Transcript
Copyright Tapan P Bagchi 1
Acceptance Sampling
Accepting or rejecting a lot (of parts, components, etc.) based
on the inspection of a sampledrawn from it
Copyright Tapan P Bagchi 2
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
A “Lot” of goods
Copyright Tapan P Bagchi 3
What is acceptance sampling?
Lot Acceptance SamplingA 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 acceptedA procedure for sentencing incoming batchesor lots of items without doing 100% inspectionThe most widely used sampling plans are given by Military Standard (MIL-STD-105E)
Copyright Tapan P Bagchi 4
What is acceptance sampling?
PurposesDetermine the quality level of an incoming shipment or at the end of productionJudge whether 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!
Copyright Tapan P Bagchi 5
Acceptance Sampling
Another area of quality control and improvementClosely connected with inspection and testing of productInspection can occur at many points in a process
Acceptance Sampling: the inspection and classification of a sample of nits selected at random from a larger batch or lot and ultimate decision about disposition of the lot –
Lot Disposition or Lot Sentencing
Acceptance Sampling: the inspection and classification of a sample of nits selected at random from a larger batch or lot and ultimate decision about disposition of the lot –
Lot Disposition or Lot Sentencing
After productionAfter production
Two common points of inspectionWhen parts are receivedWhen parts are received
Copyright Tapan P Bagchi 6
Sampling PlansAcceptance sampling: Form of inspection applied to lots or batches of items before or after a process, to judge conformance with predetermined standards
Sampling plans: Plans that specify lot size, sample size, number of samples, and acceptance/rejection criteria
Single-sampling
Double-sampling
Multiple-sampling and Sequential sampling
Copyright Tapan P Bagchi 7
Types of sampling plans
Sampling by attributes vs. sampling by variablesIncoming vs. outgoing inspectionRectifying vs. non-rectifying inspection
What is done with nonconforming items found during inspectionDefectives may be replaced by good items
Single, double, multiple and sequential plans
Copyright Tapan P Bagchi 8
How acceptance sampling works
Attributes (“go no-go” inspection)
Defectives-product acceptability across range
Defects-number of defects per unit
Variable (continuous measurement)
Usually measured by mean and standard deviation
Copyright Tapan P Bagchi 9
Why use acceptance sampling?
Can do either 100% inspection, or inspect a sample of a few items taken from the lotComplete inspection
Inspecting each item produced to see if each item meets the level desiredUsed when defective items would be very detrimental in some way
Copyright Tapan P Bagchi 10
Sampling TermsAcceptance quality level (AQL): the percentage of defects at which consumers are willing to accept lots as “good”
Lot tolerance percent defective (LTPD or RQL): the upper limit on the percentage of defects that a consumer is willing to accept
Consumer’s risk: the probability that a lot contained defectives exceeding the LTPD will be accepted
Producer’s risk: the probability that a lot containing the acceptable quality level will be rejected
Copyright Tapan P Bagchi 11
Why not 100% inspection?
Problems with 100% inspection
Generally very expensive
Can’t use when product must be destroyed to test it
Handling by inspectors can induce defects
Inspection may be very tedious so defective items may slip through even 100% inspection
Copyright Tapan P Bagchi 12
A Lot-by-Lot Sampling Plan
N(Lot) n
Count Number
Conforming
Accept orReject 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
Copyright Tapan P Bagchi 13
Quality Definitions
Acceptance quality level (AQL)The smallest percentage of defectives that will make the lot definitely acceptable. A quality level that is the base line requirement of the customer
RQL or Lot tolerance percent defective (LTPD)Quality level that is unacceptable to the customer
Copyright Tapan P Bagchi 14
How acceptance sampling works
RememberYou are not measuring the quality of the lot, but, you are to sentence the lot to either reject or accept it
Sampling involves risks:Good product may be rejectedBad product may be acceptedBecause we inspect only a sample, not the whole lot!
