燕山大学学报2024，Vol.48Issue(1)：39-47,9.DOI:10.3969/j.issn.1007-791X.2024.01.005
基于注意力机制和深度残差网络的滚动轴承故障诊断
Rolling bearing fault diagnosis based on attention mechanism and depth residual network
摘要
Abstract
Aiming at the problems of insufficient feature extraction ability and low diagnosis accuracy of existing rolling bearing diagnosis models,a fault diagnosis method combining attention mechanism and one-dimensional depth residual network was proposed.Firstly,the residual structure was introduced to prevent the performance degradation of the deep network,and then the feature extraction capability of the network was improved by combining the attention mechanism.Finally,the original rolling bearing vibration signals were used to train the fault feature classifier.In this paper,a small sample transfer learning framework was adopted for fault diagnosis in variable working conditions.The result of two open source experimental platforms shows that this method can effectively improve the accuracy of rolling bearing fault diagnosis and provide a theoretical reference for practical applications.关键词
滚动轴承/注意力机制/残差网络/特征提取/迁移学习Key words
rolling bearing/attention mechanism/residual network/feature extraction/transfer learning分类
机械制造引用本文复制引用
时培明,吴术平,于越,张宇,许学方..基于注意力机制和深度残差网络的滚动轴承故障诊断[J].燕山大学学报,2024,48(1):39-47,9.基金项目
河北省自然科学基金-青年基金资助项目(E2022203093) （E2022203093）
秦皇岛市科学技术研究与发展计划项目(202101A345) （202101A345）
