基于经验模态分解的转子故障信号熵特征提取研究
发布时间：2018-09-03 06:54
【摘要】：随着传感器技术、测试技术及信号分析的发展，故障诊断技术有了较大提高，其主要研究内容包括四个方面：信号采集、信号特征提取、故障诊断、信息融合。其中，信号特征提取是指故障诊断过程中获取与系统状态相关性较大的敏感特征的特征因子提取技术。特征提取的正确与否将直接影响故障诊断结果的准确性。针对传统信号分析方法难以准确描述转子振动信号非平稳特性以及信号特征难以定量评价的问题，本研究提出一种经验模态分解和信息熵相结合的方法，对转子故障振动信号特征提取和定量评价方法进行研究。 针对转子实验台采集到的数据，开展的主要内容及研究成果如下： 1)以轴承-转子系统模型为基础，研究旋转机械常见的故障类型及机理。针对转子故障信号的特点，，设计了中值滤波和小波消噪结合的故障信号预处理滤波器，为特征提取提供原始数据。重点研究了本文的信号处理方法—经验模态分解法(EMD)，分析该方法的实质及特点。针对EMD分解过程中存在的迭代次数难以确定及端点效应问题，提出了“能量算法”和“相似信号平移算法”相结合的算法。通过对仿真实验信号的分析表明，该算法能够对以上问题准确、有效地解决。 2)分析信息熵在故障诊断中的研究现状，对时域奇异谱熵、频域功率谱熵、时-频域小波能谱熵和小波空间状态谱熵进行了比较系统的研究和分析，并基于LabVIEW编写信息熵算法程序。 3)提出了一种基于经验模态分解法的转子故障信号熵特征提取方法。提出“能量法”与“相关系数法”相结合的算法选取出包含主要故障信息的分量。实验表明，该方法能准确地提取出后续研究所需的特征数据。再计算出包含主要故障的分量的四种信息熵值，即时域奇异谱熵、频域功率谱熵、时-频域小波能谱熵和小波空间状态谱熵。实验信号的分析结果表明，该方法能够较好的实现对转子系统故障信号的量化特征提取，所提取出的特征集合具有能够使典型故障特征量之间存在显著差异的性能。 4)基于信息融合的思想，计算EMD分解后所选取出主要故障分量的奇异谱熵、功率谱熵及小波能谱熵，提出建立故障信号信息熵状态空间分布图。实验结果表明，该状态空间模型能够直观、准确地实现转子故障的模式识别。 5)以LabVIEW软件为平台，建立了典型转子故障信号测试系统。该系统实现了对采集到的原始信号滤波处理及对原始信号和滤波后信号的频谱分析和轴心轨迹的分析。 本文以转子故障信号量化特征提取为目的，针对涉及到的数字信号处理、信息论以及智能诊断理论等内容，本文的研究工作值得进一步深入。
[Abstract]:With the development of sensor technology, test technology and signal analysis, fault diagnosis technology has been greatly improved. The main research contents include four aspects: signal acquisition, signal feature extraction, fault diagnosis, information fusion. In the process of fault diagnosis, signal feature extraction refers to the feature factor extraction technique, which can obtain sensitive features which are highly correlated with the system state. Whether the feature extraction is correct or not will directly affect the accuracy of fault diagnosis results. The traditional signal analysis method is difficult to accurately describe the non-stationary characteristics of rotor vibration signals and it is difficult to quantitatively evaluate the signal characteristics. In this paper, an empirical mode decomposition method combined with information entropy is proposed. The method of feature extraction and quantitative evaluation of rotor fault vibration signal is studied. The main contents and research results are as follows: 1) based on the bearing-rotor system model, the common fault types and mechanism of rotating machinery are studied. According to the characteristics of rotor fault signal, a fault signal preprocessing filter combining median filter and wavelet de-noising is designed, which provides the original data for feature extraction. In this paper, the essence and characteristics of the signal processing method-empirical mode decomposition method (EMD),) are studied. In view of the difficulty of determining the number of iterations and the endpoints effect in the process of EMD decomposition, a combination of "energy algorithm" and "similar signal translation algorithm" is proposed. The analysis of simulation signals shows that the algorithm can solve the above problems accurately and effectively. 2) the research status of information entropy in fault diagnosis is analyzed, and the singular spectrum entropy in time domain and power spectrum entropy in frequency domain are analyzed. The wavelet energy spectrum entropy in time-frequency domain and the state spectrum entropy in wavelet space are studied and analyzed systematically. An information entropy algorithm program based on LabVIEW is developed. 3) an information entropy feature extraction method based on empirical mode decomposition (EMD) is proposed. An algorithm combining "energy method" and "correlation coefficient method" is proposed to select the components containing the main fault information. The experimental results show that the method can extract the characteristic data exactly. Four kinds of information entropy values including the components of the main faults are calculated including the instantaneous singular spectral entropy the frequency-domain power spectral entropy the time-frequency-domain wavelet energy spectrum entropy and the wavelet spatial state spectral entropy. The experimental results show that the proposed method can extract the quantized feature of rotor system fault signal well. The extracted feature set has the capability of making significant difference between the typical fault feature quantities. 4) based on the idea of information fusion, the singular spectral entropy of the main fault components selected by EMD decomposition is calculated. Based on power spectrum entropy and wavelet energy spectrum entropy, a state space distribution map of information entropy of fault signal is proposed. The experimental results show that the state space model can realize the rotor fault pattern recognition intuitively and accurately. 5) based on LabVIEW software, a typical rotor fault signal testing system is established. The system realizes the filtering processing of the original signal, the spectrum analysis of the original signal and the filtered signal and the analysis of the axis locus. The aim of this paper is to extract the quantized feature of rotor fault signal. The research work in this paper is worthy of further research on digital signal processing, information theory and intelligent diagnosis theory.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