Copyright Tapan P Bagchi 15
Acceptance Sampling
Copyright Tapan P Bagchi 16
Three approaches for sentencing a lot
• when 100% inspection is too costly• Trusted supplier with potential risk
1. Single sampling1. Single sampling
2. Rectified sampling 2. Rectified sampling
3. Double/multiple sampling3. Double/multiple sampling
• Inspect every item in the lot, then remove the defective units by discarding, reworking, or returning to suppliers• For critical process. Defective input may result high failure cost• From new supplier
• For low cost or low impact material on the subsequent process• From trusted supplier
Copyright Tapan P Bagchi 17
Types of Sampling Plan
Sampling Plan for VariablesSampling Plan for VariablesSampling Plan for AttributesSampling Plan for Attributes
1. Single-sampling plan2. Double-sampling plan3. Multiple-sampling plan4. Sequential sampling plan
Standard sample size
Smaller sample size on average
Copyright Tapan P Bagchi 18
Advantages and Disadvantages of Sampling
SSP’s AdvantagesLess expensive because of less inspection
Works with single sample Protects both producer and consumer
Rejection on entire lot motivates quality improvement for suppliers
DisadvantagesRisk of accepting a lot of poor quality Risk of rejecting a lot of acceptable qualityRequires planning and documentationRequire extensive study on customer’s requirement
Copyright Tapan P Bagchi 19
Lot Formation
1. Lots should be such that …produced on the same machines, by same operators, from common raw materials, at approximately the same timeperiod
2. Larger lots are better than smaller lotsThese are more representative of overall quality
3. Lots should be conformable to the material handling systems and personnel
Copyright Tapan P Bagchi 20
Random SamplingInspected units should be selected at randomInspected units should represent all items in the lotPotential bad stories:
Sampled from the front, or top of pile Did not randomizeDid not stratify the lot and did not sample from each stratum
Copyright Tapan P Bagchi 21
The Single Sampling Plan
The most common and easiest plan to use but not most efficient in terms of average number of samples neededOne sample drawn from the lot and 100% inspectedSingle sampling plan
N = lot sizen = sample size (randomized)c = acceptance numberd = number of defective items in sample
Rule: If d ≤ c, accept lot; else reject the lot
Copyright Tapan P Bagchi 22
d ≤ c ?
Reject lot
YesAccept lot
Do 100% inspection
Return lot to supplier
Inspect all items in the sample
Defectives found = d
No
Take a randomized sample of size n
from the lot N
The Single Sampling procedure
Copyright Tapan P Bagchi 23
Single-Sampling Plans for Attributes
A lot of size N has been submitted for inspectionSample size nAcceptance number of defective cLot sentencing is based on one sample of size n
Example:N = 10,000
n= 89c = 2
Example:N = 10,000
n= 89c = 2
N = 10,000n = 89
Sample exactly n parts and inspect every part
Copyright Tapan P Bagchi 24
Producer’s & Consumer’s Risksdue to mistaken sentencing of lot
TYPE I ERROR = P(reject good lot)α or Producer’s risk
5% is common
TYPE II ERROR = P(accept bad lot)β or Consumer’s risk10% is typical value
Copyright Tapan P Bagchi 25
Acceptance sampling contd.
Producer’s riskRisk associated with a lot of acceptable quality rejected
Alpha α= Prob (committing Type I error)
= P(rejecting lot at AQL quality level)
= producers risk
Copyright Tapan P Bagchi 26
Acceptance sampling contd.
