From: Stan Hilliard shilliard@samplingplans.com
Date: 16 Jul 2001
Time: 22:18:49
Hi Khalil,
>> Questions : 1/ Do you find this sampling plan logical?
You could do better than that attribute plan. It wastes most of the information contained in the tensile measurements by converting the numbers to go/no-go.
I think that since you are measuring a variable -- tensile strength -- should switch to a sampling plan for variables with upper limit for tensile strength. For a given sample size, a variables plan will produce a much steeper oc curve -- which is which you would want.
You have a choice of using a plan for (1) variables percent nonconforming or (2) a plan for the mean tensile strength. Method (1) can be based on either a known historical within-lot standard deviation, or sigma unknown. Method (2) requires using a known standard deviation.
If you design a plan with the historical within-lot standard deviation "known", you still need to check the sample standard deviation of each lot against a limit to ensure that it hasn't changed significantly.
Software program TP414 for variables will design all of this based on two points on the oc curve that you specify. See
www.samplingplans.com/software_oc.htm
The software is well worth the $245.00 expense considering the importance of making the right decision economically. TP414 also develops sequential sampling plans, which are even more efficient and especially useful when testing is destructive. There is more information about sequential sampling in the tutorial:
www.samplingplans.com/modern3.htm
>> Questions : 2/ Are they special standards for destructive sampling plans?
I am not aware of any standards that deal exclusively with destructive sampling plans. The fact that the testing is destructive does not have any statistical impact. The destruction of samples only effects the economics of testing, and favors using a statistically valid plan for accept/reject decisions.
A question for you -- what is the type of product that you are testing?
Stan Hilliard