Stator turn-to-turn fault detection of induction motor by non-invasive method using generalized regression neural network

Condition monitoring and protection methods based on the analysis of the machine's current are widely used according to non-invasive characteristics of current transformers. It should be noted that, these sensors are installed by default in the machine control center. On the other hand, condition monitoring based on mathematical methods has been proposed in literature. However, they are model based and are too complex. Artificial neural network (ANN) methods are robust and less model dependent for fault diagnosis when the fault signature can be directly achieved using the sampling data. In this procedure, the state of internal process will be ignored. Therefore, generalized regression neural network (GRNN) based method is presented in this paper that uses negative sequence currents (calculated from the machine's currents) as inputs to detect and locate an inter-turn fault in the stator windings of the induction motor. Turn-to-turn fault by changing the contact resistance and various numbers of shorted turns for realizing the fault severity has been modeled by Matlab/Simulink. The simulation and experimental results show that the proposed method is effective for the diagnosis of stator inter-turn fault in induction motor under the supply voltage unbalances.