STATISTICAL VIBRATION-BASED FAULT DIAGNOSIS APPROACH APPLIED TO BRUSHLESS DC MOTORS

This paper proposes a statistical fault diagnosis approach based on vibration analysis, applied to brushless DC motors. First, the model of an inverter-fed permanent magnet motor, simulated using Matlab/Simulink, is briefly presented. The proposed model enables the generation of electrical, magnetic and vibration signals under healthy and faulty behaviors. The presented work focuses mainly on the rotor demagnetization and eccentricity faults. Then, different indicators including time, frequency, space and space harmonic characteristics are extracted from vibrations for different cases. These features are analyzed with respect to fault type and severity to select the most discriminative ones. For fault detection, a statistical test is used to compare each of the selected indicators with a decision threshold to detect the occurrence of a fault in the motor. The values of different thresholds are calculated in order to achieve a given low false alarm rate (alpha). The test power (1 - beta) of each fault indicator is also evaluated for its corresponding threshold. The fault isolation is then realized using a fault signature table. Finally, the proposed approach is tested on two sets of noisy simulated data related to different machine conditions. (C) 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.