Consumer’s riskReceive shipment, assume good quality, actually bad quality
Beta β= Prob (committing Type II error)= Prob (accepting a lot at RQL quality level) = consumers risk
The Operating Characteristic (OC) curve for a sampling plan quantifies these risks
Copyright Tapan P Bagchi 27
Take a randomized sample of size n from
the lot of unknown quality p
The Single Sampling procedure
Inspect all items in the sample
Defectives found = d
d ≤ c ?Yes
Reject lot
Accept lot
No
Return lot to supplier
Do 100% inspection
Copyright Tapan P Bagchi 28
The Operating Characteristic or OC
Curve
The OC curve indicates a sampling plan’s ability to
discriminate between good and bad lots
Copyright Tapan P Bagchi 29
What is the 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 lotShows probability of lot acceptance Pa as function of lot quality level (p)
It is based on the sampling planCurve indicates discriminating power of the planAids in selection of plans that are effective in reducing risk Helps to keep the high cost of inspection down
Copyright Tapan P Bagchi 30
OC Definitions on the CurvePr
obab
ility
of A
ccep
ting
Lot
Lot Quality (Fraction Defective)
100%
75%
50%
25%
.03 .06 .09
α = 0.1090%
β = 0.10
AQ
L
RQ
LIndifferent
Good Bad
Copyright Tapan P Bagchi 31
OC Curve TermsAcceptable Quality Level (AQL)
Percentage of defective items a customer is willing to accept from you (a property of mfg. process)
Lot Tolerance Percent Defective (LTPD)Upper limit on the percentage of defects a customer is willing to accept ( a property of the consumer)
Average Outgoing Quality (AOQ)Average of rejected lots and accepted lots
Average Outgoing Quality Limit (AOQL)Maximum AOQ for a range of fractions defective
Copyright Tapan P Bagchi 32
The Perfect OC CurvePr
obab
ility
of A
ccep
ting
Lot
Lot Quality (Fraction Defective)
100%
75%
50%
25%
.03 .06 .09
This curve distinguishes perfectly between good and bad lots.
What would allow you to achieve a curve like this?
Copyright Tapan P Bagchi 33
OC Curves
Lot Quality (Fraction Defective)
Prob
abili
ty o
f Acc
eptin
g Lo
t100%
75%
50%
25%
.03 .06 .09
OC Curves come in various shapes depending on the sample size and risk of α and β errors
This curve is more discriminating
This curve is less discriminating
Copyright Tapan P Bagchi 34
OC Curve and its interpretation
The oc-curve enables us to evaluate the probability of acceptance (Pa) for any true lot quality level-on a what-if basis. This way, one can design sampling plans that perform the way one wants. The OC curve is Interpreted as follows:
If the lot quality is 0.093 fraction defective, then Pa, is 0.05. If the lot quality is 0.018 fraction defective, then Pa, is 0.95.
Copyright Tapan P Bagchi 35
Operating Characteristic Curve
AQL LTPD
β = 0.10
α = 0.05Pr
obab
ility
of
acce
pta
nce
, P a
{
0.60
0.40
0.20
0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20
0.80
{
Proportion defective p
1.00
OC curve for n and c
Copyright Tapan P Bagchi 36
OC Curve helps visualize producer’s and consumer’s points
Type I and Type II decision errors correspond logically to the two decision points on the oc-curve. Type I error -- Wrongful RejectionA type I error is associated with the producer's point -- to reject when the true value of the quality characteristic is AQL. The risk of rejecting an AQL lot is the producer'srisk (α = alpha risk)Type II error -- Wrongful acceptance
A Type II error is to accept when the true value of the quality characteristic is RQL -- at the consumer's point. The risk of accepting a lot, if it is an RQL lot, is the consumer's risk (ß = beta risk).
Copyright Tapan P Bagchi 37
Decision Criteria
0
1.00
Prob
abili
ty o
f acc
eptin
g lo
t
Lot quality (fraction defective)
“Good” “Bad”
Ideal
Not verydiscriminating
Copyright Tapan P Bagchi 38
Consumer’s and Producer’s Risk
00.10.20.30.40.50.60.70.80.9
1
0 .05 .10 .15 .20 .25
Prob
abili
ty o
f acc
eptin
g lo
t
Lot quality (fraction defective)
α = .10
β = .10
“Good”
AQL
“Bad”Indifferent
LTPD
Copyright Tapan P Bagchi 39
Example: QC Curve for n = 10, c = 1Example: QC Curve for n = 10, c = 1
00.10.20.30.40.50.60.70.80.9
1
0 .10 .20 .30 .40 .50
Prob
abili
ty o
f acc
epta
nce
Fraction defective in lot
.9139
.7361
.5443
.3758
.2440.1493
.0860
Copyright Tapan P Bagchi 40
Types of OC Curves
Type AGives the probability of acceptance for an individual lot coming from finite production
Type BGive the probability of acceptance for lots coming from a continuous process or infinite size lot
Copyright Tapan P Bagchi 41
OC Curve Calculation
The Ways of Calculating OC Curves
Binomial distribution (Type B)
Hypergeometric distribution (Type A)
Pa = P(r defectives found in a sample of n)Poisson formula
P(r) = ( (np)r e-np)/ r!Larson nomogram
Copyright Tapan P Bagchi 42
OC Curve Calculation by Poisson distribution
A Poisson formula can be usedP(r) = ((np)r e-np) /r! = Prob(exactly r defectives in n)
Poisson is a limit Limitations of using Poisson
n ≤ N/10 total batch Little faith in Poisson probability calculation when n is quite small and p quite large.
