Detection of Incipient Inter-Turn Short-Circuit Faults by Artificial Intelligence Classifiers

This paper presents two artificial intelligence (AI) based identification methods for inter-turn short-circuit fault (ISCF) detection in induction motors (IMs), driven by voltage source inverters (VSIs). It was previously observed that, for an IM driven by finite control set model predictive control (FCS-MPC), the ISCF occurrence disturbs the balanced distribution of the resultant switching vectors, which are merely the control outputs of the FCS-MPC scheme. This effect of the ISCFs is utilized for fault detection purposes. The proposed method successfully detects the ISCF using AI methods which are fed by histograms of switching vectors along with torque and speed. This is especially convenient from the motor driver's perspective since no additional sensor or hardware is required for fault detection. The dataset utilized in this paper was obtained from an experimental IM drive test setup, on which intentional ISCFs can be created. The test results proved that the average fault detection rate is 95.8%, for an ISCF of 2-turns in a 104-turns phase winding.