Ch 12- Control Charts for Attributes

p chart fraction defective np chart number defective

c, u charts number of defects

Defect vs. Defective

Defect a single nonconforming quality characteristic.
Defective items having one or more defects.

Legal Concerns with Term Defect

Often called nonconformity. Possible Legal Dialog
Does your company make a lot of defects? Enough to track them on a chart ? If they are not bad, why do you call them defects, sounds bad to me. So you knowingly track and ship products with defects?

Summary of Control Chart Types and Limits Table 12.3

These are again 3 sigma control limits

p, np - Chart
P is fraction nonconforming. np is total nonconforming.
Charts based on Binomial distribution. Sample size must be large enough (example p=2%) Definition of a nonconformity. Probability the same from item to item.

c, u - Charts
c and u charts deal with nonconformities.
c Chart total number of nonconformities. u Chart nonconformities per unit.

Charts based on Poisson distribution. Sample size, constant probabilities.

How to Interpret Attribute Charts

Points beyond limits- primary test.
Below lower limits means process has improved.

Zone rules do not apply. Rules for trends, shifts do apply.

Only get One Chart !!

Examples of When to Use

p,np charts
Number of nonconforming cables is found for 20 samples of size 100. Number of nonconforming floppy disks is found for samples of 200 for 25 trials.

c,u charts Number of paint blemishes on auto body observed for 30 samples. Number of imperfections in bond paper by area inspected and number of imperfections.

Control Chart Selection

Quality Characteristic

variable
defective n>1? no x and MR

attribute
defect

yes
n>=10 or no computer? x and R

constant yes p or sample np size?

no p-chart with variable sample size

constant sampling unit? yes c no u

yes
x and s

13

Comparison of Variables v. Attributes

Variables
Fit certain cases. Both mean and variation information. More expensive? Identify mean shifts sooner before large number nonconforming.

Attributes
Fit certain cases taste, color, etc. Larger sample sizes. Provides summary level performance. Must define nonconformity.

When are Shifts Detected ?

Lower

Specification
Limit LCL

Process Target

Upper Specification Limit UCL

Nonconformity

Control Chart Identifies Mean Shift Here

Attribute Chart Identifies Mean Shift Here

Variables v. Attributes
Both have advantages. At High levels - Attribute charts, identify problem areas.
At Lower levels Variables charts, quantitative problem solving tools.

Intro to Acceptance Sampling

Acceptance Sampling
a historically significant topic but less used today.

Part of Ch. 11 on Inspection Methods.

Still used in some applications today.

History and Status

Used extensively in WW II. Many Mil-Spec plans developed (105-E, ANSI/ASQC Z1.4-1993).
Still popular as a defense procurement tool.
Very large lots, screening tool. Low bid suppliers no history.

Acceptance Sampling Flow Chart

Lot Received from Supplier

Random Sample of Material Selected

Items Inspected and Analyzed

Results Compared with Acceptance Criteria

Define and Analyze Sampling Plan

Accept the Lot

Reject the Lot

Send Lot to Inventory or Production

Decide Disposition, Return to Supplier

Role of Producer and Consumer

Take a Sample Size n, Accept if c or less.

Producer
Risk is a good lot will be rejected and sent back.

Consumer
Risk is a bad lot will be accepted.

Terminology
Producers risk risk associated with rejecting a lot of good quality. Acceptable Quality Level (AQL) Numerical definition of a good lot, associated with the producers risk. Consumers risk risk of accepting a poor lot. Limiting Quality Level - Numerical definition of a poor lot, associated with the consumers risk.

Examples
Producers risk is 5% for an AQL of 0.02.
Means batches that are 2% nonconforming are good and prefer to reject these no more than 5% of the time.

Consumers risk is 10% for an LQL of 0.08.

Means batches that are 8% nonconforming are bad and prefer to accept these only 10% of the time.

Operating Characteristic (OC) Curve

Defines the performance of a sampling plan.
Plots
probability of acceptance versus proportion nonconforming (p).

