Re: REQUEST: Acceptance Quality Level (AQL) and Consumer Risk

["Dr. W. J. Latzko" <Latzko@att.net>]

At 11:13 AM 11/14/2000 -0500, limsh@iqmsia.po.my wrote:

>How to correlate AQL with consumer risk? Let say we set AQL to 0.65, what 
>would be the consumer risk in this case?

The AQL (Acceptable Quality Level does not correlate to the consumer risk 
without specifying the sampling plan used. The RQL (Rejectable Quality 
Level - approximate consumer risk at 10%) ranges from 27.1% for plan F to 
1.41% for plan R in MIL-STD 105E.

To explain all this it is necessary to understand the OC (Operating 
Characteristic) curve which is a plot giving the percent of lots accepted 
at a given process average. This data is tabulated in Table X (pages 34 
through 66 in MIL-STD-105E). The sample size, acceptance and rejection 
numbers dictate the OC curve. The risks depend on these factors.

Having said all of this, I question the value of acceptance sampling in 
case of non-destructive testing. Even with non-destructive testing, 
acceptance sampling is a questionable method of achieving quality.  Since 
Mood's article in 1943 "on the dependence of sampling inspection plans upon 
population distributions" (Ann. Math. Stat. 14: 415-425) we have known that 
the theoretical basis of acceptance sampling depends on various assumption 
which are often not found in practice. For instance, the plans are based on 
random sampling in the technical sense. That is often too costly or 
physically impossible.

It is better to use process control in a System of Profound Knowledge to 
achieve the company's quality objective. For details on these concepts 
study Deming's last two books, Out of the Crises
  and the New Economics.

Bill Latzko
---------------------------------------------------------------------
Dr. William J. Latzko
215 - 79th Street, North Bergen, NJ 07047-5727
Telephone & Fax: +1-201-868-5338
E-mail: Latzko@att.net
---------------------------------------------------------------------