For Poisson, Pa = P(r ≤ c)
Copyright Tapan P Bagchi 43p
For us, Pa = P(r ≤ c)
Copyright Tapan P Bagchi 44
OC Curve Calculation by Binomial Distribution
Note that we cannot always use the binomial distributionbecause
Binomials are based on constant probabilities
N may not be infinitep changes as items are drawn from the lot
Copyright Tapan P Bagchi 45
OC Curve by Binomial FormulaPa p (%)
.998 .01
.980 .02
.930 .03
.845 .04
.739 .05
.620 .06
.502 .07
.394 .08
.300 .09
.223 .10
.162 .11
.115 .12
Using this formula with n = 52 and c=3 and p = .01, .02, ...,.12 we find data values as shown on the right. This givens the plot shown below.
Copyright Tapan P Bagchi 46
The Ideal OC Curve
p AQL
1.0
0.0
Pa● 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
Copyright Tapan P Bagchi 47
Guidelines for choosing Producer’s and Consumer’s decision points (AQL and RQL)To choose the Producer's Point (AQL) in practice
Lots at the producer's point quality level (AQL) should be accepted most of the time. Define AQL accordingly. Take into account historical quality levels and the consequence of returning a lot to the producer at AQL quality level. Choose the producers risk of rejecting a lot that is of AQL quality. Typical: α = 0.05.
To choose the Consumer's Point (RQL) in practiceLots at the consumer's point quality level (RQL) should be rejected most of the time. Take into account the economic consequence of the consumer’s accepting a lot at RQL quality level and define RQL accordingly. Choose the consumers risk of accepting a lot that is of RQL quality. Typical: β = 0.05.
Copyright Tapan P Bagchi 48
The OC Curve by binomial formula
Operating characteristic (OC) curve
Curve plots the probability of acceptingthe lot (Pa) versus the lot fraction defective (p)
Copyright Tapan P Bagchi 49
The OC Curve for Single SP
Copyright Tapan P Bagchi 50
Ideal OC curve
Theoretically , it can be achieved by 100% inspection of the lotDue to sampling, ideal OC curve cannot be perfectly achieved.If sample size (n) is large, the OC curve shape will approach the ideal OC curve
Copyright Tapan P Bagchi 51
Properties of OC Curves
The acceptance number c and sample size n are most important factors in defining the OC curveDecreasing the acceptance number (c) is preferred over increasing sample size (n)The larger the sample size n the steeper is the OC curve (i.e., it becomes more discriminating between good and bad lots)
Copyright Tapan P Bagchi 52
Effects of nn on OC curves
As sample size (n) increases,increases, the OC curve shape will approach the idealOC curve ~ more discriminating
Note that c is kept in proportional to n
movement of OC
Copyright Tapan P Bagchi 53
Effects of cc on OC curves
Increasing acceptance number (c) does not significantly change the shape of OC
But Producer’s risk (β) increases
movement of OC
Copyright Tapan P Bagchi 54
OC curveAcceptable quality level (AQL)
The poorest quality level for the supplier’s process that a consumer would consider to be acceptable A property of the supplier’s manufacturing process, not a property of the sampling plan
Lot tolerance percent defective (LTPD)The protection obtained for individual lots of poor qualityAlso called rejectable quality level (RQL)LTPD is a level of lot quality specified by consumer, not a characteristic of the sampling planSampling plans can be designed to have specified performance at the AQL and the LTPD points
Copyright Tapan P Bagchi 55
Type-A and Type-B OC Curves
Copyright Tapan P Bagchi 56
Poisson distribution for Defects
Poisson parameter: λ = npP(r) = (np)r e-np/r! = Prob(exactly r defectives in n)
This formula may be used to formulate equations involving AQL,RQL, α and β to given (n, c).
