Analysis of stator faults in induction machines using growing curvilinear component analysis

Fault detection of shorted turns in the stator windings of Induction Motors (IMs) is possible in a variety of ways. As current sensors are usually installed together with the IMs for control and protection purposes, using stator current for fault detection has become a common practice nowadays, as it is much cheaper than installing additional sensors. In this study, stator currents from the healthy and faulty IMs are obtained and analysed via MATLAB® software. The current signatures from healthy and faulty IMs are conditioned using the inbuilt DSP module of the dSPACE prior to analysis using AI techniques. This paper presents a Growing Curvilinear Component Analysis (GCCA) neural network which is able to correctly identify anomalies in the IM and follow the evolution of the stator fault using its current signature, making on-line early fault detection possible.