本文编号：2219237
[Abstract]:With the development of sensor technology, test technology and signal analysis, fault diagnosis technology has been greatly improved. The main research contents include four aspects: signal acquisition, signal feature extraction, fault diagnosis, information fusion. In the process of fault diagnosis, signal feature extraction refers to the feature factor extraction technique, which can obtain sensitive features which are highly correlated with the system state. Whether the feature extraction is correct or not will directly affect the accuracy of fault diagnosis results. The traditional signal analysis method is difficult to accurately describe the non-stationary characteristics of rotor vibration signals and it is difficult to quantitatively evaluate the signal characteristics. In this paper, an empirical mode decomposition method combined with information entropy is proposed. The method of feature extraction and quantitative evaluation of rotor fault vibration signal is studied. The main contents and research results are as follows: 1) based on the bearing-rotor system model, the common fault types and mechanism of rotating machinery are studied. According to the characteristics of rotor fault signal, a fault signal preprocessing filter combining median filter and wavelet de-noising is designed, which provides the original data for feature extraction. In this paper, the essence and characteristics of the signal processing method-empirical mode decomposition method (EMD),) are studied. In view of the difficulty of determining the number of iterations and the endpoints effect in the process of EMD decomposition, a combination of "energy algorithm" and "similar signal translation algorithm" is proposed. The analysis of simulation signals shows that the algorithm can solve the above problems accurately and effectively. 2) the research status of information entropy in fault diagnosis is analyzed, and the singular spectrum entropy in time domain and power spectrum entropy in frequency domain are analyzed. The wavelet energy spectrum entropy in time-frequency domain and the state spectrum entropy in wavelet space are studied and analyzed systematically. An information entropy algorithm program based on LabVIEW is developed. 3) an information entropy feature extraction method based on empirical mode decomposition (EMD) is proposed. An algorithm combining "energy method" and "correlation coefficient method" is proposed to select the components containing the main fault information. The experimental results show that the method can extract the characteristic data exactly. Four kinds of information entropy values including the components of the main faults are calculated including the instantaneous singular spectral entropy the frequency-domain power spectral entropy the time-frequency-domain wavelet energy spectrum entropy and the wavelet spatial state spectral entropy. The experimental results show that the proposed method can extract the quantized feature of rotor system fault signal well. The extracted feature set has the capability of making significant difference between the typical fault feature quantities. 4) based on the idea of information fusion, the singular spectral entropy of the main fault components selected by EMD decomposition is calculated. Based on power spectrum entropy and wavelet energy spectrum entropy, a state space distribution map of information entropy of fault signal is proposed. The experimental results show that the state space model can realize the rotor fault pattern recognition intuitively and accurately. 5) based on LabVIEW software, a typical rotor fault signal testing system is established. The system realizes the filtering processing of the original signal, the spectrum analysis of the original signal and the filtered signal and the analysis of the axis locus. The aim of this paper is to extract the quantized feature of rotor fault signal. The research work in this paper is worthy of further research on digital signal processing, information theory and intelligent diagnosis theory.
【学位授予单位】：兰州理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
【引证文献】
相关硕士学位论文 前1条
1 孙茂军;汽车机械变速器动力性能试验台的研究[D];武汉理工大学;2013年
本文编号：2219237
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2219237.html
