Deep learning based Incipient Stator Inter-turn Fault Diagnosis for Synchronous Reluctance Motor Drives

Synchronous Reluctance Motors (SynRMs) are becoming increasingly important in Electric Vehicle (EV) applications due to their robustness, high efficiency, high power density, and lower cost compared to PMSMs, which do require rare earth magnets. However, despite their robustness, SynRMs are also susceptible to faults. Therefore, diagnosing these faults is essential to enhance the system’s reliability. Inter-turn faults are the major type of fault that occurs in the stator windings of electric machines. Therefore, diagnosing inter-turn faults at an incipient stage can prevent unscheduled downtimes and further motor damage. This paper proposes a novel deep-learning-based architecture that diagnoses stator inter-turn faults in SynRMs at an incipient stage. The proposed architecture combines the EfficientNet deep learning model with Markov transitional fields, and it only uses the current and voltage signals generated from the developed customized motor under different speeds and loads. The algorithm demonstrated an accuracy of 97.10% and was compared to contemporary techniques to establish its robust performance.