摘要
针对滚动轴承振动信号特征提取有效性差以及传统BP神经网络识别率低的问题,提出一种将希尔伯特-黄变换(HHT)、列文伯格-马夸尔特(LM)算法和BP神经网络相结合的滚动轴承故障识别方法。该方法首先利用经验模态分解(EMD)、相关函数得到几个能充分表征原始信号信息的IMF分量,计算它们的能量特征,构成第一种特征向量组。将这几个IMF分量进行Hilbert变换,得到Hilbert边际谱,将边际谱区域变化能量特征作为第二种特征向量组。将两种特征向量组合在一起输入到LM算法优化的BP神经网络中进行训练和测试,进而实现故障的分类。结果表明,该方法能有效提取出轴承的故障特征信息,并且能准确的识别出不同的故障。
To solve the poor validity of rolling bearing vibration signal feature extraction and the low recognition rate of traditional BP neural network,a fault identification method for rolling bearings based on Hilbert-Huang Transform(HHT),Levenberg-Marquardt(LM)algorithm and BP neural network is proposed.Firstly,the empirical modal decomposition(EMD)and correlation function are used to obtain several IMF components which can fully represent the original signal information,and their energy characteristics are calculated to form the first eigenvector group.IMF components are subjected to Hilbert transform to obtain the Hilbert marginal spectrum,and the marginal spectral region variation energy feature is taken as the second eigenvector group.The two feature vectors are combined and input into the BP neural network optimized by LM algorithm for training and testing,so as to realize fault classification.The results show that the method can effectively extract the fault characteristic information of the bearing and can accurately identify different faults.
作者
胡泽
张智博
王晓杰
吴雨宸
谢心蕊
Hu Ze(Southwest Petroleum University,Chengdu 610500,Sichuan)
出处
《电动工具》
2020年第1期11-18,共8页
Electric Tool
作者简介
胡泽(1966—),男,博士,教授,主要研究方向为油气测控工程、计算机控制、自动化等。
