Thermal Behavior-Informed Inter-Turn Fault Detection of PMSMs using Explainable AI in an Attention-Based Deep Learning Framework

Permanent Magnet Synchronous Motors (PMSMs) are widely used in high-performance applications in the industry due to their efficiency and compact design. However, given the non-linear behavior of PMSMs under fault conditions, the development of Inter-Turn Faults (ITFs) can jeopardize motor performance and safety. Understanding the influence of important variables on temperature rise following faults is difficult due to the inability of traditional fault detection techniques to handle these non-linear dynamics and their lack of interpretability. To fill this gap, this paper proposes a robust ITF detection method that uses thermal behavior analysis via an attention-based long short-term memory network to overcome these difficulties. To enhance model transparency, an explainable artificial intelligence approach is employed to interpret how motor variables affect temperature changes after ITFs, which helps to improve operational safety and efficiency in industrial settings by providing a dependable solution for real-time ITF identification in PMSMs as well as an understanding of underlying variables. Simulation results confirm the proposed method’s superiority over conventional approaches in both performance and interpretability.