .0000 .0000 .0000 .0000
1 .9428 .8131 .6572 .5033 .3671 .2553 .1691 .1064 .0632 .0352 .0181 .0085 .0036 .0013 .0004 .0001 .0000 .0000
2 .9942 .9619 .8948 .7969 .6785 .5518 .4278 .3154 .2201 .1445 .0885 .0498 .0253 .0113 .0042 .0012 .0002 .0000
3 .9996 .9950 .9786 .9437 .8862 .8059 .7064 .5941 .4470 .3633 .2604 .1737 .1061 .0580 .0273 .0104 .0029 .0004
4 1.0000 .9996 .9971 .9896 .9727 .9420 .8939 .8263 .7396 .6367 .5230 .4059 .2936 .1941 .1138 .0563 .0214 .0050
5 1.0000 1.0000 .9998 .9988 .9958 .9887 .9747 .9502 .9115 .8555 .7799 .6848 .5722 .4482 .3215 .2031 .1052 .0381
6 1.0000 1.0000 1.0000 .9999 .9996 .9987 .9964 .9915 .9819 .9648 .9368 .8936 .8309 .7447 .6329 .4967 .3428 .1869
7 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9998 .9993 .9983 .9961 .9916 .9832 .9681 .9424 .8999 .8322 .7275 .5695
8 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
PART FOUR Quality

9. . . . 0 .6302 .3874 .2316 .1342 .0751 .0404 .0207 .0101 .0046 .0020 .0008 .0003 .0001 .0000 .0000 .0000 .0000 .0000
1 .9288 .7748 .5995 .4362 .3003 .1960 .1211 .0705 .0385 .0195 .0091 .0038 .0014 .0004 .0001 .0000 .0000 .0000
2 .9916 .9470 .8591 .7382 .6007 .4628 .3373 .2318 .1495 .0898 .0498 .0250 .0112 .0043 .0013 .0003 .0000 .0000
3 .9994 .9917 .9661 .9144 .8343 .7297 .6089 .4826 .3614 .2539 .1658 .0994 .0536 .0253 .0100 .0031 .0006 .0001
4 1.0000 .9991 .9944 .9804 .9511 .9012 .8283 .7334 .6214 .5000 .3786 .2666 .1717 .0988 .0489 .0196 .0056 .0009
5 1.0000 .9999 .9994 .9969 .9900 .9747 .9464 .9006 .8342 .7461 .6386 .5174 .3911 .2703 .1657 .0856 .0339 .0083
6 1.0000 1.0000 1.0000 .9997 .9987 .9957 .9888 .9750 .9502 .9102 .8505 .7682 .6627 .5372 .3993 .2618 .1409 .0530
7 1.0000 1.0000 1.0000 1.0000 .9999 .9996 .9986 .9962 .9909 .9805 .9615 .9295 .8789 .8040 .6997 .5638 .4005 .2252
8 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9997 .9992 .9980 .9954 .9899 .9793 .9596 .9249 .8658 .7684 .6126
9 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
10. . . . 0 .5987 .3487 .1969 .1074 .0563 .0282 .0135 .0060 .0025 .0010 .0003 .0001 .0000 .0000 .0000 .0000 .0000 .0000
1 .9139 .7361 .5443 .3758 .2440 .1493 .0860 .0464 .0233 .0107 .0045 .0017 .0005 .0001 .0000 .0000 .0000 .0000
2 .9885 .9298 .8202 .6778 .5256 .3828 .2616 .1673 .0996 .0547 .0274 .0123 .0048 .0016 .0004 .0001 .0000 .0000
3 .9990 .9872 .9500 .8791 .7759 .6496 .5138 .3823 .2660 .1719 .1020 .0548 .0260 .0106 .0035 .0009 .0001 .0000
4 .9999 .9984 .9901 .9672 .9219 .8497 .7515 .6331 .5044 .3770 .2616 .1662 .0949 .0473 .0197 .0064 .0014 .0001
5 1.0000 .9999 .9986 .9936 .9803 .9527 .9051 .8338 .7384 .6230 .4956 .3669 .2485 .1503 .0781 .0328 .0099 .0016
6 1.0000 1.0000 .9999 .9991 .9965 .9894 .9740 .9452 .8980 .8281 .7340 .6177 .4862 .3504 .2241 .1209 .0500 .0128
7 1.0000 1.0000 1.0000 .9999 .9996 .9984 .9952 .9877 .9726 .9453 .9004 .8327 .7384 .6172 .4744 .3222 .1798 .0702
8 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9995 .9983 .9955 .9893 .9767 .9536 .9140 .8507 .7560 .6242 .4557 .2639
9 1,0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9997 .9990 .9975 .9940 .9865 .9718 .9437 .8926 .8031 .6513
10 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
15. . . . 0 .4633 .2059 .0874 .0352 .0134 .0047 .0016 .0005 .0001 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000
P
n x .05 .10 .15 .20 .25 .30 .35 .40 .45 .50 .55 .60 .65 .70 .75 .80 .85 .90

