Power MOSFET Remaining Useful Life Prediction Model Based on Fractional Poisson Process and Failure Threshold Variability

Remaining useful life (RUL) prediction of power MOSFET modules has critical significance for the intelligent maintenance of various industrial systems. To this aim, this paper proposes to use fractional Poisson process (fPp) to derive the adaptive RUL prediction model for the power MOSFET modules, considering the fractal characteristics of the degradation data. Firstly, a novel simulation algorithm for the fPp is proposed, which is more realistic and computationally efficient than the previous random walk accumulation method. The drift and diffusion coefficients in the degradation model are continuously updated to represent the environmental impact. Furthermore, a novel failure threshold (FT) variability strategy is adopted in our model, illustrating the unit‐to‐unit variability. Real thermal‐accelerated power MOSFET degradation data are employed to validate the proposed model. Results show that our model has better performance than the existing methods due to its better description of the degradation feature.