Electronics failure rates typically follow the exponential distribution and models like Prism, Mil-Hdbk-217 or 217Plus (newer) are used to predict failure rates.
We are currently in the production phase and fielding our electro-mechanical systems but there seems to be a manufacturing defect within the capacitors by the vendor where much higher failures are being experienced attributed to knit-line defects or delamination at the electrode-dielectric junction, and the failures are occurring during final testing just prior fielding or even after they have been fielded.
My question is, if the capacitor have been modeled using our assumptions (exponential dist, power, vibe and thermal stresses), and manufacturing defects or screening escapes have occurred, then doesn't this mean that our modeled failure rate is NOT relevant anymore? Even calculating a point estimated failure rate (number of failures/accrued number of hours) is pointless (pun, ha) since this particular failure trend is un-modeled and a point estimate only represents a failure rate up until that point in time, and tomorrow we could realize we have many more failures in the field and that failure rate isn't a rate as expected.
How can this new failure rate be calculated/modeled given this manufacturing defect/screening escape? I realize the best thing to do is to chase the root cause and have corrective actions in place but at this point that is difficult as there are many fielded systems already in place.
Best Answer
So if you’re positively sure the failure comes from the vendor your original assumption that the capacitor will eventually fail given an exponential distribution is false. This distribution applies to random events caused by the component’s exposure to time and stresses. What you seem to be experiencing is a quality issue normally expressed in ppm by the manufacturer. What is odd is that from your explanation this is being unexpectedly high. My suggestion is to get in touch with the manufacturer and request for an analysis of one of your defective capacitors. At the same time check the ppm of this component to realize the correct failure rate you should expect before going into field.