Enriched fault detection and classification for MOSFET via Wavelet Visu transform and isolate network

Multilevel inverters use MOSFETs for power distribution, but removing the noises, and identifying genuine transient faults in the MOSFET is more challenging. Hence a novel “Enriched Fault Detection and Classification for MOSFET via Wavelet Visu Transform and Isolate Network” is proposed. The existing fault detection method struggles to differentiate between noise-induced voltage fluctuations and intentional signal transitions due to parasitic capacitances in MOSFETs. Thus, a novel Wavelet Bayesian Visu Transform is implemented to minimize noise-induced switching events, remove precise noise from signal division, and improve signal clarity by denoising gate signals. Furthermore, classifying faults is challenging due to the potential for transient faults to be mistaken for noise, and interference in measurement data masking these faults. So, a novel Isolate Contractive Enduring Network is introduced, enhancing fault classification precision by eliminating interference and focusing on distinctive features. The proposed method exhibits superiority in fault detection and classification, with a low MAE of 2.4, a high recall of 0.98 %, and an accuracy of 98.8 %.