CNN-Based Location Detection of Static Eccentricity in 5-X Resolver
Mahdi Emadaleslami, Farid Tootoonchian · 2023
According to the significant effects of faulty resolver condition and rotor angle error on Permanent Magnet Synchronous Motor (PMSM) drive system stability, the diagnosis of the fault occurrence and, more critically, the fault location is crucial. Among the resolver’s faults, static eccentricity has more probability and severity. So, the following study focuses on proposing a Convolutional Neural Networks (CNN)-based static eccentricity location diagnosis under stator slots. Accordingly, using Finite Element Analysis (FEA), various locations and severity of static eccentricity are generated. The Fast Fourier Transform (FFT) and Frequency Occurrence Plot (FOP) are regarded as the data preprocessing stages, applying on resolver signal voltages. Then, in the training stage, various CNNs are trained to obtain a highly accurate network. The feasibility of the suggested CNN in fault location diagnosis is shown by the confusion matrix, accuracy metric, and F1-score criterion.