Predicting Failure Precursor Parameters of SiC MOSFETs Based on the RTNET Neural Network

Predicting the failure precursor parameter of silicon carbide metal-oxide-semiconductor field-effect transistors (SiC MOSFETs) is of practical significance due to the potential failure and economic losses resulting from their aging and deterioration. Therefore, there has been significant interest in investigating the prediction of failure precursor parameters in SiC MOSFETs. This paper introduces the RTNET model for Predicting the failure precursor parameter of SiC MOSFETs and the predicted parameter value can be employed to estimate the Remaining Useful Life (RUL) of SiC MOSFETs. By analyzing the failure mechanisms of SiC MOSFETs, on-resistance (Rds,on) has been selected as the failure precursor parameter and extracted through power cycling tests. The RTNET model, Autoformer model, and DLinear model have been evaluated using Mean Squared Error (MSE) and Mean Absolute Error (MAE) as performance metrics. The results indicate that the RTNET model outperforms the DLinear model by 7.2% to 32%. Due to the continuous operation of SiC MOSFET, a large amount of data is generated, and the RTNET model is capable of handling long-term time series forecasting (LTSF) problems effectively. Therefore, the RTNET model can meet the prediction requirements for failure precursor parameters of SiC MOSFET.