Fault detection in large AC machines

Abstract

The emerging techniques of artificial neural networks (ANNs) are applied to the problem of developing an artificial neural system capable of detecting interlayer faults in large AC machines using line-end coil voltage measurements. The proposed ANN system is a two-layer back propagation neural network, which is basically a classifier capable of recognizing data vectors buried in noise. The developed ANN system is fast to train and produced reliable fault detection and localization with noisy measurements. Furthermore, the proposed system needs neither data pre-processing nor feature extraction networks.