Purpose of Lot Inspection

Acceptance Sampling

u Not

used to estimate lot quality Quality Control Tool be used as an audit tool

Reject/Accept decision only

u Not a u Should

incoming supply meets specifications output from a process meets specifications

WOOD/STAT 335 Quality Improvement

Lot Inspection
3 Methods u Do Not Inspect u 100% Inspection u Acceptance Sampling

Why Use Acceptance Sampling?

u Destructive

Testing

u Cost of inspection is high u 100%

inspection not feasible from a time standpoint u Vendor has good performance, but some checking should be done u When there are liability issues

Advantages & Disadvantages

Advantages u Less expensive u Less handling & damage u Reduces inspection error u Lot rejection motivates supplier u Can do destructive tests Disadvantages u Risk of accepting bad, rejecting good u Less information is known about the product u Requires planning & documentation of the sampling procedure

Types of Sampling Plans

u Type of Measurement

Attributes - go/no go Variables - measured quantity

u Sampling Schemes

Single Sampling Double Sampling Multiple Sampling

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page1

Lot Formation
u Should be homogeneous

Single Sampling Plans

Terminology
N= n= c= d= Lot Size Sample Size Allowed number of defects Observed number of defects

same machines, operators, etc, makes problem detection easier

u Larger

lots preferred

u Handling systems should be designed to

facilitate sampling u Samples must be random

Procedure
From a homogeneous lot of size N, take a random sample of n items. If the number of defects found in the sample (d) exceeds the allowable number defective (c) - reject the lot. Otherwise accept the lot.

Procedure
From a homogeneous lot of size N, take a random sample of n items. If the number of defects found in the sample (d) exceeds the allowable number defective (c) - reject the lot. Otherwise accept the lot.

Procedure - Important Points

From a homogeneous lot of size N, take a random sample of n items. If the number of defects found in the sample (d) exceeds the allowable number defective (c) - reject the lot. Otherwise accept the lot.

Measure of Performance
Operating Characteristics Curve u Plots Pa vs. p
Pa - Probability of Accepting the Lot p - Fraction Defective in the Lot

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page2

Calculating Pa
Pa = P{d c} = n! p d (1 p ) d !( n d )! d =0
c n d

Example
Sample of Furniture Parts
N = Lot Size is sufficiently large to use binomial n = 89 c= 2 d = Observed from Sample

When sampling from large lots and the binomial can be used.

Data for OC Curve

p 0.000 0.005 0.010 0.010 0.015 0.020 0.025 0.030 0.035 0.040 0.045 0.050 0.055 0.060 0.065 0.070 0.075 0.080 0.085 0.090 0.095 0.100 Pa 1.0000 0.9897 0.9397 0.9397 0.8502 0.7366 0.6153 0.4985 0.3936 0.3042 0.2307 0.1721 0.1265 0.0919 0.0659 0.0468 0.0329 0.0230 0.0159 0.0109 0.0074 0.0050

Resulting OC Curve
OC Curve
1.000 0.900

n = 89 c=2

n! Pa = p1d 1 p1d d =0 d !(n d )! =

or Pa = binomdist(c,n,p,true )
2

n d

0.800 0.700 0.600 Pa

89! p d 1 p d d =0 d !(89 d ) !

89 d

0.500 0.400 0.300 0.200 0.100 0.000 0.000 0.020 0.040 0.060 Lot Fraction Defective 0.080 0.100 0.120

Perfect Sampling Plan

100% Inspection What would the OC Curve Look Like?

Perfect Sampling Plan

Perfect Plan OC Curve - 100% Inspection
1.000 0.900 0.800 0.700 0.600 Pa 0.500 0.400 0.300 0.200 0.100 0.000 0.000 0.020 0.040 0.060 Lot Fraction Defective 0.080 0.100 0.120

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page3

Types of Curves
u Type A OC Curve

Type A and B OC Curves

Lot Size = 200

Type A & B OC Curves for n = 89, c = 2

1.0000 0.9000 0.8000 0.7000 0.6000 Pa 0.5000 0.4000 0.3000 0.2000 0.1000 0.0000 0.000 0.020 0.040 0.060 Fraction Defective (p) 0.080 0.100 0.120 Type B Type A

Probabilities for an individual lot Use the hypergeometric distribution Always lies below the Type A Curve
u Type B OC Curve

For a large lot, or a stream of lots Binomial is correct distribution

Type A and B OC Curves

Lot Size = 1000

Points on the OC Curve

u AQL - Acceptable

Type A & B OC Curves for n = 89, c = 2

1.0000 0.9000 0.8000 0.7000 0.6000 Pa 0.5000 0.4000 0.3000 0.2000 0.1000 0.0000 0.000 0.020 0.040 0.060 Fraction Defective (p) 0.080 0.100 0.120 Type B Type A

Quality Level

poorest quality consumer would accept as a process average

u LTPD - Lot

Tolerance Percent Defective

poorest quality consumer would accept in a given lot this is higher than the AQL also know as:
RQL - Rejectable Quality Level LQL - Limiting Quality Level