Ideal OC Curve
1.2 1

P acceptance

0.8 0.6 0.4 0.2 0 -0.2 0

Always Accept

Always Reject
0.01 0.02 0.03 0.04 0.05

proportion nonconforming - p

Actual OC Curves
Are determined by sample size [n] and acceptance number [c].
Accept the lot if c or fewer nonconforming are obtained, reject if more.

OK to assume Binomial distribution (if lot size is 10x sample size). Calculate Paccept for range of incoming p levels.

Sample problem
Given a lot size of N=2000, a sample size n=50, and an acceptance number c=2.
Calculate the OC curve for this plan.

Create OC Curve
b(x) = n x

( )

px (1-p)(n-x)

x = 0,1,..n
1.2 1
P acceptance

Probability of accepting is obtaining c=2 or less nonconforming items in samples of size n=50.
Vary p from 0 to 0.15 (what if p = .)

0.8 0.6 0.4 0.2 0 0 0.03 0.06 (p) 0.09 0.12 0.15

Acceptance Sampling
Pros
1. Vary level of risk in decisions. Inexpensive, less work than 100% inspection. Flexibility vary plan based on history. Lots rejected pressure on supplier.

Cons
1. 2. 3. 4. Plan to accept bad quality. Detects bad quality, not prevention or improvement. Deming views on inspection. Risk of rejecting good lots.

2.
3. 4.

Sample Calculations
Binomial table only goes up to n=20. Approximate Binomial by Poisson, u=np. Calculate p(2 or less). This is Paccept.
Example n=50, p=0.03, u=1.5. P(x2)=0.809.

Vary p from 0 to 0.15.

Single sampling plan n = 50, c= 2

1.2 1
P acceptance

0.8 0.6 0.4 0.2 0 0 0.03 0.06 (p) 0.09 0.12 0.15

Producer and Consumer Risk

Assume AQL(acceptable quality level) is 0.01. Then Paccept = .986. Producers Risk is 1-0.986 = 0.014.
Assume LQL(limiting quality level) is 0.11. Then Paccept = 0.076. Consumers Risk is 0.076.

Designing Plan Performance

Vary n and c to obtain different OC curves.
Single and multiple sampling. Refer to standard published sampling plans.

Double Sampling Plan

Application of double sampling requires that a first sample of size n1 is taken at random from the (large) lot. The number of defectives is then counted and compared to the first sample's acceptance number a1 and rejection number r1. Denote the number of defectives in sample 1 by d1 and in sample 2 by d2, then:
If d1<= a1, the lot is accepted. If d1 >= r1, the lot is rejected. If a1 < d1 < r1, a second sample is taken.

If a second sample of size n2 is taken, the number of defectives, d2, is counted. The total number of defectives is D2 = d1 + d2. Now this is compared to the acceptance number a2 and the rejection number r2 of sample 2. In double sampling, r2 = a2 + 1 to ensure a decision on the sample.
If D2 <= a2, the lot is accepted. If D2 >= r2, the lot is rejected.

Vary n and c
1.2 1 0.8 0.6 0.4 0.2 0 0 0.05 p 0.1 0.15 n=50 c=2 n=80 c=2 n=50 c=1 n=100 c=2

P accept

Vary n and c
1.2 1 0.8
P Accept

N=50, C=2 N=100, C=4 N=200, C=8 N=500, C=20

0.6 0.4 0.2 0 -0.2 0 0.05 p 0.1 0.15

Class Problem
Acceptance Sampling Plan n=30,c=1
Draw the OC Curve What is Producers risk if AQL is 0.02. What is Consumers risk if LQL is 0.1.

OC Curve Worksheet n=30, c=1 p 0 0.02 0.04 0.06 0.08 0.1 0.12 np (=u) Paccept (x<=1) 1

Plot Paccept vs. p

Homework
Read Toyota Production System case and think about application of the Deming 14 points we will discuss this next class.