Robust Design Optimization for Failure Analysis

Failure analysis is important especially for the reliability of electronic components. Low failure probabilities and high dimensional input parameter spaces, including for example also manufacturing tolerances, usually require an unfeasible number of simulations and/or experiments to reach a high degree of certainty, at least when taking Monte-Carlo or similar approaches. With Robust Design Optimization we can study, as presented in this paper, the transition region between failure and non-failure very efficiently. For many applications, like Physical Health Management in combination with Digital Twins, information especially about the transition region from non-failure to failure is very important. This information must be passed along the value chain while keeping IP non-disclosed. This can be appropriately done using metamodels of failure probabilities (fragility surfaces).