Multiswitch Fault-detection for VSI fed Multiphase Motor Drive Based on Machine Learning

Detecting early faults in an electric drive system in order to maintain reliability and uninterrupted post-fault operation is an extremely difficult task. In recent years, early fault detection has become one of the most important areas of cutting-edge research in the fields of electric vehicles, offshore-ship propulsion, air taxis, electric vertical take-off and landing, etc. In this paper, a generalized fault detection method for the five-phase induction motor drive is presented based on an optimized support vector machine (SVM) learning algorithm. Using the second low-frequency processing method, low-frequency signals were extracted from fault currents and employed for SVM training. This expedites fault detection and reduces memory allocation. The proposed fault detection algorithm has been validated through simulation in steady-state and dynamic loading conditions of the drive for a variety of fault scenarios.