基于共振稀疏分解的滚动轴承故障诊断方法研究

发布时间：2018-07-03 04:32

本文选题：滚动轴承 + 故障诊断　；参考：《北京交通大学》2017年硕士论文

【摘要】：滚动轴承被广泛应用于机械设备中,是旋转设备的重要部件,同时又是一个主要的故障源,其工作状态正常与否直接影响到机械设备的运行稳定性和安全性。因此,为了及时发现故障,降低经济损失,对滚动轴承进行运行状态监测和故障诊断具有十分重要的意义。本文在对滚动轴承结构和振动特性深入研究的基础上,研究了共振稀疏分解方法,并针对其在参数选择问题,提出了优化方法,取得了较好的信号分解结果。另外,对模式识别方法进行了研究并提出了优化方法,能够对滚动轴承故障信号进行有效地模式识别。本文主要内容如下:阐述了滚动轴承故障诊断的研究背景和意义,总结了故障诊断技术的发展过程,系统介绍了滚动轴承故障特征提取方法和模式识别方法的研究状况。研究了滚动轴承的故障形式和故障诊断方法,根据滚动轴承的结构和振动机理,给出了故障特征频率的计算公式,并总结了基于振动信号的滚动轴承故障诊断的基本步骤。深入研究了共振稀疏分解方法的基本原理,针对其参数选择问题,提出采用PSO算法对品质因子的确定过程进行优化。为加强全局寻优能力,引入了模拟退火算法和调整惯性权重因子的方法,对PSO算法作出了改进,得到了基于改进PSO算法优化的共振稀疏分解方法。采用不同方法对模拟信号进行分解和频谱分析,得到故障特征频率,通过分解结果的对比,验证了本文所提方法的有效性。研究了支持向量机的分类原理,针对支持向量机在处理大样本问题上的局限性,提出最小二乘支持向量机分类方法,利用改进的PSO算法对其进行参数优化。利用优化的分类方法对Wine数据进行分类识别,证明了优化的分类方法的有效性。利用滚动轴承的故障振动信号,对本文提出的故障特征提取方法和模式识别方法进行了实验验证。对滚动轴承故障信号进行共振稀疏分解,一方面,对分解得到的低共振分量进行频谱分析,提取出故障特征频率;另一方面,将低共振分量对应的系数作为支持向量机的输入,进行故障模式识别。利用不同方法对滚动轴承信号进行分解,通过故障特征频率提取结果和模式识别分类准确率的对比,表明了本文所提优化方法的优越性和鲁棒性。
[Abstract]:Rolling bearing is widely used in mechanical equipment. It is an important part of rotating equipment and also a main fault source. Whether its working condition is normal or not has a direct impact on the operation stability and safety of mechanical equipment. Therefore, in order to find fault in time and reduce economic loss, it is very important to monitor and diagnose rolling bearing running condition. Based on the deep study of the structure and vibration characteristics of rolling bearings, the resonance sparse decomposition method is studied in this paper, and an optimization method is proposed to solve the problem of parameter selection, and a better signal decomposition result is obtained. In addition, the method of pattern recognition is studied and the optimization method is put forward, which can effectively recognize the fault signals of rolling bearings. The main contents of this paper are as follows: the research background and significance of rolling bearing fault diagnosis are expounded, the development process of fault diagnosis technology is summarized, and the research status of fault feature extraction method and pattern recognition method of rolling bearing is systematically introduced. The fault form and fault diagnosis method of rolling bearing are studied. According to the structure and vibration mechanism of rolling bearing, the calculation formula of fault characteristic frequency is given, and the basic steps of fault diagnosis of rolling bearing based on vibration signal are summarized. In this paper, the basic principle of the resonance sparse decomposition method is deeply studied, and the PSO algorithm is proposed to optimize the process of determining the quality factor in view of the problem of parameter selection. In order to enhance the ability of global optimization, the simulated annealing algorithm and the method of adjusting inertia weight factor are introduced. The PSO algorithm is improved, and the resonance sparse decomposition method based on the improved PSO algorithm is obtained. Different methods are used to decompose the analog signal and the frequency spectrum is analyzed, and the fault characteristic frequency is obtained. The validity of the proposed method is verified by the comparison of the decomposition results. In this paper, the classification principle of support vector machine is studied. Aiming at the limitation of support vector machine in dealing with large sample problem, the least square support vector machine classification method is proposed, and the improved PSO algorithm is used to optimize its parameters. The optimized classification method is used to classify and identify Wine data, which proves the effectiveness of the optimized classification method. The fault feature extraction method and pattern recognition method proposed in this paper are verified by using the fault vibration signal of rolling bearing. The resonance sparse decomposition of rolling bearing fault signal is carried out. On the one hand, the frequency spectrum of the low resonance component is analyzed to extract the fault characteristic frequency; on the other hand, the coefficient corresponding to the low resonance component is used as the input of the support vector machine. Fault pattern recognition is carried out. Different methods are used to decompose the rolling bearing signals. The comparison of fault feature frequency extraction results with the classification accuracy of pattern recognition shows the superiority and robustness of the optimization method proposed in this paper.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【参考文献】