Intelligent Fault Diagnosis Method of Rolling Bearings Based on Transfer Residual Swin Transformer with Shifted Windows

作     者：Haomiao Wang Jinxi Wang Qingmei Sui Faye Zhang Yibin Li Mingshun Jiang Phanasindh Paitekul 

作者机构：The Institute of Marine Science and TechnologyShandong UniversityQingdao26623China The School of Control Sciences and EngineeringShandong UniversityJinan250061China Thailand Institute of Scientific and Technological ResearchAmphoe Khlong LuangPathum Thani12120Thailand 

出 版 物：《Structural Durability & Health Monitoring》 (结构耐久性与健康监测（英文）)

年 卷 期：2024年第18卷第2期

页      面：91-110页

核心收录：

学科分类：0710[理学-生物学] 08[工学] 080203[工学-机械设计及理论] 0802[工学-机械工程] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 080201[工学-机械制造及其自动化] 

基　　金：supported in part by the National Natural Science Foundation of China(General Program)under Grants 62073193 and 61873333 in part by the National Key Research and Development Project(General Program)under Grant 2020YFE0204900 in part by the Key Research and Development Plan of Shandong Province(General Program)under Grant 2021CXGC010204 

主　　题：Rolling bearing fault diagnosis transformer self-attention mechanism 

摘      要：Due to their robust learning and expression ability for complex features,the deep learning(DL)model plays a vital role in bearing fault ***,since there are fewer labeled samples in fault diagnosis,the depth of DL models in fault diagnosis is generally shallower than that of DL models in other fields,which limits the diagnostic *** solve this problem,a novel transfer residual Swin Transformer(RST)is proposed for rolling bearings in this *** has 24 residual self-attention layers,which use the hierarchical design and the shifted window-based residual *** with transfer learning techniques,the transfer RST model uses pre-trained parameters from ImageNet.A new end-to-end method for fault diagnosis based on deep transfer RST is ***,wavelet transform transforms the vibration signal into a wavelet time-frequency *** signal’s time-frequency domain representation can be represented ***,the wavelet time-frequency diagram is the input of the RST model to obtain the fault ***,our method is verified on public and self-built *** results show the superior performance of our method by comparing it with a shallow neural network.