基于变分模态分解的旋转机械故障诊断研究
发布时间：2018-09-04 16:19
【摘要】：转子、滚动轴承等是工业生产机械中许多机器设备的重要零部件,对机器的正常运转有重要影响。故障振动信号一般具有非平稳、非线性、非高斯等特性,单一的方法难以提取故障振动信号中有效的特征信息。因此本文采用变分模态分解(Variational Mode Decomposition,VMD)这种全新的信号处理方法,并结合一些其他信号处理手段,对机械故障振动信号进行分析处理。主要研究内容如下:1、基于变分模态分解的转子故障时频分析方法针对转子故障诊断问题,使用一种基于变分模态分解的信号处理方法,该方法在获取分解分量的过程中通过迭代搜寻变分模型最优解来确定每个分量的频率中心及带宽,从而能够自适应地实现信号的频域剖分及各分量的有效分离,对各单分量信号进行希尔伯特变换即可得到瞬时频率和幅值信息。针对仿真信号和典型转子故障信号进行VMD方法和EMD方法的分析比较,以验证所提方法的有效性。仿真信号的分解结果表明,变分模态能够准确分离出信号中的固有模态分量且不存在模态混叠;转子故障实验信号的分析结果表明,所提方法能够有效提取出明显的故障特征,从而准确诊断出转子存在的故障。2、基于VMD和1.5维Teager能量谱的滚动轴承故障特征提取为准确提取滚动轴承故障信号中的故障特征,使用基于VMD和1.5维Teager能量谱的滚动轴承故障特征提取方法。故障特征提取过程:首先,对滚动轴承故障信号进行VMD分解得到一组分量,根据峭度-相关系数准则筛选分量进行信号重构;再次,对重构信号进行1.5维Teager能量谱分析,根据能量谱图的分析,提取出滚动轴承的内圈和滚动体故障特征。仿真和实验信号的分析验证了所提方法的有效性。与EEMD比较,采用VMD和1.5维Teager能量谱的分析方法更具有区分性,可以有效识别滚动轴承的故障特征。3、基于VMD、模糊熵和模糊C均值聚类的滚动轴承故障诊断使用一种基于VMD、模糊熵和模糊C均值聚类(FCM)算法的模式识别方法。首先采用VMD方法对信号进行分解,取相关性较大的分量组成初始特征向量矩阵;而后对初始特征向量矩阵求取模糊熵值,组成模糊熵值特征向量矩阵;最后将模糊熵值特征向量矩阵作为数据源输入FCM进行故障模式识别。将该方法应用于滚动轴承的故障模式识别,并与基于EMD和FCM的模式识别方法进行对比,验证了所提方法的有效性。
[Abstract]:Rotor, rolling bearing and so on are important parts of many machines and equipments in industrial production machinery, which have an important effect on the normal operation of the machine. The fault vibration signal usually has the characteristics of non-stationary, nonlinear and non-Gao Si, so it is difficult to extract the effective characteristic information from the fault vibration signal by a single method. Therefore, this paper adopts variational mode decomposition (Variational Mode Decomposition,VMD), a new signal processing method, and combines some other signal processing methods to analyze and process the vibration signals of mechanical faults. The main research contents are as follows: 1. The rotor fault time-frequency analysis method based on variational mode decomposition is used to solve the rotor fault diagnosis problem, and a signal processing method based on variational mode decomposition is used. In the process of obtaining decomposed components, the frequency center and bandwidth of each component can be determined by iterative search for the optimal solution of the variational model, so that the frequency domain partition and the effective separation of each component can be realized adaptively. The instantaneous frequency and amplitude information can be obtained by Hilbert transform for each single component signal. In order to verify the effectiveness of the proposed method, the VMD method and the EMD method are compared between the simulated signal and the typical rotor fault signal. The decomposition results of simulation signals show that the inherent modal components in the signals can be separated accurately and there is no modal aliasing, and the analysis results of the rotor fault signals show that the proposed method can extract the obvious fault characteristics effectively. The fault features of rolling bearings based on VMD and 1.5-D Teager energy spectrum are extracted to accurately extract fault features from fault signals of rolling bearings. The fault feature extraction method of rolling bearing based on VMD and 1.5 D Teager energy spectrum is used. Fault feature extraction process: first, the rolling bearing fault signal is decomposed into a group of components by VMD decomposition, and the signal is reconstructed according to the kurtosis correlation coefficient criterion. Thirdly, the 1.5 dimensional Teager energy spectrum analysis of the reconstructed signal is carried out. According to the analysis of energy spectrum, the fault characteristics of inner ring and rolling body of rolling bearing are extracted. The effectiveness of the proposed method is verified by simulation and experimental signal analysis. Compared with EEMD, VMD and 1.5-dimensional Teager energy spectrum analysis methods are more discriminative. Rolling bearing fault diagnosis based on VMD, fuzzy entropy and fuzzy C-means clustering uses a pattern recognition method based on VMD, fuzzy entropy and fuzzy C-means clustering (FCM) algorithm. Firstly, the signal is decomposed by VMD method, and the components with high correlation are used to form the initial eigenvector matrix, then the fuzzy entropy value is obtained for the initial eigenvector matrix, and the fuzzy entropy eigenvector matrix is formed. Finally, the fuzzy entropy eigenvector matrix is used as the input of FCM for fault pattern recognition. The method is applied to the fault pattern recognition of rolling bearing, and compared with the pattern recognition method based on EMD and FCM, the effectiveness of the proposed method is verified.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17
