重庆理工大学学报2024，Vol.38Issue(11)：133-140,8.DOI:10.3969/j.issn.1674-8425(z).2024.06.016
利用EEMD和深度置信网络的滚动轴承故障诊断方法
Rolling bearing fault diagnosis method based on EEMD and deep belief network
摘要
Abstract
Combined with the nonlinear characteristics of vibration signals of rolling bearings,in order to restore the nonlinear dynamic characteristics of signals,the fault diagnosis method of rolling bearings based on EEMD and Deep Belief Network (DBN)is studied,and the diagnosis and prediction of different fault types and damage degrees of rolling bearings are realized.First,after EEMD is used to decompose the signal,the signal reconstruction is completed by combining the correlation coefficient and kurtosis analysis.On this basis,the acceleration-velocity matrix is constructed as the input of the DBN model to realize the diagnosis and prediction of different fault types and damage degrees of rolling bearings.In the process,the model generalization ability,layer feature extraction ability,model analysis results and analysis efficiency are compared and analyzed.The original data,acceleration-velocity matrix and phase diagram are employed as the input of DBN model to analyze and compare the diagnostic effect.The comparative analysis of different network models of SVM,XGboost,ResNet and DBN is made.Our results show using EEMD for signal processing and constructing acceleration-velocity matrix as the input of DBN model effectively achieve the diagnosis of different fault types and fault damage degrees of rolling bearings with the average diagnostic accuracy reaching 97.23％.This model may provide some insightful reference for the engineering application of bearing condition monitoring and intelligent fault diagnostic technology.关键词
深度置信网络/滚动轴承/故障诊断/状态评价Key words
deep belief network/rolling bearings/fault diagnosis/state evaluation分类
信息技术与安全科学引用本文复制引用
郑鑫辉,马超,王少红,徐小力..利用EEMD和深度置信网络的滚动轴承故障诊断方法[J].重庆理工大学学报,2024,38(11):133-140,8.基金项目
国家自然科学基金项目(51975058) （51975058）
北京信息科技大学勤信人才项目(QXTCPC202120) （QXTCPC202120）
北京市科学技术概念验证项目(20220481077) （20220481077）
