Artificial neural network based on-line partial discharge monitoring system for motors

Corona discharge (CD) and partial discharge (PD) indicate early stages of insulation problems in motors and generators. Early detection of CD/PD will enable better coordination and planning of resources such as maintenance personnel, ordering of parts, etc. Most importantly, one can prevent catastrophic failures during normal operations. In decades, on-line PD measurement has been used to find loose, delaminated, overheated, and contaminated defects before these problems lead to failures. As a result, on-line PD monitoring has become an important tool for planning machine maintenance. Many methods are available to measure the PD activities in the operating machines. The electrical techniques usually measure the currents by means of a high frequency current transformer at neutral points or detect the PD pulses via high voltage capacitors connected to the phase terminals. Those methods are generally expensive and easy to be interfered by the noise due to the considerations of the high frequency and low signal levels. Instead of using high frequency analysis, this paper extracts the low frequency characteristics of PD/CD faults and develops a low cost PD/CD on-line health monitoring system for motors. The system employs an artificial neural network (ANN) with multiple sensors inputs for PD/CD diagnostic task. The proposed algorithms and circuits are implemented and tested in the laboratory environment. Results show that the system is sensitive and accurate