本文编号：2222700
[Abstract]:Rotor, rolling bearing and so on are important parts of many machines and equipments in industrial production machinery, which have an important effect on the normal operation of the machine. The fault vibration signal usually has the characteristics of non-stationary, nonlinear and non-Gao Si, so it is difficult to extract the effective characteristic information from the fault vibration signal by a single method. Therefore, this paper adopts variational mode decomposition (Variational Mode Decomposition,VMD), a new signal processing method, and combines some other signal processing methods to analyze and process the vibration signals of mechanical faults. The main research contents are as follows: 1. The rotor fault time-frequency analysis method based on variational mode decomposition is used to solve the rotor fault diagnosis problem, and a signal processing method based on variational mode decomposition is used. In the process of obtaining decomposed components, the frequency center and bandwidth of each component can be determined by iterative search for the optimal solution of the variational model, so that the frequency domain partition and the effective separation of each component can be realized adaptively. The instantaneous frequency and amplitude information can be obtained by Hilbert transform for each single component signal. In order to verify the effectiveness of the proposed method, the VMD method and the EMD method are compared between the simulated signal and the typical rotor fault signal. The decomposition results of simulation signals show that the inherent modal components in the signals can be separated accurately and there is no modal aliasing, and the analysis results of the rotor fault signals show that the proposed method can extract the obvious fault characteristics effectively. The fault features of rolling bearings based on VMD and 1.5-D Teager energy spectrum are extracted to accurately extract fault features from fault signals of rolling bearings. The fault feature extraction method of rolling bearing based on VMD and 1.5 D Teager energy spectrum is used. Fault feature extraction process: first, the rolling bearing fault signal is decomposed into a group of components by VMD decomposition, and the signal is reconstructed according to the kurtosis correlation coefficient criterion. Thirdly, the 1.5 dimensional Teager energy spectrum analysis of the reconstructed signal is carried out. According to the analysis of energy spectrum, the fault characteristics of inner ring and rolling body of rolling bearing are extracted. The effectiveness of the proposed method is verified by simulation and experimental signal analysis. Compared with EEMD, VMD and 1.5-dimensional Teager energy spectrum analysis methods are more discriminative. Rolling bearing fault diagnosis based on VMD, fuzzy entropy and fuzzy C-means clustering uses a pattern recognition method based on VMD, fuzzy entropy and fuzzy C-means clustering (FCM) algorithm. Firstly, the signal is decomposed by VMD method, and the components with high correlation are used to form the initial eigenvector matrix, then the fuzzy entropy value is obtained for the initial eigenvector matrix, and the fuzzy entropy eigenvector matrix is formed. Finally, the fuzzy entropy eigenvector matrix is used as the input of FCM for fault pattern recognition. The method is applied to the fault pattern recognition of rolling bearing, and compared with the pattern recognition method based on EMD and FCM, the effectiveness of the proposed method is verified.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17
【参考文献】
相关期刊论文 前10条
1 唐贵基;王晓龙;;IVMD融合奇异值差分谱的滚动轴承早期故障诊断[J];振动.测试与诊断;2016年04期
2 姚家驰;向阳;李胜杨;王帅;;基于VMD-ICA-CWT的内燃机噪声源识别方法[J];华中科技大学学报(自然科学版);2016年07期
3 马增强;李亚超;刘政;谷朝健;;基于变分模态分解和Teager能量算子的滚动轴承故障特征提取[J];振动与冲击;2016年13期
4 曹莹;段玉波;刘继承;;Hilbert-Huang变换中的模态混叠问题[J];振动.测试与诊断;2016年03期
5 王志搏;李富才;孟立立;张希;;基于ITD和模糊熵的滚动轴承智能诊断[J];噪声与振动控制;2016年01期
6 向玲;李媛媛;;经验小波变换在旋转机械故障诊断中的应用[J];动力工程学报;2015年12期
7 李学军;何能胜;何宽芳;何雷;;基于小波包近似熵和SVM的圆柱滚子轴承诊断[J];振动.测试与诊断;2015年06期
8 张亢;程军圣;;基于局部均值分解和峭度图的滚动轴承包络分析方法[J];航空动力学报;2015年12期
9 张淑清;胡永涛;李盼;包红燕;姜万录;钱磊;;基于MEMD互近似熵及FCM聚类的轴承故障诊断方法[J];中国机械工程;2015年19期
10 马文朋;张俊红;马梁;刘昱;贾晓杰;;改进的经验模式分解在机械故障诊断中的应用[J];振动.测试与诊断;2015年04期
相关博士学位论文 前2条
1 胡爱军;Hilbert-Huang变换在旋转机械振动信号分析中的应用研究[D];华北电力大学（河北）;2008年
2 程军圣;基于Hilbert-Huang变换的旋转机械故障诊断方法研究[D];湖南大学;2005年
相关硕士学位论文 前2条
1 王振威;基于变分模态分解的故障诊断方法研究[D];燕山大学;2015年
2 顾小军;面向旋转机械的支持向量机方法及智能故障诊断系统研究[D];浙江大学;2006年
，本文编号：2222700
本文链接：https://www.wllwen.com/jixiegongchenglunwen/2222700.html
