Location of an inter-turns short-circuit fault in the stator windings of induction motor by neural network

In order to avoid a complete degradation of an induction motor (IM) and to direct the maintenance task, the detection and location of an early fault is of great concern. This presentation deals with the location of the inter turns short-circuit in the stator windings of a three phases induction motor using an artificial neural network (ANN). A multi layer perceptron feedforward ANN is proposed. The shape and the root mean square (RMS) value of line's residual currents for several numbers of shorted turns on each of the three phases and under different load conditions are learned. The diagnostic procedure consists on correlate the amplitude of the three residual currents to the faulty phase. The residual current is the difference between the faulty current and the healthy one. In this fact, the obtained ANN is able to locate correctly any number of shorted turns at any phase. In noisy environment, the ANN structure is not able to provide us good results. The originality consists on overcoming this problem, on computing the RMS values of the residual over an adequate cycle. The test results reveal the efficiency of locating an inter turns short circuit fault even under different load conditions and noisy signals.