Predictive Fault Diagnosis Inter-Turn Short Circuit PM Motor based on Machine Learning Algorithms for Elevator Applications

The occurrence of short circuits in the stator windings is one of the main issues of Permanent Magnet Synchronous Motors (PMSMs) used particularly in elevator systems. Therefore, it is important to develop reliable techniques and tests for fault diagnosis in time before passenger safety problems arise. Many research efforts have been focused on the utilization of diagnostic techniques and methodologies for motor tolerance and faults of the drive system. The use of system predictive maintenance, by developing Machine Learning (ML) models, has been widely adopted as a diagnostic procedure of internal short circuit analysis. In this paper, the authors will evaluate the motor operation in healthy and faulty conditions and propose a predictive maintenance methodology using ML algorithm to monitor the system condition. While other similar techniques have been shown to fail in reliability and accuracy issues, this work proposes an integrated predictive maintenance framework based on sensor data acquisition, and hybrid supervised and unsupervised methods, demonstrating classification accuracy up to 97 %.