摘要
轴承的健康状态与设备安全可靠运行息息相关,在现代制造系统中,轴承剩余使用寿命(Remaining Useful Life,RUL)预测已成为研究热点。文中提出了通过数据增强来提取轴承衰退特征并利用卷积神经网络(Convolutional Neural Network,CNN)进行轴承RUL预测的方法。该方法首先将均方根、峰值和峰度作为时域特征,频谱分区求和(FSPS)特征作为频域特征,经过数据增强将18维时域和频域特征增加到108维,从而得到全面反应轴承退化过程的信息。通过搭建卷积神经网络(CNN),利用CNN处理高维特征的能力实现轴承RUL预测。最后,试验结果证明文中所提方法相对DNN模型、SVM模型具有更高的预测精度。
The health of bearing is closely related to the safe and reliable operation of the equipment.In the modern manufacturing systems,prediction of the bearing’s remaining useful life(RUL)is a research focus.This article proposes a method to extract the bearing’s decay features through data enhancement and adopt convolutional neural network(CNN)to predict the bearing’s RUL.Firstly,the root mean square,the peak value and the kurtosis serve as the time-domain features,and the frequency spectrum partition summation(FSPS)serves as the frequency-domain feature.After data enhancement,the 18-dimensional timedomain and frequency-domain features increase to 108-dimensional features,in order to obtain comprehensive information about the bearing’s degradation process.Based on the convolutional neural network(CNN),the ability of processing the high-dimensional features ensures the prediction of the bearing’s RUL.Finally,the experimental results prove that this method has higher prediction accuracy compared with the DNN model and the SVM model.
作者
邹旺
吉畅
陈伟兴
郑凯
ZOU Wang;JI Chang;CHEN Wei-xing;ZHENG Kai(Engineering Training Center,Liupanshui Normal University,Liupanshui 553004;School of Physics and Electrical Engineering,Liupanshui Normal University,Liupanshui 553004;School of Mining and Civil Engineering,Liupanshui Normal University,Liupanshui 553004;Hangzhou Eohi Cyber System Co.,Ltd.,Hangzhou 310000)
出处
《机械设计》
CSCD
北大核心
2021年第8期84-90,共7页
Journal of Machine Design
基金
贵州省教育厅青年人才成长项目(黔教合KY字[2020]121,黔教合KY字[2020]114)。
关键词
数据增强
卷积神经网络
特征提取
剩余寿命预测
data enhancement
convolutional neural network
feature extraction
prediction of remaining useful life
作者简介
邹旺(1993—),男,讲师,硕士,研究方向:智能制造、机械健康状态管理。E-mail:781023098@qq.com;通信作者:郑凯(1992—),男,讲师,硕士,研究方向:故障诊断。E-mail:442093185@qq.com。
