Diagnosis of Stator Interturn Fault in SPMSM Using Machine Learning

摘要

Permanent magnet synchronous motors (PMSM) are becoming popular in several high-power applications, such as industrial drives, vehicular propulsion, robotics, etc., due to benefits such as high efficiency, strong torque, and compact design. Though dependable, these motors operate in harsh environments and can experience electrical, mechanical, and magnetic faults. Among the various faults, one of the most common in PMSM is the stator winding inter-turn short circuit, which falls under electrical faults and should be ascertained at the beginning level, or else leads to catastrophic failure. In this work, a finite element analysis on stator inter-turn fault is carried out on a 550 W, 220 V surface-mounted PMSM using the ANSYS Maxwell tool, and elements such as torque, current, and flux distribution are investigated. Inter-turn fault severities such as 3%, 6%, and 12.5% are incorporated in the finite element model and considered for examination. An experimental analysis is made with a setup comprising 1 kW, 48 V PMSM with inter-turn faults, inverter, controller, and electrical loading arrangement. Motor current data for normal and faulty PMSM are acquired from finite element model and experimental setup and imported to the machine learning toolbox in MATLAB for analysis. Decision Tree and Random Forest achieved maximum accuracy for the finite element model whereas Neural Network and Decision Tree attained maximum accuracy for the experimental set-up.