Fault diagnosis of low-speed and heavy-load bearings based on STFT-SE-ResNet18 model
ZHENG Xin
SUN Xianbin
YIN Gang
KONG Liya
Abstract：[Objective]Aiming at the problems of weak fault features and difficult fault feature extraction in vibration signals of low-speed and heavy-load rolling bearings,a fault diagnosis method based on ResNet18 residual network with squeeze-and-excitation(SE)attention mechanism was proposed.[Methods]Firstly,the vibration signals collected by the sensor were converted into two-dimensional time-frequency diagrams via short-time Fourier transform(STFT).By virtue of its advantages in fault feature extraction,noise resistance and visualization,more fault feature information was captured.Then,the two-dimensional time-frequency diagrams were input into the improved STFT-SE-ResNet18 model.With the help of the SE attention mechanism,adaptive channel weights were learned,which enabled the model to pay more attention to useful channel information and improved the learning ability and recognition accuracy of the model.Finally,comparative experiments were conducted between this model and other network models.[Results]The results show that the proposed model has high fault diagnosis accuracy and strong anti-interference ability under different working conditions,with remarkable fault diagnosis effect,demonstrating high superiority and great application potential.
Keywords：Low-speed heavy-load bearingSTFTAttention mechanismResidual networkFault diagnosis
Publication Date：2025-11-15
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：10( 167-176 )