基于EMD分解的小波脊线法在故障诊断中的应用

发布时间：2018-12-26 12:40

【摘要】：随着社会生产的发展，要求设备生产效率不断提高。这就对液压系统以及轴承元件的性能提出了更高的要求。因此液压泵与轴承元件的实时故障诊断就更加重要。轴向柱塞泵与滚动轴承故障信号是一种典型的非平稳、非线性信号，这就需要一种适合处理非平稳、非线性信号的方法准确地提取故障特征。EMD分解与小波脊线在处理这类信号中具有其独特的优势。经验模态分解(EMD)可以把原始信号分解为一系列固有模态函数(IMF)之和。各个IMF分量起到了对数据特征放大的效果，对各个IMF分量进行研究能够更清楚的发现故障特征。小波脊线基于小波变换理论，是由在时间—频率面上满足每个时刻点小波系数的模极大值点所形成的集合。这些点往往更能清楚地表征故障的特征信息。因此把经验模态分解与小波脊线结合起来对故障信号的分析能够更清楚地分析故障信息。为了验证该方法的有效性和优越性，本文对斜盘式轴向柱塞泵采集的振动信号与美国Case Western Reserve University轴承故障模拟试验台采集的轴承故障振动信号进行了分析研究。通过边际谱对比提取故障信号EMD分解后的敏感IMF分量进行小波脊线包络解调分析，准确地提取了液压泵以及滚动轴承各种状态时的敏感频率。并通过敏感IMF分量小波脊线解调后的时频谱与Hilbert变换解调时频谱进行了对比，证明了该方法较Hilbert变换解调具有更高的时频定位精度。本文提出了一种基于EMD分解的小波脊线解调信号能量特征向量提取方法，通过EMD分解后敏感IMF分量进行小波脊线解调得到包络信号，对降低采样频率后的包络信号再次进行EMD分解，利用再次分解后的IMF分量信号能量有效地提取了故障特征向量。利用K均值聚类方法对液压泵与轴承的各种状态进行了故障模式识别，通过与Hilbert变换解调提取的特征向量对比证明了该方法具有一定的优势。
[Abstract]:With the development of social production, equipment production efficiency has been improved. This puts forward higher requirements for the performance of hydraulic system and bearing components. Therefore, the real-time fault diagnosis of hydraulic pump and bearing components is more important. The fault signal of axial piston pump and rolling bearing is a typical nonstationary and nonlinear signal. EMD decomposition and wavelet ridge have unique advantages in processing such signals. Empirical mode decomposition (EMD) can decompose the original signal into the sum of a series of intrinsic modal functions (IMF). Each IMF component has the effect of magnifying the data features, and the study of each IMF component can find fault features more clearly. The wavelet ridge is based on the wavelet transform theory and is a set of modulus maximum points which satisfy the wavelet coefficients at every time point on the time-frequency plane. These points can more clearly represent the characteristic information of the fault. Therefore, the fault signal can be analyzed more clearly by combining empirical mode decomposition with wavelet ridge. In order to verify the effectiveness and superiority of this method, the vibration signals collected by oblique disc axial piston pump and bearing fault vibration signal collected by Case Western Reserve University bearing fault simulator in USA are analyzed and studied in this paper. The sensitive IMF component of fault signal after EMD decomposition is extracted by edge spectrum contrast and the wavelet ridge envelope demodulation analysis is carried out. The sensitive frequency of hydraulic pump and rolling bearing is accurately extracted. The time-frequency spectrum of wavelet ridge demodulation of sensitive IMF component is compared with that of Hilbert transform demodulation. It is proved that this method has higher time-frequency localization accuracy than Hilbert transform demodulation. In this paper, an energy eigenvector extraction method for wavelet ridge demodulation signal based on EMD decomposition is proposed. The envelope signal is obtained by demodulating the sensitive IMF component of wavelet ridge by EMD decomposition. The envelope signal with lower sampling frequency is decomposed by EMD again, and the fault eigenvector is extracted effectively by using the energy of the IMF component signal after being decomposed again. The K-means clustering method is used to identify the fault patterns of hydraulic pump and bearing. The comparison with the eigenvector extracted by Hilbert transform and demodulation proves that this method has some advantages.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;TN911.7

【参考文献】