From: Stan Hilliard
Date: 10 Jan 2002
Time: 13:47:10
Sanjay,
I prefer the two-point method to develop sampling plans. You specifying a producer's point (AQL and Alpha) and a consumer's point (RQL and Beta). This determines the sample size and acceptance number. For more on this, see the tutorial at:
www.samplingplans.com/modern3.htm
I also use the same procedure in reverse. That is, choose by trial and error the sample size (n) and the acceptance number (C). This determines the AQL and RQL.
The thing that makes a sampling plan valid is that, for a lot, which is at some level of percent defective, you know the probabilities of accepting that lot. The oc curve shows this. The oc curve for attributes is the probability of acceptance versus true fraction defective of the lot.) The software program TP105 does these things and more.
www.samplingplans.com/software_oc.htm
Other than that, if you want to just look up a plan (as with a cookbook), Mil-Std-105E = Z1.4 is the most recent US standard. Be forewarned though that the AQL values in Mil-Std-105/Z1.4 do not tell you anything about the consumer's point. Therefore they do not tell you anything about the plan's exposure to acceptance of off-grade lots. For more on this, see:
www.samplingplans.com/aqlprimer.htm
I hope that this helps you to choosing a method for designing attribute sampling plans.
Stan Hilliard