I assume you are looking at time to get to the failure from a bench test? In that case, Exponential Distribution is used, or if that is disproved, then Weibull. If you are talking how much load did the item get loaded to, pretty much a "normal" Student T could be used, though depends on the rigor/complexity of the experimental design. Or are you just talking - I'm going to dissect this completed product and see what the internals look like?