Resulting OC Curve
OC Curve
1.000 0.900 0.800 0.700 0.600 Pa 0.500 0.400 0.300 0.200 0.100 0.000 0.000

Designing a Sampling Plan

Want an OC Curve Passing Through 2 Points u Where Pa = 1 , for lots with P1 Defective u Where Pa = , for Lots with P2 Defective

This gives us 2 equations and 2 unknowns

AQL 0.020

0.040

Lot Fraction Defective

0.060 LTPD

0.080

0.100

0.120

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page4

Designing a Sampling Plan

n! 1 = p1d (1 p1d ) d = 0 d !( n d ) ! n! d = p d 1 p2 ) (n d )! 2 ( d =0 d !
c c n d

Resulting OC Curve
OC Curve
1.000 0.900 0.800 0.700 0.600 Pa 0.500 0.400 0.300

Will reject a lot better than P1 with probability

n d

Solve these 2 equations for n and c

0.200 0.100 0.000 0.000

Will accept a lot worse than P2 with probability

P1

0.020

0.040

2 Lot Fraction Defective

0.060 P

0.080

0.100

0.120

Ways to Solve the Equations

u Nomograph

Rectifying Inspection

Approach
Incoming Lots P0 Rejected Lots Inspection Accepted Lots Outgoing Lots

u Tabular Approach

Duncans Dodge-Romig Tables Military Standards

Rectifying Inspection
Fraction Defective 0

Rectifying Inspection
Fraction Defective 0

Incoming Lots P0

Rejected Lots Inspection Accepted Lots

Outgoing Lots

Incoming Lots P0

Rejected Lots Inspection Accepted Lots

Outgoing Lots

Fraction Defective P0

Fraction Defective P0

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page5

Rectifying Inspection
Fraction Defective 0

Measures of Effectiveness
u Average

Outgoing Quality

Incoming Lots P0

Rejected Lots Inspection Accepted Lots

Outgoing Lots p1 < p0

Average Fraction Defective of Outgoing Lots

u Average

Outgoing Quality Limit

Fraction Defective P0

Maximum Fraction Defective of Outgoing Lots

u Average

Total Inspection

Average Number of Samples Taken

Average Outgoing Quality

Outgoing Quality u if lot accepted, outgoing quality is same as incoming quality, p
Incoming Lots P0 Rejected Lots Inspection Accepted Lots
Fraction Defective P0 Fraction Defective 0

Average Outgoing Quality

Outgoing Quality u if lot accepted, outgoing quality is same as incoming quality, p u if lot rejected, all defects are fixed, outgoing quality p = 0
Incoming Lots P0 Rejected Lots Inspection Accepted Lots
Fraction Defective P0 Fraction Defective 0

Outgoing Lots p1 < p0

Outgoing Lots p1 < p0

Average Outgoing Quality

In general,

Average Outgoing Quality

In general,

AOQ = Pa * p + (1 Pa ) * 0

AOQ = Pa * p + (1 Pa ) * 0

Accepted Lots

Rejected Lots

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page6

Average Outgoing Quality

Leaving,

Average Outgoing Quality

Leaving,

AOQ = Pa * p

AOQ = Pa * p
However, you must correct for the sample taken since defects in the sample will be fixed or replaced.

Average Outgoing Quality

Final AOQ equation:

AOQ Curve
AOQL Curve
0.0180 0.0160

Outgoing Fraction Defective

P p( N n ) AOQ = a N

0.0140

n = 89, c = 2

0.0120

0.0100 n = 89, c = 1 0.0080

0.0060

0.0040

n = 89, c = 0

0.0020

0.0000 0.000 0.020 0.040 0.060 0.080 0.100 0.120

Incoming Lot Fraction Defective

Average Outgoing Quality Limit

Maximum AOQ Possible u As lot quality worsens, you are more likely to reject the lot. u When the lot is rejected, it is rectified. u Therefore, at some point outgoing quality starts to improve.

Average Outgoing Quality Limit

AOQL Curve
0.0180 0.0160

0.0140

n = 89, c = 2

Outgoing Fraction Defective

0.0120

0.0100 n = 89, c = 1 0.0080

0.0060

0.0040

n = 89, c = 0

0.0020

0.0000 0.000 0.020 0.040 0.060 0.080 0.100 0.120

Incoming Lot Fraction Defective

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page7

Average Total Inspection

Inspection u always must inspect n units for the sample
Fraction Defective 0

Average Total Inspection

Inspection u always must inspect n units for the sample u must detail rejected lots, inspecting N units
Incoming Lots P0 Inspect n units Rejected Lots
Fraction Defective 0

Incoming Lots P0

Inspect n units

Rejected Lots

Outgoing Lots p1 < p0

Outgoing Lots p1 < p0

Accepted Lots

Fraction Defective P0

Accepted Lots

Fraction Defective P0

Average Total Inspection

In general,

Average Total Inspection

In general,

ATI = n + (1 Pa )( N n)

