Finite Element-Based Inter-turn Fault Analysis in Closed Loop Induction Motor Drive Using Machine Learning

This work utilizes machine learning techniques to investigate the finite element method-constituted stator phase winding inter-turn short fault in an induction motor (IM) operated in a closed-loop speed-controlled approach. One of the biggest challenges for electrical machine researchers is diagnosing inter-turn problems that involve fewer than 5% of turns. Due to the machine's varying speed and fluctuating load, diagnosing such defects becomes more complicated when the machine is regulated from a PWM inverter in a closed-loop speed-controlled technique. In this framework, ANSYS Maxwell finite element tool is utilized to simulate a 7.5 kW IM with several inter-turn fault situations, such as 30T (12.5%), 15T (6.25%), IOT (4.2%), and 5T (2%) and is integrated with ANSYS Simplorer for power source and loading configurations. In order to conduct a realistic simulation, a closed-loop controlled PWM inverter is designed in MATLAB and cosimulated with ANSYS Simplorer. Subsequently, machine learning methods are utilized to analyze the stator current data collected for different levels of fault severity. The outcome shows that ensemble bagged trees achieved a maximum accuracy of 90% across different algorithms. The number of faulty samples in the stator current histogram increases with the severity of the fault.