Remaining Useful Life Prediction Method for MOSFET Based on Time Series Model

Abstract

This paper presents a remaining useful life prediction method for MOSFET based on time series model. First, the degradation data of MOSFET is acquired from mathematical model. Next, correlation test is conducted to determine difference order of time-series model and Akaike information criterion (AIC) is used to determine the order of autocorrelation model and moving average model, thereby determining the parameters of time series model. Then, short-term cycle prediction is added to improve prediction accuracy and reduce accumulated error. Finally, the effectiveness of the developed life prediction model is verified using Matlab/Simulink.