Investigation of Dynamic Eccentricity in Switched Reluctance Motors: A Fusion of Finite Element Analysis and Machine Learning

Switched reluctance motors (SRM) are widely considered for their reliability and fault-tolerant capability. Dynamic eccentricity fault is one of the major mechanical faults that occur in an SRM which induces electrical imbalances leading to increased losses and reduced performance. This work combines the finite element SRM model with machine learning to investigate dynamic eccentricity faults. A 550 W, 220 V finite element SRM model in ANSYS Maxwell is considered for examination. Eccentric fault severities ranging from 10% to 40% are incorporated in the model and considered for analysis. A comparison of motor parameters involving winding currents and airgap flux is conducted between normal and eccentric SRM. The main contribution of this work is to detect the presence of dynamic eccentricity at the beginning stage which causes a slight asymmetry in the distribution of flux lines in the SRM compared to a healthy SRM. Data on motor current and airgap flux density under both healthy and dynamic eccentric conditions are acquired from the machine model. Subsequently, they are subjected to examination using a machine learning tool in MATLAB, and of all algorithms, the Bagged Ensemble obtained an accuracy of 88.5% based on stator current and 66.3% based on air gap flux density.

I. Introduction

The Switched Reluctance Motor (SRM) operates on the principle of converting reluctance torque into mechanical power. Due to the salient structure of stator and rotor poles, SRM provides high output torque pulsed in nature due to the alignment tendency of poles [1]. Though SRM is reliable and fault tolerant, it is sensitive to various faults as it operates in harsh industrial and traction environmental conditions [2]. Eccentricity is one of the major faults in SRM, which is mechanically based and constitutes nearly 10% of mechanical faults [3].