摘要
为提高复杂工业过程故障诊断的精度和泛化能力,提出了一种基于变分模态分解(VMD)、局部均值分解(LMD)和卷积神经网络(CNN)的集合型故障诊断方法。首先,利用VMD对采集的数据进行降噪处理;然后,运用LMD将降噪后信号分解为一系列的PF分量;最后,通过相关系数选择合适的PF分量,输入卷积神经网络进行特征提取和分类。通过2种轴承的故障诊断试验验证了VMD-LMD-CNN方法的有效性,与其他算法的对比表明VMD-LMD-CNN方法具有更高的故障诊断精度。
To improve accuracy and generalization of fault diagnosis in complex industrial processes,an ensemble fault diagnosis method is proposed based on variational mode decomposition(VMD),local mean decomposition(LMD)and convolutional neural network(CNN).Firstly,VMD is used to reduce noise of collected data.Then,LMD is used to decompose denoised signal into a series of PF components.Finally,the appropriate PF components are selected by correlation coefficients as input of CNN for feature extraction and classification.The effectiveness of VMD-LMD-CNN method is verified by two kinds of bearing fault diagnosis experiments,and the VMD-LMD-CNN method has higher fault diagnosis accuracy compared with other algorithms.
作者
吴东升
杨青
张继云
纪振平
WU Dongsheng;YANG Qing;ZHANG Jiyun;JI Zhenping(School of Automation and Electrical Engineering,Shenyang Ligong University,Shenyang 110159,China)
出处
《轴承》
北大核心
2020年第10期57-63,共7页
Bearing
基金
国家自然科学基金项目(61273178)
辽宁省自然科学基金指导计划(20180550801)
辽宁省教育厅科学研究项目计划(LG201917)。
关键词
滚动轴承
故障诊断
变分模态分解
局部均值分解
卷积神经网络
rolling bearing
fault diagnosis
variational mode decomposition
local mean decomposition
convolutional neural network
