Optimization of the feedforward neural network for rotor cage fault diagnosis in three-phase induction motors

We present the results of our investigation in the use of the multi-layer feed forward back-propagation artificial neural networks (ANNs) for induction machine faults diagnosis. ANNs are used effectively to determine the classification of induction machine rotor faults tested at different loads. First the raw signals are collected and features are extracted from the collected signals allowing the development of a data base necessary for the training of the ANNs. However, determining the ANN structure is a fundamental design issue and can be critical for the classification performance. The novelty in our works is that the genetic algorithms (GAs) can be used to select a smaller sub-set of ANN structure features that together form a genetically fit family for successful fault identification and classification tasks, at the same time, an appropriate simple structure of the ANN, in terms of the number of nodes in the hidden layer, can be determined. The proposed methodology is experimentally tested on 4kW/1500rpm induction machines at 50Hz/380V ; the obtained results provide a high level of accuracy.