基于卷积图神经网络的滚动轴承故障诊断算法  

Fault Diagnosis Algorithm for Rolling Bearing Based on Convolution Graph Neural Network

作　　者：丁汕汕 陈仁文[1] 刘飞[1] 刘昊 Ding Shanshan;Chen Renwen;Liu Fei;Liu Hao(State Key Laboratory of Mechanical Structure Mechanics and Control,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学机械结构力学及控制国家重点实验室,南京210016

出　　处：《单片机与嵌入式系统应用》2022年第4期38-41,51,共5页Microcontrollers & Embedded Systems

摘　　要：针对常规深度学习方法在直接处理一维时域振动信号进行故障诊断时诊断准确度较低的问题,提出了一种基于一维卷积神经网络(Convolution Neural Network,CNN)与图神经网络(Graph Neural Network,GNN)的滚动轴承故障诊断算法(CGNN)。首先通过一维卷积层对振动信号做自适应滤波与数据压缩预处理,然后将预处理后的一维特征数据转换为图结构数据,最后使用三层图神经网络来进行滚动轴承的故障诊断。在凯斯西储大学滚动轴承数据集(CWRU)上开展实验验证,结果表明,CGNN在各个工况下都能具有90%以上的故障诊断准确率。Aiming at the problem of low diagnostic accuracy of conventional deep learning methods when they directly processing one-dimensional time-domain vibration signals for rolling bearing fault diagnosis,a rolling bearing fault diagnosis algorithm based on one-dimensional convolutional neural network(CNN)and graph neural network(GNN)is proposed.First,perform adaptive data filtering and data compression preprocessing of the vibration signal through the one-dimensional convolutional neural network layers,then convert the preprocessed one-dimensional feature data into graph structure data,and in the end a three-layer graph neural network is used for fault diagnosis of rolling bearings.The experimental verification is carried out on the Case Western Reserve University Fault Rolling Bearing Data Set(CWRU).The results show that CGNN can have more than 90%accuracy of fault diagnosis under all operating conditions.

关 键 词：图神经网络 卷积神经网络 深度学习 滚动轴承 故障诊断 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]