Test Samples Selection

Test sample selected should be selected RANDOMLY only. Otherwise, you will not get representative results.

For defining number of samples to be tested, (sample plan), there are International/ national standards. This depends upon many things, including your batch quantity and risk factors

Dear Guest -

a) if you get time, it might be worth your while to register, and tell us a little more about what it is you would like to accomplish - minus details it is hard to help other than to give you the most general direction

b) gsuhas is correct - a random (continual not continuous) sampling will need to be generated and tested - and the resulting data should be incorporated back into your process capability to understand what if any additional 'control' you have on the outcome - I'm wondering how many significant posts will happen in this thread before it wanders off topic -

c) I suspect this group will get into a big discussion about 'random' in general -

let me offer a couple of manufacturing thoughts to help open up your quest a bit -

1) 'truly random' against a standard AQL random testing procedure could have you select the first, say 125 parts out of 2500, run your tests, accept on 7 bad, reject on 8 and then put all the parts good and bad back into the containers for shipping - or it could be the last 125, or one every some many except when divisible by 7 etc - random is a tough idea to get across; and then the randomness in box two might be a different sampling variation with the same acceptance criteria.

2) on the other hand - the first 125 pieces in a box are not truly representative to any great degree of the last 125 going into the box, unless you have established that the PROCESS is both 'capable' and 'stable' - here were some other threads discussed and offered up by bright CR4'ers previously about process capability discussion as well -

http://cr4.globalspec.com/thread/15149/Process-Capability?frmtrk=CR4digest

http://www.qualityamerica.com/knowledgecente/articles/cqeIVH4_1.htm