Diagnosis of Stator Fault in Asynchronous Machine Using Soft Computing Methods

Stator Winding Fault can be detected by monitoring any abnormality of the Park’s spectrum. In this paper, a fault-detection performance comparison is presented between the Support Vector Machine (SVM) and backpropagation algorithm (BP) using experimental data for a healthy and faulty case. Support Vector Machine and Back propagation Algorithm provide environments to develop fault-detection schemes because of their multi-inputprocessing and its good generalization capability. The training patterns are obtained using motor current signature analysis (MCSA) and using Spectral Park’s Vector. The neural networks are evaluated by means of the crossvalidation technique to determine easily the diagnosis and severity of turn-toturn faults.