We can use Poisson tables to approximately solve these equations. Poisson can approximate binomial probabilities if n is large and p small.
Q. If we sample 50 items from a large lot, what is the probability that 2 are defective if the defect rate (p) = .02? What is the probability that no more than 3 defects are found out of the 50?
Copyright Tapan P Bagchi 57
Hypergeometric DistributionHypergeometric formula:
r defectives in sample size n when M defectives are in N.This distribution is used when sampling from a small population. It is used when the lot size is not significantly greater than the sample size. (Can’t assume here each new part picked is unaffected by the earlier samples drawn).
Q. A lot of 20 tires contains 5 defective ones (i.e., p = 0.25). If an inspector randomly samples 4 items, what is the probability of 3 defective ones?
⎟⎟⎠
⎞⎜⎜⎝
⎛
⎟⎟⎠
⎞⎜⎜⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛−−
=
Nn
Mr
MNrn
rP )(
Copyright Tapan P Bagchi 58
Properties of OC Curves for finite size lots
Copyright Tapan P Bagchi 59
Properties of OC Curves
If the acceptance level c is changed, the shape of the curve will change. All curves permit the same fraction of sample to be nonconforming.
Copyright Tapan P Bagchi 60
“Rectifying” Inspection
When the lot is rejected, it requires corrective action: 100% inspection and replacement of defectives by good parts
Rectified
Passed
Copyright Tapan P Bagchi 61
Average Outgoing Quality (AOQ)
Expected proportion of defective items passed to customer
Average outgoing quality limit (AOQL) isThe “maximum” point on AOQ curve
NnNpPinspectionrectifyingwithAOQ a )( −
=
Copyright Tapan P Bagchi 62
Average outgoing quality -AOQ
Given the lot size of N, sample size of n, and fraction defective of pn items in the sample that, after inspection, contain no defectives, because all discovered defectives are replacedN – n items that, if the lot is rejected, also contain no defectivesN – n items that, if the lot is accepted p(N-n) defectives
If N >> n, then
AOQ = Pa p
Copyright Tapan P Bagchi 63
Average Quality
Average outgoing quality (AOQ): Average of inspected lots (100%) and uninspected lots
⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛ −×= NnNpPaAOQ
Pa = Probability of accepting lotp = Fraction defectiveN = Lot sizen = Sample size
Copyright Tapan P Bagchi 64
AOQLimit (AOQL) Estimation
0 00.05 0.0460.1 0.074
0.15 0.0820.2 0.075
0.25 0.0610.3 0.045
0.35 0.030.4 0.019
0
0.02
0.04
0.06
0.08
0.1
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Approximate AOQL = .082
AO
Q (F
ract
ion
defe
ctiv
e ou
t)
Incoming fraction defective
Copyright Tapan P Bagchi 65
AOQ Curve
0.015
0.010
0.005
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10
AOQLAverageOutgoingQuality
(Incoming) Percent Defective
AQL LTPD
Copyright Tapan P Bagchi 66
Average outgoing quality -AOQ
Copyright Tapan P Bagchi 67
Double Sampling Plans
Take small initial sampleIf # defectives < lower limit, accept
If # defectives > upper limit, reject
If # defectives between limits, take second sample
Accept or reject lot based on 2 samples
Less inspection than in single-sampling
Copyright Tapan P Bagchi 68
Multiple Sampling Plans
Advantage: Uses smaller sample sizes
Take initial sample
If # defectives < lower limit, accept
If # defectives > upper limit, reject
If # defectives between limits, re-sample
Continue sampling until accept or reject lot based on all sample data
Copyright Tapan P Bagchi 69
Sequential Sampling
The ultimate extension of multiple samplingItems are selected from a lot one at a timeAfter inspection of each sample a decision is made to accept the lot, reject the lot, or to select another item
In Skip Lot Sampling only a fraction of the lots submitted are inspected
Copyright Tapan P Bagchi 70
Sequential Sampling PlansSequential Analysis is the technique by which we build up our sample one item at a time, and after inspecting each item, ask ourselves: "Can we be sure enough to accept or reject this batchon the information so far collected?“Seq SP offers Its value is in enabling reliable conclusions to be wrung from a minimum of data. This was deemed sufficient to require that it be classified "Restricted " within the meaning of the Espionage Act during the war of 1939-45.Used for lot-by-lot acceptance plans for applications needing the minimum possible average sample numbers.The SPR type sequential sampling plan is also known as Columbia sequential sampling because the methodology was developed at Columbia University.