1 .8290 .5490 .3186 .1671 .0802 .0353 .0142 .0052 .0017 .0005 .0001 .0000 .0000 .0000 .0000 .0000 .0000 .0000
2 .9638 .8159 .6042 .3980 .2361 .1268 .0617 .0271 .0107 .0037 .0011 .0003 .0001 .0000 .0000 .0000 .0000 .0000
3 .9945 .9444 .8227 .6482 .4613 .2969 .1727 .0905 .0424 .0176 .0063 .0019 .0005 .0001 .0000 .0000 .0000 .0000
4 .9994 .9873 .9383 .8358 .6865 .5155 .3519 .2173 .1204 .0592 .0255 .0093 .0028 .0007 .0001 .0000 .0000 .0000
5 .9999 .9978 .9832 .9389 .8516 .7216 .5643 .4032 .2608 .1509 .0769 .0338 .0124 .0037 .0008 .0001 .0000 .0000
6 1.0000 .9997 .9964 .9819 .9434 .8689 .7548 .6098 .4522 .3036 .1818 .0950 .0422 .0152 .0042 .0008 .0001 .0000
7 1.0000 1.0000 .9994 .9958 .9827 .9500 .8868 .7869 .6535 .5000 .3465 .2131 .1132 .0500 .0173 .0042 .0006 .0000
8 1.0000 1.0000 .9999 .9992 .9958 .9848 .9578 .9050 .8182 .6964 .5478 .3902 .2452 .1311 .0566 .0181 .0036 .0003
9 1.0000 1.0000 1.0000 .9999 .9992 .9963 .9876 .9662 .9231 .8491 .7392 .5968 .4357 .2784 .1484 .0611 .0168 .0022
10 1.0000 1.0000 1.0000 1.0000 .9999 .9993 .9972 .9907 .9745 .9408 .8796 .7827 .6481 .4845 .3135 .1642 .0617 .0127
11 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9995 .9981 .9937 .9824 .9576 .9095 .8273 .7031 .5387 .3518 .1773 .0556
12 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9997 .9989 .9963 .9893 .9729 .9383 .8732 .7639 .6020 .3958 .1841
13 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9995 .9983 .9948 .9858 .9647 .9198 .8329 .6814 .4510
14 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9995 .9984 .9953 .9866 .9648 .9126 .7941
15 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
20. . . . 0 .3585 .1216 .0388 .0115 .0032 .0008 .0002 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000
1 .7358 .3917 .1756 .0692 .0243 .0076 .0021 .0005 .0001 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000
2 .9245 .6769 .4049 .2061 .0913 .0355 .0121 .0036 .0009 .0002 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000
3 .9841 .8670 .6477 .4114 .2252 .1071 .0444 .0160 .0049 .0013 .0003 .0000 .0000 .0000 .0000 .0000 .0000 .0000
4 .9974 .9568 .8298 .6296 .4148 .2375 .1182 .0510 .0189 .0059 .0015 .0003 .0000 .0000 .0000 .0000 .0000 .0000
5 .9997 .9887 .9327 .8042 .6172 .4164 .2454 .1256 .0553 .0207 .0064 .0016 .0003 .0000 .0000 .0000 .0000 .0000
6 1.0000 .9976 .9781 .9133 .7858 .6080 .4166 .2500 .1299 .0577 .0214 .0065 .0015 .0003 .0000 .0000 .0000 .0000
7 1.0000 .9996 .9941 .9679 .8982 .7723 .6010 .4159 .2520 .1316 .0580 .0210 .0060 .0013 .0002 .0000 .0000 .0000
8 1.0000 .9999 .9987 .9900 .9591 .8867 .7624 .5956 .4143 .2517 .1308 .0565 .0196 .0051 .0009 .0001 .0000 .0000
9 1.0000 1.0000 .9998 .9974 .9861 .9520 .8782 .7553 .5914 .4119 .2493 .1275 .0532 .0171 .0039 .0006 .0000 .0000
10 1.0000 1.0000 1.0000 .9994 .9961 .9829 .9468 .8725 .7507 .5881 .4086 .2447 .1218 .0480 .0139 .0026 .0002 .0000
11 1.0000 1.0000 1.0000 .9999 .9991 .9949 .9804 .9435 .8692 .7483 .5857 .4044 .2376 .1133 .0409 .0100 .0013 .0001
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling

