Induction motor bearing fault detection using a sensorless approach

Abstract

Continuous condition assessment of induction motors is very important due to its potential to reduce down-time and manpower needed in industry. Rolling element bearing faults result in more than 40% of all induction motor failures. Vibration analysis has been utilized to detect bearing faults for years. However, vibration sensors and expert vibration interpretation are expensive. This limitation prevents widespread monitoring of continuous bearing conditions in induction motors, which provides better performance compared to periodic monitoring, a typical practice for motor bearing maintenance in industry. A strong motivation exists for finding a costeffective approach for the detection of bearing faults. Motor terminal signals have attracted much attention. However, not many papers in the literature address this issue as it relates to bearing faults, because of the difficulties in effective detection. In this research, an incipient bearing fault detection method for induction motors is proposed based on the analysis of motor terminal voltages and currents. The basic idea of this method is to detect changes in amplitude modulation between the spatial harmonics caused by bearing faults and the supply fundamental frequency. This amplitude modulation relationship can be isolated using the phase coupling property. An Amplitude Modulation Detector (AMD), developed from higher order spectrum estimation, correctly captures the phase coupling and isolates these modulation relationships. In this research, in-situ bearing damage experiments are conducted so that the accelerated life span of the bearing can be recorded and investigated. Experimental results shown in this dissertation are based on different power supplies, load levels, VSI control schemes, and motor operating conditions. Taking the mechanical vibration indicator as a reference for fault detection, the proposed method is demonstrated to be effective in detecting incipient bearing faults in induction motors. If motors are operating at near steady state conditions, then experimental results show that the bearing fault detection rate of the proposed approach is 100%, while no false alarms are recorded.