Finite Element-based Dynamic Eccentricity Fault Analysis in Switched Reluctance Motor Using Machine Learning Approach

Switched Reluctance Motor (SRM) is becoming popular in several fields, such as electric vehicles, renewable energy-dependent machines, domestic appliances, etc. Among the faults in SRM, bearing and eccentricity-related failures are predominant and must be addressed at the incipient stage. In this work, dynamic eccentricity fault is considered for investigation, which occurs due to the cylinder’s displacement with the stator and rotor rotational axis being fixed at the origin. A 550W, 220V SRM in ANSYS Maxwell Finite Element tool is considered for the analysis, and various dynamic eccentric fault severities such as 5%, 10%, and 25% are incorporated. Magnetic flux density over the machine and airgap increase when the fault severity increases. Additional harmonics are induced in the spatial FFT spectrum of airgap flux due to the fault. Stator current and radial airgap flux density data of SRM under healthy and faulty conditions in finite element model is exported to the machine learning toolbox in MATLAB. Machine learning analysis shows that the ensemble boosted trees algorithm has an accuracy of 70.4% for radial flux density and 57% for stator current when used as a parameter for fault detection.