12 1.0000 1.0000 1.0000 1.0000 .9998 .9987 .9940 .9790 .9420 .8684 .7480 .5841 .3990 .2277 .1018 .0321 .0059 .0004
13 1.0000 1.0000 1.0000 1.0000 1.0000 .9997 .9985 .9935 .9786 .9423 .8701 .7500 .5834 .3920 .2142 .0867 .0219 .0024
14 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9997 .9984 .9936 .9793 .9447 .8744 .7546 .5836 .3828 .1958 .0673 .0113
15 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9997 .9985 .9941 .9811 .9490 .8818 .7625 .5852 .3704 .1702 .0432
16 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9997 .9987 .9951 .9840 .9556 .8929 .7748 .5886 .3523 .1330
17 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9998 .9991 .9964 .9879 .9645 .9087 .7939 .5951 .3231
18 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9999 .9995 .9979 .9924 .9757 .9308 .8244 .6083
19 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 .9998 .9992 .9968 .9885 .9612 .8784
20 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
23
24 PART FOUR Quality

Larsons Binomial Nomograph

n
0.1
.01 N
0 1000
c n!
700 P = P{m c} = pm(1 p) n m
.02 5 m = 0 m!(n m)!
500
400 p

Nu
ces (c)
.03 10 300 P

mb
.02
.04 200 .10

er
20

ccurren
.001

of
.05 140 .08

tria
40 100 .95
.06 .005

ls
Probability of occurrence in a single trial (p)

er of o
70

or
.07 50 .01

Probability of c or fewer occurrences in n trials (P )


sa
.08 50
.02

m
70 40 n = 90

ple
.09 Numb
.10 30 c=3 .05

siz
100
20

e(
.10

n)
140 10 .20
.15
.30
5 .40
.50
.20 200 2 .60
0 .70
.25 .80

1 .90
.30
.95
2

(c)
.35 3 .98

ces
4 .99

ren
5 .995
.40

cur
7
Oc .999
.45 9
30 20 10
70 50 40
.50 140 100
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 25

1000
800

600 31
500 26 To

400 21 To 35
To 30
300 19 25
17 20
15 18
200 13
16
11 14
10 12
100 9
8

er
[PtN = (Tolerance fraction defective) (Lot size)]

mb
80 7
nu
ce
Tolerance number of defectives

6
an

60
pt
ce

50 5
Ac

40 4

30 3

2
20

10 0

6
5

3
Single sampling
consumers risk, 0.10
2

1
0 0.1 0.2 0.3 0.4 0.5 0.6
Ratio of process average to tolerance
[P/Pt = (Process average fraction defective)/(Tolerance fraction defective)]

