Number of Lifetimes for Testing

[HOTMETAL]

I recently had some vibration test done on a part. The person who performed the test recommends that the accelerated lifetime test be performed for 2 effective product lifetimes. He claims that testing for one lifetime only gives a 50% confidence that units will survive their operating life. Testing for 2 lifetimes (suppossedly industry standard) will supposedly give a 95% confidence. Is this assertion correct?

[Dr. Dave Olwell]

Your question can be summed up as: how many parts do I need to test, and for how long, to demonstrate a certain reliability with a certain confidence. To answer that question, we need to know the anticipated life distribution for the part, the reliability that is to be demonstrated, the acceleration mechanism (to verify the claim that one knows the "effective product lifetime"), and the confidence level. Your post only provides the confidence level, so it is hard to be of service without more information.

I am not aware of any industry standard that says that if you test one part for twice its lifetime, you demonstrate a certain reliability with 95% confidence. I doubt it is accurate -- for example, with an exponential distribution, to demonstrate 95% reliability with 95% confidence, you would have to have a single unit last 59 effective lives without failure.

A general comment: the design of reliability tests tool in Weibull++6.0 is very helpful for answering this type of question, once one has all the information.

Best wishes,

Dr. Dave Olwell