Prognosis of power MOSFET in continuous time framework

Abstract

The reliability of the MOSFET is a concern as it has been extensively used for the critical power supply applications in hybrid electric vehicles, high-speed electric trains, and electric aircraft. The MOSFET is subject to the multiple electro-mechanical stresses simultaneously during its operation. The degradation in MOSFET is random in nature due to these stresses along with manufacturing imperfection. The noise in the fault precursor measurement system results in large variation in the remaining useful life (RUL). In this paper, a continuous time (CT) framework based RUL estimation technique is proposed using the drain to source on-time resistance (R ds, on ) as the fault precursor. The CT framework is utilized to capture the dynamics of the system. The CT Markov Process (CTMP) in conjunction with Continuous time (CT) Sequential Importance Resampling (SIR) Particle Filter (PF) is used to simulate the trajectory of the R ds, on considering noise in measurement system. This simulated trajectory is then used to estimate the RUL of the MOSFET. The RUL estimation of the proposed method is compared with the RUL estimation of Kalman filter (KF) and Discrete Time Sequential Importance Sampling (SIS) PF using the experimental data of the R DS, ON of the MOSFET.