Figure DR1 Dodge-Romig Curves for Finding the Acceptance Number

Source: H. F. Dodge and H. G. Romig, Sampling Inspection Tables, 2nd ed, 1959,
NewYork: John Wiley, p. 14.
26 PART FOUR Quality

24

22 Simple sampling
[Ptn = (Tolerance fraction defective) (Sample size)] consumers risk, O.IO 19
20
18
17
18
16
15
Tolerance times sample size

16
14

ber
13

num
14
12

ce
11
12

tan
10

cep
9
10

Ac
8
7
8
6
5
6
4
3
4
2
1
2 0
0
1 2 3 4 5 7 10 20 30 50 70 100 200 300 500 1000
Tolerance number of defectives
[PtN = (Tolerance fraction defective) (Lot size)]

Figure DR2 Dodge-Romig Curves for Finding the Size of Sample

Source: H. F. Dodge and H. G. Romig, Sampling Inspection Tables, 2nd ed, 1959, New York:
JohnWiley, p. 15.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 27

1000

800

600
500

400

300

200 40

[PtImin = (Tolerance fraction defective) (Minimum average number inspected per lot)]
35

100 30
[PtN = (Tolerance fraction defective) (Lot size)]

80

Tolerance times minimum average inspected


25
60
Tolerance number of defectives

50
20
40

30

15
13
20
11
10
9
8
10 7

8 6

5
6
5 4

4
3

2
2
1.5
Single sampling
consumers risk, 0.10

1
0 0.1 0.2 0.3 0.4 0.5 0.6
Ratio of process average to tolerance
[P/Pt = (Process average fraction defective)/(Tolerance fraction defective)]

Figure DR3 Dodge-Romig Curves for Finding the Minimum Amount of Inspection per Lot

Source: H. F. Dodge and H. G. Romig, Sampling Inspection Tables, 2nd ed, 1959,
New York: John Wiley, p. 16.
28 PART FOUR Quality

10 000
8000
6000
5000
4000
3000

2000

C
21
1000 to
800 19 25
15 20
600 1617
M = pN (process average fraction defective times lot size)

500 14 18
400 12
300 10
9
8
200
7

6
100 5
80
4
60
50
40
3
30

20
2

10
8
6
5
1
4
3

C=0
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5
p
k = (ratio of process average fraction defective to AOQL)
pL

Figure DR4 Dodge-Romig Curves for Determining the Acceptance Number c; AOQL
Protection
Source: H. F. Dodge and H. G. Romig, Sampling Inspection Tables, 1944, New York:
John Wiley, p. 50.
SUPPLEMENT TO CHAPTER 10 Acceptance Sampling 29

Given c x y Given c x y Given c x y Given c x y

0 1.00 0.3679 10 8.05 6.528 20 15.92 13.89 30 24.11 21.70


1 1.62 0.8400 11 8.82 7.233 21 16.73 14.66 31 24.95 22.50
2 2.27 1.371 12 9.59 7.948 22 17.54 15.43 32 25.78 23.30
3 2.95 1.942 13 10.37 8.670 23 18.35 16.20 33 26.62 24.10
4 3.64 2.544 14 11.15 9.398 24 19.17 16.98 34 27.45 24.90
5 4.35 3.168 15 11.93 10.13 25 19.99 17.76 35 28.29 25.71
6 5.07 3.812 16 12.72 10.88 26 20.81 18.54 36 29.13 26.52
7 5.80 4.472 17 13.52 11.62 27 21.63 19.33 37 29.97 27.33
8 6.55 5.146 18 14.31 12.37 28 22.46 20.12 38 30.82 28.14
9 7.30 5.831 19 15.12 13.13 29 23.29 20.91 39 31.66 28.96
10 8.05 6.528 20 15.92 13.89 30 24.11 21.70 40 32.51 29.77

Table DR1 Dodge-Romig values of x and y for Determining AOQL

Source: H. F. Dodge and H. G. Romig, Sampling Inspection Tables, 1944, New York: John Wiley, p. 49.

You might also like