基于形态小波和最小二乘支持向量机的滚动轴承故障诊断研究

发布时间：2018-05-01 22:31

本文选题：滚动轴承 + 故障诊断　；参考：《电子科技大学》2012年硕士论文

【摘要】：滚动轴承是机械中应用最广泛的零件之一，也是非常容易发生故障的零件。当滚动轴承发生故障时，不仅会影响到企业的正常生产过程，严重情况下还会造成生产事故，危及人生安全。因此，对于滚动轴承的故障诊断的研究是很有必要的。发生故障的滚动轴承在运行过程中，其振动往往更为强烈。实际上，振动是轴承运行时不可避免的现象。故障轴承的振动信号包含着丰富的故障信息，由此可通过分析其振动信号进行故障的诊断。滚动轴承的振动信号通常含有较多的噪声以及冲击成分，在分析之前先要进行去噪处理。传统的分析方法如傅里叶分析只适合应用于平稳信号的分析，对于具有时变性质的滚动轴承故障信号无能为力。小波是近来发展起来的，可用于分析时变信号的分析方法，在低频处可获得较好的频域分辨率，高频处可获得较好的时域分辨率。但是，小波固有的算法会产生频率折叠的问题，这对于提取故障信号的分量来说是不利的。形态小波兼顾了数学形态学和小波的优良特性，能在去噪的同时，更好的保持细节，本文采用形态中值小波提取振动信号的能量作为模式识别的特征向量，通过比较可以发现，形态小波的效率要优于小波，且能提取到小波所不能提取到的特征。传统的神经网络等分类方法需要大量的训练样本，这对于数据样本较少的轴承故障数据来说，是不适用的。采用支持向量机进行分类的优点是训练所需要的样本数量相对较少，能解决通过向高位空间的映射解决非线性问题，，而且不会带来维数的灾难。最小二乘支持向量机是基础的支持向量机的拓展，其算法的效率和精度要更优于基础的支持向量机。
[Abstract]:Rolling bearing is one of the most widely used parts in machinery. When the rolling bearing breaks down, it will not only affect the normal production process of the enterprise, but also cause the production accident and endanger the safety of life. Therefore, it is necessary to study the fault diagnosis of rolling bearings. The vibration of the malfunctioning rolling bearing is often stronger. In fact, vibration is an inevitable phenomenon in bearing operation. The vibration signal of the fault bearing contains abundant fault information, so the fault diagnosis can be carried out by analyzing the vibration signal of the bearing. The vibration signal of rolling bearing usually contains more noise and shock components. The traditional analysis methods such as Fourier analysis are only suitable for the analysis of stationary signals, but there is no way to deal with the fault signals of rolling bearings with time-varying properties. Wavelet is developed recently and can be used to analyze time-varying signals. It can obtain better frequency domain resolution at low frequency and better time domain resolution at high frequency. However, the inherent wavelet algorithm produces the problem of frequency folding, which is unfavorable for extracting the component of fault signal. Morphological wavelet combines the excellent characteristics of mathematical morphology and wavelet, and can better keep details while de-noising. In this paper, the energy of vibration signal is extracted as the feature vector of pattern recognition by means of morphological median wavelet, which can be found by comparison. Morphological wavelet is more efficient than wavelet and can extract features which cannot be extracted by wavelet. Traditional classification methods such as neural networks require a large number of training samples, which is not suitable for bearing fault data with less data samples. The advantage of using SVM to classify is that the number of samples needed for training is relatively small which can solve the nonlinear problem by mapping to the high space and will not bring about the disaster of dimension. Least square support vector machine (LS-SVM) is an extension of basic SVM, and its efficiency and accuracy are better than that of basic SVM.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】