RT Journal Article
SR Electronic
A1 Zhang, Mei
A1 Sun, Zilong
A1 Zheng, Wenchao
T1 Fault Diagnosis of Electric Motor Rotor Systems Based on Feature Extraction and CNN-BiGRU-Attention
JF Manufacturing Technology Journal
YR 2025
VO 25
IS 4
SP 559
OP 568
DO 10.21062/mft.2025.048
UL https://journalmt.com/artkey/mft-202504-0003.php
AB To enhance the accuracy of fault diagnosis (FD) in motor rotor systems, this study introduces a novel method that leverages feature extraction (FE) combined with a CNN-BiGRU-Attention deep learning model. Initially, the time-domain features of the vibration signals are extracted using Variational Mode Decomposition (VMD), which also effectively denoises the data. Subsequently, the frequency-domain features of the vibration signals are extracted via Fast Fourier Transform (FFT). The aggregated features are then fed into the CNN-BiGRU-Attention model to perform fault classification. In this model, the Convolutional Neural Network (CNN) module extracts local spatial features, the Bidirectional Gated Recurrent Unit (BiGRU) module models the temporal dependencies, and the Attention mechanism enhances the focus on critical fault information, thereby improving the model's classification performance. Experimental results demonstrate that the proposed FD method achieves an accuracy of 99.58%. Compared to other commonly used models, the performance metrics of our model show significant advantages and superior performance.