Fault Diagnosis Method of Rolling Bearing Based on MSCNN-CAM

In the fault diagnosis of rolling bearings, traditional deep learning models such as convolutional neural network (CNN) may be difficult to extract multi-scale information from complex vibration signals effectively by using single-scale convolution kernel. Aiming at this problem, a fault diagnosis m...

Full description

Saved in:
Bibliographic Details
Published in2024 Prognostics and System Health Management Conference (PHM) pp. 328 - 333
Main Authors Wang, Hongchao, Lu, Shuaiwei, Yu, Li, Li, Simin, Guo, Zhiqiang, Du, Wenliao, Xue, Guoqing
Format Conference Proceeding
LanguageEnglish
Published IEEE 28.05.2024
Subjects
Anti-noise diagnosis
Channel attention mechanism
Convolution
Data mining
Fault diagnosis
Feature extraction
multi-scale convolution
Noise
Rolling bearing
Rolling bearings
Vibrations
Online AccessGet full text

Cover

More Information
Summary:In the fault diagnosis of rolling bearings, traditional deep learning models such as convolutional neural network (CNN) may be difficult to extract multi-scale information from complex vibration signals effectively by using single-scale convolution kernel. Aiming at this problem, a fault diagnosis method based on multi-scale attention mechanism is proposed. Multi-scale convolutional neural network (MSCNN) is introduced in this method, and information is extracted from different scales by using convolutional kernels of different sizes to prevent the loss of important features in the extraction process. In addition, a channel attention mechanism module (CAM) is introduced to selectively filter out irrelevant features after multi-scale information extraction. The effectiveness of this method is tested on the bearing data sets of CWRU and QPZZ-II. The experiment of adding noise to data sets shows that the model has high fault classification accuracy, robustness and strong anti-noise ability.
ISSN:2166-5656
DOI:10.1109/PHM61473.2024.00065