机械科学与技术2025，Vol.44Issue(2)：262-270,9.DOI:10.13433/j.cnki.1003-8728.20230188
双通道交叉密集连接的滚动轴承故障诊断
Fault Diagnosis of Rolling Bearings with Two-channel Cross-dense Connection
摘要
Abstract
To address the problems for traditional convolutional networks having inadequate learning capability of critical faults and low diagnostic accuracy,a dual-channel cross-densely connected fault diagnosis model(DCCNN)incorporating parallel ECA modules is proposed,which builds a dual-channel structure on the basis of a densely connected network,designs a multi-convolutional residual module and a multi-scale densely connected network to extract fault features and realize the interaction and integration of fault information.The network is embedded with a parallel channel attention module,which is reweighted by the channel attention mechanism to form multi-weighted features that can suppress the interference of noise and irrelevant signals from multiple perspectives.Finally,the training is conducted on the bearing data and gear data from Case Western Reserve University;the experimental results show that the accuracy of bearing fault recognition is 99.31%,which verifies that the model has adaptive diagnostic capability;the network model also maintains a good diagnostic performance in a noisy and loaded environment,and the proposed method has good generalization and noise immunity compared with other methods.关键词
密集连接网络/注意力机制/故障诊断/残差模块Key words
dense connected network/mechanism of attention/fault diagnosis/residual module分类
计算机与自动化引用本文复制引用
王庆荣,王媛,朱昌锋,周禹潼..双通道交叉密集连接的滚动轴承故障诊断[J].机械科学与技术,2025,44(2):262-270,9.基金项目
国家自然科学基金项目(71961016)与甘肃省自然科学基金项目(20JR10RA214) （71961016）
