From: Stan Hilliard
Date: 13 Oct 2000
Time: 18:20:44
I am sorry for taking so long to reply.
It seems reasonable to me to use a sampling plan verify the supplier's decisions. Both random and rotating selection methods are valid. You did not say whether you would apply the same Z1.4 sampling plans as the supplier uses.
Here is an idea for another approach. Use a variables plan and sample each of the 40 dimensions once. Then calculate a K=(X-Spec)/SD for each dimension based on that dimension's specification. (K corresponds to Cpk times 3.)
Since K is dimensionless (units = standard deviation), you can treat the Ks of different dimensions as a single variable. In other words, treat K as you would any measured variable, and use a variables sampling plan for K. The advantage of this approach is that it enables you to make one variables acceptance decision for 40 dimensions. (Or a sample of the dimensions.)
A disadvantage is that very high (good) Ks weaken the ability to detect small (bad) Ks, since the Ks are averaged. To counteract this, you could set all Ks above a certain value -- like 3 -- to that maximum. You should also have a decision rule for the range (or standard the deviation) of the sample -- just as you would for any variables plan with known standard deviation.
Since the various dimensions could be correlated, you should consider taking each measurement on a different part.
Stan Hilliard