Product Reliability Calculation

[Raffat Ali]

I am working on belts "retractor" reliability. Anytime you pull on the belt and buckle up,it is a "pass condition". You pull again after unbuckling and it sticks, it is a "failure condition".
14 retractor and each belt/retractor went thru 21 cycles.
Assume sample 3 and 7 failed on cycles 9 and 13.

What would be a good, quick way to calculate reliability and waht distribution, Weibull, Normal etc.
Thanks,

Raffat

[Pantelis]

The first step is to create the data set:

I am assuming that all specimens – with the exception of 3 and 7 went to 21 cycles without failure. (thus you have 12 items suspended – or right censored – at 21 cycles).

Specimens 3 and 7 failed at 9 and 13 cycles respectively thus you have a failure at 3 and 7.

With this data - fit a model.
A quick way to do it is to use the itools on Weibull.com (http://www.weibull.com/itools/index.htm) with the following data set:

3, 7, -21, -21, -21, -21, -21, -21, -21, -21, -21, -21, -21, -21

Alternatively use Weibull++.

Now I should point out that fitting a Weibull to this data - with only two failure points - is risky.

[Dennis Craggs]

How often is the customer going to use the retractor? I think many more than 21 times, perhaps thousands. Therefore testing 21 cycles is not a reliability or durability test.

This test is more of a functional evaluation. How frequent are operational failures on a brand new retractor? If the design is not robust, then failures will be frequent. I recommend that design alternatives should be evaluated in a DOE test plan.