XGBoost for Multi-Fault Diagnosis and Prediction in Permanent Magnet Synchronous Machines

In this study, we propose a data-driven diagnostic system that uses Extreme Gradient Boosting (XGBoost) to detect, classify, and assess the severity of multiple faults in permanent magnet synchronous motors (PMSMs). The three main fault categories that are the focus of the suggested method are inter-turn short-circuit (ITSC) faults, stator open-circuit faults, and permanent magnet demagnetization. To capture fault-specific characteristics and their development with severity, discriminative electrical features are retrieved from stator currents, flux linkage, and dq-axis values. Next, using the chosen electrical indications, an aggregated diagnostic index is created to facilitate defect diagnosis and severity quantification in a single learning process. The XGBoost-based model has been shown to produce excellent diagnostic accuracy and robust separation between various fault causes via extensive assessment. It also maintains dependable performance under previously unknown operating or fault situations. These findings show that an XGBoost-only approach offers a scalable and efficient way to monitor advanced PMSM conditions in industrial and safety-critical applications.