Copyright Tapan P Bagchi 71
Example of Sequential Decision Diagram
Results of sequential sampling data is plotted on the diagram sequentially as it is collected. The accept or reject decision is made when sample data plots on or over an acceptance line or rejection line. The stair step pattern is caused by the discrete nature of the sample size and the number of defectives in the n-table upon which this diagram is based.
Copyright Tapan P Bagchi 72
Sequential Sampling: Acceptance and Rejection Limit lines
Copyright Tapan P Bagchi 73
Equations for the Limit Lines
The equations for the two limit lines are functions of the parameters p1, , p2, and .
where
Copyright Tapan P Bagchi 74
OC-Curve for Count Data in Sequential Sampling—similar to other sampling plans
Copyright Tapan P Bagchi 75
Choosing A Sampling Method:Single, Double or Sequential?
It is an economic decisionSingle sampling plans
high sampling costs
Double/Multiple sampling planslow sampling costs
More complex to implement
Copyright Tapan P Bagchi 76
Designing a Sampling Plan
The plan must protect both the consumer (the buyer) and the
supplier (the seller)
Copyright Tapan P Bagchi 77
d ≤ c ?
Reject lot
YesAccept lot
Do 100% inspection
Return lot to supplier
Inspect all items in the sample
Defectives found = d
No
Take a randomized sample of size n from
the lot of unknown quality p
Designing The Single Sampling plan
Copyright Tapan P Bagchi 78
Sampling Plan Design by Binomial Distribution
Binomial distribution:
P(x defectives in n) = [n!/(x!(n-x))!]px(1- p)n-x
Recall n!/(x!(n-x))! = ways to choose x in n
Q. If 4 samples (items) are chosen from a population with a defect rate = .1, what is the probability that a) exactly 1 out of 4 is defective?
b) at most 1 out of 4 is defective?
Copyright Tapan P Bagchi 79
Designing a single-sampling plan with a specified OC curve
To construct sampling plan such that
Probability of acceptance for lots with fraction defective p1
Probability of acceptance for lots with fraction defective p2
Two simultaneous conditions that the (n, c) plan should meet
Copyright Tapan P Bagchi 80
Solving for (n, c)
To design a single sampling plan we need two points. Typically these are p1 = AQL, p2 = LTPD and , are the Producer's Risk (Type I error) and Consumer's Risk (Type II error), respectively. By binomial formulas, n and c are the solution to
These two simultaneous equations are nonlinear so there is no simple, direct solution. The Larson nomogram can help us here.
Copyright Tapan P Bagchi 81
SSP Design by Larson Nomogram
● Applies to single sampling plan
● Based on binomial distribution
● Uses1-α = Pa at AQL
β = Pa at RQL● Can produce OC
curve
Copyright Tapan P Bagchi 82
Two simultaneous conditions that the (n, c) plan will meet
Designing a single sampling plan by Larson’s Binomial Nomogram
Inputs are AQL, α, RQL and β.Outputs are n and c
Copyright Tapan P Bagchi 83
The MIL-STD-105E approachA Query from a Practitioner: Selecting AQL (acceptable quality levels)“I'd like some guidance on selecting an acceptable quality level and inspection levels when using sampling procedures and tables. For example, when I use MIL-STD-105E, how do I to decide when I should use GI, GII or S2, S4?”
