Implementation of Machine Learning to Analyze Static Eccentricity Fault in SPMSM using FEM

The high performance of Surface Mount Permanent Magnet Synchronous Motors (SPMSM) considers an excellent choice for using them in industries. Among the faults encountered in SPMSM, eccentricity faults account for nearly 10%. Diagnosing the eccentricity fault at an early stage is challenging since the fault would not significantly alter the terminal parameters of motor until the fault intensity was extremely high. This paper uses the finite element approach to study various motor parameters like phase current, flux lines, and flux density of a 0.55 kW, 220 V SPMSM during normal and static eccentricity fault conditions in the ANSYS Maxwell tool. The study is conducted for 10% to 25% of eccentricity in the rotor of SPMSM. Due to the eccentricity fault, a slight deformity is observed in the flux lines distribution and radial airgap flux compared to the healthy motor. The severity of eccentricity is observed in the FFT (spatial) spectrum of airgap flux (radial component) as redundant harmonic noise. Data pertaining to motor current and flux density in airgap are extracted from finite element model and diagnosed using machine learning tool in MATLAB. Among the algorithms, the medium k – NN attained 58.4% accuracy for estimating the static eccentricity fault based on airgap flux density, and Cubic SVM achieved 50.1% accuracy based on stator current.