Bearing fault diagnosis method based on dual-channel feature fusion

Intelligent diagnosis method based on convolution neural network (CNN) has been widely used in bearing fault diagnosis. However, most existing diagnostic models rely on single-source information inputs, limiting their accuracy and reliability. To solve this limitation, this paper presents a rolling...

Full description

Saved in:
Bibliographic Details
Main Authors: ZHANG Xiaoning, ZHU Huilong, XIN Liang, YANG Muchen, WANG Hao
Format: Article
Language:zho
Published: Editorial Department of Electric Drive for Locomotives 2023-11-01
Series:机车电传动
Subjects:
Online Access:http://edl.csrzic.com/thesisDetails#10.13890/j.issn.1000-128X.2023.06.005
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Intelligent diagnosis method based on convolution neural network (CNN) has been widely used in bearing fault diagnosis. However, most existing diagnostic models rely on single-source information inputs, limiting their accuracy and reliability. To solve this limitation, this paper presents a rolling bearing fault diagnosis method based on dual-channel feature fusion. Firstly, the time-frequency analysis diagrams of rolling bearing vibration signals were constructed by using multiple Q-factor continuous Gabor wavelet transform (CMQGWT) and fast spectral coherence (Fast-SC), respectively. Subsequently, a CNN model with dual input channels was constructed, allowing for the fusion of deep time-frequency features extracted from each channel into a new feature at a feature fusion layer. Finally, the diagnosis results were output using a classifier. Through classification and recognition experiments involving single and compound faults in rolling bearings for high-speed trains, compared with the CNN model with a single input channel, the proposed model demonstrates superior diagnostic accuracy and robustness.
ISSN:1000-128X