A Robust Fault Detection and Isolation of A DC Motor

Abstract

This paper presented the results of the experiments on an electrical motor in order to perform on-line diagnosis. The implementation of a dynamic decision is interesting in the case of closed-loop systems for which the residual responses may be transient. It was shown that recurrent neural networks (RNNs) are well adapted to fault isolation of the electrical machine. A neural network classifier was developed to diagnose the sensor and actuator faults from the residuals generated, and simulation results were presented to demonstrate the satisfactory achieved results. The classifier construction has been designed to cope with transient behaviours of the residuals generated by abrupt faults. The robustness with respect to torque disturbance and the sensitivity with respect to drift fault amplitude were experimented.