Technical Paper
Machine Learning Approach for Open Circuit Fault Detection and Localization in EV Motor Drive Systems
2024-04-09
2024-01-2790
Semiconductor devices in electric vehicle (EV) motor drive systems (MDSs) are considered the most vulnerable component to physical faults. In this regard, open circuit fault (OCF) accounts for 38% of the total faults that may occur in a power converter. Various model-based studies for OCF localization have been presented in previous articles, mostly centered on the output current characteristics of the power converter. However, this proposed work presents a machine learning (ML) approach for OCF localization associated with active thermal management (ATM) based model predictive control (MPC) in a 2-level three-phase inverter. In this regard, the inherent property of ATM-based MPC to estimate the conduction losses for each power transistor has been employed. These conduction losses are then used to train a machine learning (ML) classifier, which can locate the OCF in any power transistor of MDS based on the dynamics associated with the conduction losses.