P. 1
Control Charts

|Views: 8|Likes:

See more
See less

11/02/2012

pdf

text

original

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

c. • Charts based on Poisson distribution.Charts • c and u charts deal with nonconformities. u . • Sample size. . – u Chart – nonconformities per unit. – c Chart – total number of nonconformities. constant probabilities.

.

.

.

.

How to Interpret Attribute Charts • Points beyond limits. – Below lower limits means process has improved. shifts do apply.primary test. Only get One Chart !! . • Rules for trends. • Zone rules do not apply.

– Number of nonconforming floppy disks is found for samples of 200 for 25 trials.u charts– Number of paint blemishes on auto body observed for 30 samples.np charts– – Number of nonconforming cables is found for 20 samples of size 100. • c. .Examples of When to Use • p. – 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 .

. – Must define nonconformity. Attributes • Variables – Fit certain cases. – Provides summary level performance. • Attributes – Fit certain cases – taste. etc. – More expensive? – Identify mean shifts sooner before large number nonconforming.Comparison of Variables v. – Both mean and variation information. color. – Larger sample sizes.

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 .

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

.

• Part of Ch. • Still used in some applications today. 11 on Inspection Methods.Intro to Acceptance Sampling • Acceptance Sampling – a historically significant topic but less used today. .

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

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 .

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. .Role of Producer and Consumer Take a Sample Size ‘n’.

• Consumer’s risk – risk of accepting a ‘poor’ lot. • Limiting Quality Level . • Acceptable Quality Level (AQL) – Numerical definition of a ‘good’ lot. .Terminology • Producer’s risk – risk associated with rejecting a lot of ‘good’ quality. associated with the producer’s risk. associated with the consumer’s risk.Numerical definition of a ‘poor’ lot.

Examples • Producer’s risk is 5% for an AQL of 0. – Means batches that are 2% nonconforming are good and prefer to reject these no more than 5% of the time. – Means batches that are 8% nonconforming are bad and prefer to accept these only 10% of the time.02.08. . • Consumer’s risk is 10% for an LQL of 0.

Operating Characteristic (OC) Curve • Defines the performance of a sampling plan. . • Plots – probability of acceptance versus – proportion nonconforming (p).

4 0.01 0.p .8 0.04 0.03 0.2 0 Always Accept Always Reject 0.Ideal OC Curve 1.6 0.02 0.05 proportion nonconforming .2 1 P acceptance 0.2 0 -0.

reject if more. . – Accept the lot if ‘c’ or fewer nonconforming are obtained. • Calculate Paccept for range of incoming p levels. • OK to assume Binomial distribution (if lot size is 10x sample size).Actual OC Curves • Are determined by sample size [n] and acceptance number [c].

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

2 0 0 0.Create OC Curve b(x) = n x ( ) px (1-p)(n-x) x = 0.09 0.1.06 (p) 0. Vary p from 0 to 0.6 0.n 1.2 1 P acceptance Probability of accepting is obtaining c=2 or less nonconforming items in samples of size n=50.) 0.15 (what if p = ….03 0.8 0.15 .12 0...4 0.

3. 4. Inexpensive. 3.Acceptance Sampling • Pros 1. Risk of rejecting ‘good’ lots. Detects bad quality. less work than 100% inspection. Vary level of risk in decisions. 2. 4. Deming views on inspection. • Cons 1. not prevention or improvement. 2. . Flexibility – vary plan based on history. Plan to accept bad quality. Lots rejected – pressure on supplier.

03. P(x≤2)=0. This is Paccept.15. • Approximate Binomial by Poisson. • Vary p from 0 to 0.809. • Calculate p(2 or less). u=np. • Example n=50. u=1. p=0. .5.Sample Calculations • Binomial table only goes up to n=20.

15 .8 0. c= 2 1.6 0.Single sampling plan n = 50.12 0.2 0 0 0.03 0.4 0.06 (p) 0.2 1 P acceptance 0.09 0.

.11. • Producer’s Risk is 1-0.076. Then Paccept = . • Consumer’s Risk is 0.986 = 0. Then Paccept = 0.Producer and Consumer Risk • Assume AQL(acceptable quality level) is 0.01.014.076.986. • Assume LQL(limiting quality level) is 0.

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

r2 = a2 + 1 to ensure a decision on the sample. If a1 < d1 < r1. Denote the number of defectives in sample 1 by d1 and in sample 2 by d2. then: – If d1<= a1. Now this is compared to the acceptance number a2 and the rejection number r2 of sample 2. the lot is accepted. If D2 >= r2. The total number of defectives is D2 = d1 + d2. the lot is rejected. In double sampling. If d1 >= r1. d2. . the number of defectives. • If a second sample of size n2 is taken. the lot is rejected. the lot is accepted. – If D2 <= a2. a second sample is taken.Double Sampling Plan • Application of double sampling requires that a first sample of size n1 is taken at random from the (large) lot. is counted. The number of defectives is then counted and compared to the first sample's acceptance number a1 and rejection number r1.

1 0.8 0.6 0.2 0 0 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 0.4 0.05 p 0.

15 .2 1 0. C=4 N=200.2 0 0. C=8 N=500.8 P Accept N=50.Vary n and c 1.4 0. C=2 N=100.05 p 0.6 0.1 0. C=20 0.2 0 -0.

c=1 • Draw the OC Curve • What is Producer’s risk if AQL is 0.1. . • What is Consumer’s risk if LQL is 0.Class Problem • Acceptance Sampling Plan – n=30.02.

02 0.06 0. c=1 p 0 0.12 np (=u) Paccept (x<=1) 1 Plot Paccept vs.OC Curve Worksheet n=30.08 0.04 0.1 0. p .

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

scribd
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->