Monitoring and diagnosis of induction motors electrical faults using a current Park's vector pattern learning approach

Various applications of artificial neural networks (ANN's) presented in the literature prove that such technique is well suited to cope with online faults diagnosis in induction motors. The aim of this paper is to present a methodology by which induction motors electrical faults can be diagnosed. The proposed methodology is based on the so-called Park's vector approach. In fact, stator current Park's vector patterns are first learned, using ANN's, and then used to discern between "healthy" and "faulty" induction motors. The diagnosis process was tested on both classical and decentralized approaches. The purpose of a decentralized architecture is to facilitate a satisfactory distributed implementation of new types of faults to the initial NN monitoring system, The generality of the proposed methodology has been experimentally tested on a 4-kW squirrel-cage induction motor. The obtained results provide a satisfactory level of accuracy, indicating a promising industrial application of the hybrid Park's vector-neural networks approach.