-- Confused in Columbus, Ohio
W. Edwards Deming observed that the main purpose of MIL-STD-105 was to beat the vendor over the head.
"You cannot improve the quality in the process stream using this approach," cautions Don Wheeler, author of Understanding Statistical Process Control (SPC Press, 1992). "Neither can you successfully filter out the bad stuff. About the only place that this procedure will help is in trying to determine which batches have already been screened and which batches are raw, unscreened, run-of-the-mill bad stuff from your supplier. I taught these techniques for years but have repented of this error in judgment. The only appropriate levels of inspection are all or none. Anything else is just playing roulette with the product."
Copyright Tapan P Bagchi 84
MIL-STD-105E
Original version (MIL STD 105A) issued in 1950 as tables; Last version (MIL STD 105E) in 1989; ISO adopted it as ISO 2859Plan covers sampling by attributes for given lot size (N) and acceptable quality level (AQL).Prescribes sample size n, acceptance number c, and rejection number rStandard included three types of inspection—normal, tightened and reduced and gives switching rulesPlans assure producer’s risk (α) of 0.01 – 0.1. The only way to control the consumer’s risk (β) is to change inspection level
Copyright Tapan P Bagchi 85
Military Standard 105ESampling procedure for inspection by attributes developed duringWorld War II and is the most widely used acceptance-sampling system for attributes in the world todayA collection of sampling schemes; therefore an acceptance-sampling systemProvides for three types of sampling: single, double, and multiplePrimary focal point is the acceptable quality level (AQL)
Different AQLs may be designated for different types of defects: critical, major, and minorGenerally specified in contract or by authority responsible for sampling
Sample size is determined by lot size and by choice of inspection level
Copyright Tapan P Bagchi 86
AQL Acceptance Sampling by Attributes by MILSTD 105E
Determine lot size N and AQL for the task at handDecide the type of sampling—single, double, etc.Decide the state of inspection (e.g. normal)Decide the type of inspection level (usually II)Look at Table K for sample sizesLook at the sampling plans tables (e.g. Table IIA)Read n, Ac and Re numbers
Copyright Tapan P Bagchi 87
Military Standard 105E Procedure1. Choose the AQL2. Choose the inspection level3. Determine the lot size4. Fine the appropriate sample size code letter (from
the MIL-STD 105E Table)5. Determine the appropriate type of sampling plan to
use (single)6. Enter the appropriate table to find the type of plan to
be used7. Determine the corresponding normal and reduced
inspection plans to be used when required
Copyright Tapan P Bagchi 88
MIL STD 105E Inspection levels and Sample Size codes
Copyright Tapan P Bagchi 89
OC Curves under different Inspection Conditions
Copyright Tapan P Bagchi 90
MIL STD 105E “Normal” Sample Size n and defective limit c
Copyright Tapan P Bagchi 91
The OC curve of a MIL STD 105E Sampling Plan
Copyright Tapan P Bagchi 92
MIL-STD-105E Normal Inspection
Copyright Tapan P Bagchi 93
MIL-STD-105E Reduced Inspection
Copyright Tapan P Bagchi 94
MIL-STD-105E Tightened Inspection
Copyright Tapan P Bagchi 95
Military Standard 105ESampling rules for switching between normal, tightened and reduced inspection
Copyright Tapan P Bagchi 96
OC Curves under different MIL Inspection Conditions
Copyright Tapan P Bagchi 97
Discussion of Military Standard
Several points about MIL STD 105E should be emphasized:MIL STD 105E is AQL-oriented, no concern for βNot all possible sample sizes are possible (2,3,5,8,13,20,32,50,etc.)Sample sizes are related to lot sizesSwitching rules are subject to criticism for both misswitchingbetween inspection plans and discontinuation even though there has been no actual quality deteriorationMIL STD 105E has been replaced by MIL STD 1916 by the US Department of Defense as of October 2008 http://assist.daps.dla.mil/
Copyright Tapan P Bagchi 98
How/When would you use Acceptance Sampling?
