Dynamic Remaining Useful Lifetime (RUL) Estimation of Power Converters based on GaN Power FETs

Abstract

The rapid advancement of power Gallium Nitride (GaN) devices is making them an attractive option in the industry to achieve high power density and efficiency. However, their reliability has been a concern for the industry, primarily since these devices have been used commercially only for a few years. Hence, this paper presents a method for the remaining useful lifetime (RUL) prediction of GaN-based converter system in real-time. Considering the most critical parts of failure in the power converters are the GaN devices and the capacitors, experimental degradation data for 10 GaN devices, and 4 capacitors were extracted over several weeks of operation in a laboratory setup. The degradation data mainly includes the measurements of the drain-source resistance for the GaN devices, ESR for capacitors, and the components' temperature. A statistical approach using probability density functions (PDFs) in uniform distribution is proposed to predict the probability of survival at the system level, based on the experimental degradation data. The RUL prediction accuracy using the proposed PDFs method is high since it utilizes real qualification data. Under different operating profiles (temperature and current stress levels), the degradation data is extracted. Finally, detailed analysis and discussions are pointed out based on the experimental results.