Neural network broken bar detection using time domain and current spectrum data

The induction motor is the most widely used in industry, covering approximately 85% of the total electric loads. As an unpredicted shutdown can be very costly, early detection and diagnosis of electric motors faults yields to economical losses reduction and operational efficiency improvement. Neural networks have been proposed to solve the induction motor broken bar detection and diagnosis problem. Current spectrum data are generally used but some statistical features of time domain data have also been considered. In this paper, a performance comparison of both types of incoming data for the neural network is accomplished.