Advantages of acceptance samplingLess handling damagesFewer inspectors to put on payroll100% inspection costs are to high100% testing would take to long
Acceptance sampling has some disadvantagesRisk included in chance of bad lot “acceptance” and good lot “rejection”Sample taken provides less information than 100% inspection
Copyright Tapan P Bagchi 99
Summary
Many basic terms you need to know to be able to understand acceptance sampling
SPC, Accept a lot, Reject a lot, Complete Inspection, AQL, LTPD, Sampling Plans, Producer’s Risk, Consumer’s Risk, Alpha, Beta, Defect, Defectives, Attributes, Variables, ASN, ATI.
Copyright Tapan P Bagchi 100
Some Key Definitions and TermsReference: NIST Engineering Statistics Handbook
Acceptable Quality Level (AQL): A percent defective that is the base line requirement for the quality of the producer's product. The producer would like the buyer to design a sampling plan such that there is a high probability of accepting a lot that has a defect level less than or equal to the AQL.
Lot Tolerance Percent Defective (LTPD) also calledRQL (Rejection Quality Level): A designated high defect level that would be unacceptable to the consumer. The consumer would like the sampling plan to have alow probability of accepting a lot with a defect level as high as the LTPD.
Copyright Tapan P Bagchi 101
Type I Error (Producer's Risk): The probability for a given (n, c) sampling plan of rejecting a lot that has a defect level equal to AQL. The producer suffers when this occurs, because α lot with acceptable quality was rejected. The symbol a is commonly used for the Type I error and typical values for α to range from 0.2 to 0.01.
Type II Error (Consumer's Risk): This is the probability, for a given (n, c) sampling plan, of accepting a lot with a defect level equal to the LTPD. The consumer suffers when this occurs, because a lot with unacceptable quality was accepted. The symbol β is commonly used for the Type II error and typical values range from 0.2 to 0.01.
Copyright Tapan P Bagchi 102
Average Outgoing Quality (AOQ):A common procedure, when sampling and testing is non-destructive, is to 100% inspect rejected lots and replace all defectives with good units. In this case, all rejected lots are made perfect and the only defects left are those in lots that were accepted. AOQ's refer to the long term defect level for this combined LASP and 100% inspection of rejected lots process. If all lots come in with a defect level of exactly p, and the OC curve for the chosen (n, c) LASP indicates a probability Pa of accepting such a lot, over the long run the AOQ can easily be shown to be:
where N is the lot size.
Copyright Tapan P Bagchi 103
Average Outgoing Quality Level (AOQL): A plot of the AOQ (Y-axis) versus the incoming lot p (X-axis) will start at 0 for p = 0, and return to 0 for p = 1 (where every lot is 100% inspected and rectified). In between, it will rise to a maximum. This maximum, which is the worst possible long term AOQ, is called the AOQL.
Average Total Inspection (ATI): When rejected lots are 100% inspected (rectifying inspection), it is easy to calculate the ATI if lots come consistently with a defect level of p. For a sampling plan (n, c) with a probability Paof accepting a lot with defect level p, we have
ATI = n + (1 - Pa) (N - n)where N is the lot size.
Copyright Tapan P Bagchi 104
Average Sample Number (ASN):For a single sampling plan (n, c) we know each and every lot has a sample of size n taken and inspected or tested.
For double, multiple and sequential plans, the amount of sampling varies depending on the number of defects observed.
For any given double, multiple or sequential plan, a long term ASN can be calculated assuming all lots come in with a defect level of p.
A plot of the ASN, versus the incoming defect level p, describes the sampling efficiency of a given lot sampling scheme.
Copyright Tapan P Bagchi 105
Useful links
http://www.bioss.sari.ac.uk/smart/unix/mseqacc/slides/frames.htmAcceptance Sampling Overview Text and Audio
http://iew3.technion.ac.il/sqconline/milstd105.htmlOnline calculator for acceptance sampling plans
http://www.stats.uwo.ca/courses/ss316b/2002/accept_02red.pdfAcceptance sampling mathematical background
http://assist.daps.dla.mil/The new US military sampling procedure
top related