Lifetime prognostics for power module bonding wires under non-constant stress conditions: A compound stochastic process degradation modeling approach

This paper proposes a compound stochastic process degradation modeling approach for lifetime prognostics of power module bonding wires under non-constant stress conditions. The primary challenge in predicting the remaining useful life (RUL) of bonding wires lies in accurately characterizing both continuous thermal fatigue evolution and abrupt degradation jumps caused by random bond wire lift-offs, especially under dynamic stress conditions. To address this, a hybrid modeling framework integrating compound Poisson processes (CPP) with a state-dependent degradation dynamics model is developed. The continuous degradation increments are modeled using a Wiener process with drift, whose parameters are estimated via Relevance Vector Machine (RVM) to capture complex state-load coupling effects. Stochastic jumps induced by bond wire lift-offs are characterized by a state-dependent CPP, with parameters estimated through Variational Bayesian (VB) approach. Monte Carlo simulation is employed to approximate the first passage time (FPT) for RUL prediction, with the median of the RUL probability distribution adopted as the final result to ensure robustness in asymmetric distributions. Validation with 6 devices under test (DUTs) under different stress conditions shows that this method significantly outperforms traditional empirical models (e.g., LESIT) and time-series data-driven models (e.g., LSTM) in terms of prediction accuracy, especially in the near-end-of-life stage. This framework supports real-time updatable RUL prediction, providing strong support for predictive maintenance in power electronics systems. © 2012 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of Global Science and Technology Forum Pte Ltd.