CEEMDAN结合排列熵在滚动轴承故障特征分析中的应用  

Application of CEEMDAN combined with permutation entropy to fault characteristics analysis of rolling bearing

作　　者：朱敏 段志善[1] 郭保良 Zhu Min;Duan Zhishan;Guo Baoliang(College of Mechanical and Electrical Engineering, Xi’an University of Architecture andTechnology, Xi'an 710055 , China)

机构地区：[1]西安建筑科技大学机电工程学院,西安710055

出　　处：《现代制造工程》2019年第4期151-154,74,共5页Modern Manufacturing Engineering

基　　金：国家自然科学基金青年科学基金项目(51105292);教育部博士点基金项目(20126120110009);陕西省科技攻关项目(2013K07-09);陕西省教育厅专项基金项目(2013JK1032)

摘　　要：提出了一种新的信号分类识别的方法,首先将信号进行CEEMDAN分解得到多个固有模态分量,对每个分量求排列熵值。通过大量的数据对比分析,选择出在不同工况下,排列熵值差异最大的几阶模态分量,然后将这几组排列熵值组成特征向量,利用RBF人工神经网络实现对信号的分类识别。最后通过数据验证,比较了三种不同的分类识别方法,证明了新方法的有效性。A new method of signal classification and recognition is proposed. First, the signal is decomposed by CEEMDAN to get multiple intrinsic modal components,find the permutation entropy of each component. By comparison and analysis of a large number of data, several modal components with maximum entropy difference are selected under different working conditions. Then the entropy value of these groups is composed of feature vectors, and the classification and recognition of signals are realized by RBF artificial neural network. Finally, through data verification, three different classification methods are compared, and the effectiveness of the new method is proved.

关 键 词：经验模态分解 排列熵 模态分量 滚动轴承 人工神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]