Reliability prediction sensitivity analysis — How to perform reliability prediction time efficiently

Reliability prediction according to reliability standards is often applied to estimate the reliability of electronic systems. Failure rates are typically dependent on their stress conditions (predominantly thermal and electrical stresses). However, getting the exact temperature and stress value for every single component is extremely time consuming, or sometimes impossible. Here the sensitivity of reliability prediction on the stresses is investigated in a case study of a voltage-source converter. It is found that for electrical stress designers´ rules are sufficient. For thermal stress, measured temperatures where used as input. It was found that reliability prediction using averaged temperatures can provide very good agreement, compared to reliability prediction with measured component temperatures. These drastic simplifications allow to perform reliability prediction in a time efficient way. The case study was performed using IEC 62380, Telcordia SR 322 (2), and MIL-HDBK 217F(2), including the ANSI/VITA 51.1 (2008) report that updates the 15 year old MIL-HDBK 217F(2).