Feature extraction for vibration-based fault detection in Permanent Magnet Synchronous Motors

This paper presents a study of different time and frequency indicators, extracted from vibration signals of a Permanent Magnet Synchronous Machine (PMSM), for fault detection and diagnosis purpose. First, a multi-physical model of the machine, able to generate the vibration displacement under different operating conditions, is simulated. Then, a brief state-of-the-art on the most encountered machine faults, with their electromagnetic and mechanical signatures, is presented. In this study, both rotor eccentricity and Permanent Magnet (PM) demagnetization faults are considered and introduced in the model by changing some of its parameters. After that, time- and frequency-domains signal processing techniques are applied to extract crucial features sets, related to healthy and faulty cases, from vibration signals. The evolution of these indicators with respect to faults degrees is analyzed to select proper criteria for each one, which can be effectively used latter in fault detection and diagnosis applications.