基于经验小波变换和奇异值分解的旋转机械故障诊断

发布时间：2018-05-06 21:33

本文选题：机械故障诊断 + 经验小波变换　；参考：《西南交通大学》2017年硕士论文

【摘要】：旋转机械的故障信号通常是非平稳、非线性的含噪振动信号,对于机械故障诊断目前应用较为广泛的是窗口傅里叶变换、Wigner分布、小波变换(WT)等时频分析方法,但这些方法都具有一定的局限性,而且容易受到干扰的影响。经验模态分解(EMD)较其他方法具有自适应分解的优势,但是由于其自身存在模态混叠效应、端点效应以及缺乏一定的理论基础,所以在应用方面还存在一定问题。而经验小波变换(EWT)则既具有EMD的自适应性又具有可靠的理论基础,其极大地减弱了 EMD方法中存在模态混叠现象,克服了端点效应问题,在旋转机械故障诊断中具有较高的应用价值。笔者利用经验小波变换的自适应性结合奇异值分解的滤噪特性提出了新的旋转机械故障诊断方法。文章首先介绍了经验小波变换理论,将经验小波变换和经验模态分解对多模态混叠含噪信号的分解结果进行对比,验证经验小波变换较经验模态分解存在的优势。然后对奇异值分解和奇异值包分解理论进行深入研究。最后提出EWT-SVD和EWT-SVDP算法,并通过仿真信号验证算法的有效性。文章将联合算法应用在旋转机械故障诊断的实例分析中,选取轴承、转子、万向轴作为研究对象,利用试验对轴承的内圈故障、转子碰摩故障、万向轴的动不平衡故障振动信号进行提取,应用文章提出的算法对故障信号进行分析研究,验证联合算法对于工程实际应用的有效性。通过应用文章提出的联合算法对仿真信号和实际工程的故障信号分析可知:基于经验小波变换结合奇异值分解(EWT-SVD)的算法表现出很好的自适应性和良好的滤噪性能,能够将多模态含噪仿真信号有效的分解成含有不同频率特性的信号,并且和单独经验小波变换方法相比,联合算法表现出良好的滤噪特性,通过对轴承故障信号、转子碰摩故障信号、万向轴动不平衡故障信号分析,基于经验小波变换结合奇异值分解能够有效地将原始振动信号分解成不同频带中的分量信号;基于经验小波变换结合奇异值包分解(EWT-SVDP)相比经验小波分解,不但表现出良好的自适应性,滤噪性能也有显著提高,而且对于信号的细节成分分析表现出极大的优势,对于工程实际信号的处理能力也大大提高。
[Abstract]:The fault signals of rotating machinery are usually non-stationary and nonlinear noise bearing vibration signals. The time frequency analysis methods such as window Fourier transform, Wigner distribution and wavelet transform (WT) are widely used for mechanical fault diagnosis, but these methods all have certain local limit and are easily affected by interference. Empirical mode decomposition (EMD) has the advantages of adaptive decomposition compared with other methods, but because of its own existence of modal aliasing effect, endpoint effect and lack of a certain theoretical basis, there are still some problems in application. The empirical wavelet transform (EWT) has both the adaptability of EMD and a reliable theoretical basis, which greatly weakens the EMD In this method, the phenomenon of modal aliasing has been found to overcome the endpoint effect problem and has high application value in the fault diagnosis of rotating machinery. The author uses the adaptive characteristic of the empirical wavelet transform to combine the filter noise characteristics of singular value decomposition to put forward a new method of diagnosis of rotating machinery fault. The author introduces the theory of empirical wavelet transform, which will be used by Zhang Shouxian. The empirical wavelet transform and empirical mode decomposition are used to compare the decomposition results of multimodal mixed signals. The advantages of empirical wavelet transform are verified. Then the singular value decomposition and the singular packet decomposition theory are studied. Finally, the EWT-SVD and EWT-SVDP algorithms are proposed, and the simulation signals are used to verify the algorithm. In this paper, the joint algorithm is applied to the case analysis of rotating machinery fault diagnosis. Bearing, rotor and universal axis are selected as the research object. The test is used to extract the inner ring fault of the bearing, the rotor rubbing fault, the dynamic unbalance fault vibration signal of the universal axis, and the fault signal is analyzed and researched by the algorithm proposed in this paper. To verify the effectiveness of the combined algorithm for the practical application of the project, the combined algorithm proposed in the application of the paper shows that the algorithm based on the empirical wavelet transform combined with the singular value decomposition (EWT-SVD) shows good adaptability and good filtering performance, and can multimodal denoising. The simulation signal is effectively decomposed into signals with different frequency characteristics. Compared with the independent empirical wavelet transform, the combined algorithm shows good noise filtering characteristics. Through the analysis of the bearing fault signal, the rotor rubbing fault signal, the universal axis dynamic unbalance fault signal analysis, the empirical wavelet transform combined with singular value decomposition can be used. The original vibration signal is effectively decomposed into component signals in different frequency bands; based on empirical wavelet transform combined with singular packet decomposition (EWT-SVDP), compared with empirical wavelet decomposition, not only good adaptability is shown, but also the noise performance is greatly improved, and it has a great advantage for the analysis of the detail component of the signal. The processing ability of the actual signal is also greatly improved.

【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17

【参考文献】