Vibration-current data fusion and gradient boosting classifier for enhanced stator fault diagnosis in three-phase permanent magnet synchronous motors

Permanent magnet synchronous motors (PMSMs) are widely recognized for their precise control capabilities, making them indispensable in numerous industrial applications. Yet, their susceptibility to faults, particularly stator faults, can lead to severe operational challenges. Effective health monitoring and prompt fault detection are, therefore, pivotal for preserving the performance and lifespan of PMSMs. In this research, we delve into an innovative approach to fault diagnosis in PMSMs by fusing vibration and current data. For the experimental setup, stator faults were simulated as inter-turn short circuits. Comprehensive datasets encompassing three-phase current signals and vibration signals were acquired from the PMSM test rig. These datasets were subsequently processed into statistical features. Leveraging the information gain for feature selection, we discerned crucial attributes for the fault assessment. A gradient boosting-based machine learning model was then employed to distinguish between various fault states, utilizing the selected features. Our findings unveiled that the combined vibration-current data fusion approach stands out, achieving an impressive diagnostic accuracy of 90.7% and an area under the curve of 95.1%. This underscores the efficacy of data fusion in conjunction with gradient boosting for fault diagnosis. The methodology presented herein promises to pave the way for timely fault detection, enabling proactive maintenance regimes, and bolstering the reliability of PMSMs in critical industrial settings.