Adaptive adjustment graph representation learning method for rotating machinery fault diagnosis under noisy signals  

作　　者：Lei WANG Peijie YOU Xin ZHANG Li JIANG Yibing LI 

机构地区：[1]School of Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan 430070,China [2]Mechanical and Aerospace Engineering,The Hong Kong University of Science and Technology,Hong Kong 999077,China

出　　处：《Frontiers of Mechanical Engineering》2025年第1期147-162,共16页机械工程前沿(英文版)

基　　金：financially supported by the National Natural Science Foundation of China(Grant No.51905396).

摘　　要：Recently,intelligent fault diagnosis methods have been employed in the condition monitoring of rotating machinery.Among them,graph neural networks are emerging as a new feature extraction tool that can mine the relationship characteristics between samples.However,many existing graph construction methods suffer from structural redundancy or missing node relationships,thus limiting the diagnosis accuracy of the models in practice.In this paper,an adaptive adjustment k-nearest neighbor graph-driven dynamic-weighted graph attention network(AAKNN-DWGAT)is proposed to address this problem.First,time-domain signals are transformed into frequency-domain features by using fast Fourier transformation.Subsequently,a frequency similarity evaluation method based on dynamic frequency warping is proposed,which enables the conversion of distance measurements into a frequency similarity matrix(FSM).Then,an adaptive edge construction operation is conducted on the basis of FSM,whereby the effective domain is captured for each node using an adaptive edge adjustment method,generating an AAKNN graph(AAKNNG).Next,the constructed AAKNNG is fed into a dynamic-weighted graph attention network(DWGAT)to extract the fault features of nodes layer by layer.In particular,the proposed DWGAT employs a dynamic-weighted strategy that can update the edge weight periodically using high-level output features,thereby eliminating the adverse impacts caused by noisy signals.Finally,the model outputs fault diagnosis results through a softmax classifier.Two case studies verified the effectiveness and the superiority of the proposed method compared with other graph neural networks and graph construction methods.

关 键 词：rotating machinery fault diagnosis graph neural network adaptive adjustment 

分 类 号：O15[理学—数学]