Physics-Informed Machine Learning for Robust Remaining Useful Life estimation of Power MOSFETs

Power semiconductor devices are vital in modern power converters, accounting for a fifth of reported system downfalls. Failures in these devices, especially power MOSFETs, pose therefore significant reliability challenges of power converter systems. In this context, predicting the Remaining Useful Life (RUL) of components like power MOSFETs is crucial for effective predictive maintenance. Conventional data-driven approaches using neural networks have been applied to address this challenge. However, due to ignoring the physical properties of the target RUL function, neural networks can result in unreasonable RUL estimates. In response, we propose a Physics-Informed Machine Learning (PIML) implementation that integrates physical laws into the learning process. Our framework not only demonstrates substantial improvements in both precision and physical consistency for the RUL estimation but also exhibits faster learning convergence and better generalizability. This method lays a robust foundation for enhancing RUL prediction in power MOSFETs, addressing critical needs in predictive maintenance for power systems.