ATI = n + (1 Pa )( N n)

Accepted Lots

Rejected Lots

Average Total Inspection

Average Total Inspection
12000.0

Double Sampling
Idea Behind Double Sampling u Take 1st Sample
If decision is clear, accept or reject the lot
u If

10000.0

Average Total Inspection

8000.0

n = 89, c = 0

n = 89, c = 1

n = 89, c = 2

not, Take 2nd Sample

6000.0

4000.0

2000.0

0.0 0.000 0.020 0.040 0.060 0.080 0.100 0.120

Lot Fraction Defective

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page8

Double Sampling
Idea Behind Double Sampling u Take 1st Sample
If decision is clear, accept or reject the lot
u If
Incoming Lots P0

Double Sampling
Terminology
n1 c1 n2 c2 sample size on 1st sample acceptance number on 1st sample sample size on 2nd sample acceptance number on both samples

not, Take 2nd Sample

Inspect n1 units Rejected Lots Inspect n2 units Rejected Lots

Accepted Lots

Accepted Lots

Double Sampling
Procedure
Incoming Lots P0 If d1 > c2 reject Inspect n1 units If d1 c1 accept Inspect n2 units If d1+d2 c2 accept If d1+d2 > c2 reject

First Sample
Inspect n1 units
Incoming Lots P0 If d1 > c2 reject Inspect n1 units If d1 c1 accept

Accept the lot if no more than c 1 defects occur in 1st sample.

First Sample
Inspect n1 units
Incoming Lots P0 If d1 > c2 reject Inspect n1 units If d1 c1 accept

Double Sampling
Otherwise, Proceed to 2nd Sample
Incoming Lots P0 If d1 > c2 reject Inspect n1 units If d1 c1 accept Inspect n2 units If d1+d2 c2 accept If d1+d2 > c2 reject

Reject the lot if more than the total allowable number of defects (c2) occur in 1st sample.

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page9

Double Sampling
Procedure
Incoming Lots P0 If d1 > c2 reject Inspect n1 units If d1 c1 accept Inspect n2 units If d1+d2 c2 accept If d1+d2 > c2 reject

Double Sampling
Procedure
Incoming Lots P0 If d1 > c2 reject Inspect n1 units If d1 c1 accept Inspect n2 units If d1+d2 c2 accept If d1+d2 > c2 reject

Accept the lot if no more than c 2 defects occur in both samples.

Reject the lot if more than c 2 defects occur in both samples.

OC Curve for Double Sampling

Pa = PaI + PaII

Example
Consider the Double Sampling Plan: n1 = 50 c1 = 1 n2 = 100 c2 p =3 = .05

Accepted on 1st Sample

Accepted on 2nd Sample

Example, continued
Pa for first sample is the probability of observing no more than 1 defect on first sample.

Example, continued
Pa for first sample is the probability of observing no more than 1 defect on first sample. Pa (first) = binomdist(c1, n1, p, TRUE) = 0.279

PaI =

d1 = 0 1

50! d (50 d )! p (1 p)
1 d1 1

50 d1

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page10

Example, continued
Pa for second sample is the probability of observing no more than 3 total defects on both samples, given that you observed d 1 defects on the first. Pa (second) = Prob(d 1=2) * Prob(d 2<=1) + Prob(d 1=3) * Prob(d 2=0) = 0.01 + 0.001

Example, continued
Pa for both samples is the sum of Pa (first) and Pa(second). Pa (both) = Pa (First) + Pa (Second) = 0.279 + 0.011 = 0.290

All points on the OC Curve are calculated the same way.

Average Sample Number

In Double Sampling, number of samples taken is variable. ASN = = Where: PI = + P(lot is accepted on first sample) P(lot is rejected on first sample) n1 PI + (n1+n 2) (1-PI ) n1 + n2 (1- PI )

Average Sample Number

In Double Sampling, number of samples taken is variable. ASN = = Where: PI = + P(lot is accepted on first sample) P(lot is rejected on first sample) n1 PI + (n1+n 2) (1-PI ) n1 + n2 (1- PI )

Prob. of making any decision on the 1st sample

Average Sample Number

In Double Sampling, number of samples taken is variable. ASN = = 50 + 100 (1- 0.519 ) 98

Curtailment
Usually, the decision is made as soon as the number of defects found exceeds d2. This is called a curtailed sample. ASN curve for a curtailed Double Sample always lies below the Single Sampling Plan.

On average, 98 samples are taken with this plan

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page11

AOQ - Double Sampling

With Rectifying Inspection:

ATI - Double Sampling

With Rectifying Inspection:

AOQ =

[P

I a

( N n1 ) + PaII ( N n1 n2 ) p N

ATI = n1PaI + ( n1 + n2 ) PaII + N (1 Pa )

Wood 335 Quality Improvement Course Notes Dr. Thomas Maness

#10 Acceptance Sampling November, 2001 Page12