Track and hunt metaheuristic based deep neural network based fault diagnosis model for the voltage source inverter under varying load conditions

Abstract

Induction motors act as the pillars for industries and are preferable in various applications due to the characteristics of stability and sturdiness. The stability level should be efficiently maintained in order to monitor and analyze the faults that occur in three-phase Voltage Source Inverter. In this research, fault prediction occurs in the inverter under various load conditions is performed using the proposed track and hunt meta-heuristic-based deep neural network. The open circuit fault is generated in the three-phase voltage source inverter transistor under varying load conditions and is determined using the proposed track and hunt optimization. The proposed track and hunt meta-heuristic is developed by hybridizing the characteristics of hunting and tracking. After the data acquisition, the features are extracted from the collected faulty data in the form of three-phase current, voltage, torque, and speed. The Deep Neural Network classifier is trained using the obtained features and the faulty switches are identified through the average value of the three-phase current. The three-phase induction motor is connected as a load for the voltage source inverter with reference to the frequency variation. The prediction performance of track and hunt- deep learning classifier reveals that the percentage improvement of 10–15% is acquired for the developed method.