Induction motor fault diagnosis based on the k-NN and optimal feature selection

The k-nearest neighbour (k-NN) rule is applied to diagnose the conditions of induction motors. The features are extracted from the time vibration signals while the optimal features are selected by a genetic algorithm based on a distance criterion. A weight value is assigned to each feature to help select the best quality features. To improve the classification performance of the k-NN rule, each of the k neighbours are evaluated by a weight factor based on the distance to the test pattern. The proposed k-NN is compared to the conventional k-NN and support vector machine classification to verify the performance of an induction motor fault diagnosis.