基于EMD的轴承故障诊断
发布时间：2018-09-18 21:28
【摘要】：滚动轴承是旋转机械中的关键部件，也是容易出现故障的部件，对其进行故障诊断是国内外工程技术领域一直非常关注的课题。据大量的研究事实证明，，目前对滚动轴承的状态进行监测与诊断，最实用的方法是振动信号分析法。测取的振动信号是非平稳、非线性的，经验模态分解方法是自适应的分解方法，尤其适用于非线性、非平稳信号的分解。 本文首先介绍了滚动轴承的故障模式和发生故障的机理，计算了实验所用轴承的特征频率和固有频率，通过对采集的振动信号进行时频分析得到了故障频率，通过对比理论计算和软件计算得到的结果，对轴承的故障模式做了初步诊断，得到了各故障模式的频谱特征。 本文主要利用小波精确的频带换分优势对采集的振动信号进行去噪，尝试使用平移不变量方法对采集的信号去噪，取得了不错的效果。然后对去噪后的信号进行经验模态分解法分解获得了本征模函数分量，由于信号的特征主要集中在前几个分量，计算了前7个IMF的能量组成特征向量，结合神经网络和支持向量机两种方法对滚动轴承的工作状况做出诊断，对两种方法的识别效果做了比较。
[Abstract]:Rolling bearing is one of the key parts in rotating machinery, and it is also prone to malfunction. Fault diagnosis of rolling bearing is a subject of great concern in the field of engineering and technology at home and abroad. According to a large number of research facts, the most practical method to monitor and diagnose the status of rolling bearings is vibration signal analysis. The measured vibration signal is nonstationary and nonlinear. The empirical mode decomposition method is an adaptive decomposition method, which is especially suitable for the decomposition of nonlinear and non-stationary signals. In this paper, the fault mode and fault mechanism of rolling bearing are introduced, the characteristic frequency and natural frequency of bearing used in experiment are calculated, and the fault frequency is obtained by time-frequency analysis of vibration signal collected. By comparing the results of theoretical calculation and software calculation, the fault modes of bearings are preliminarily diagnosed, and the spectrum characteristics of each fault mode are obtained. In this paper, we mainly use the advantage of frequency band switching of wavelet to Denoise the collected vibration signal, and try to use the method of translation invariant to de-noise the collected signal, and get good results. Then the eigenmode function components of the denoised signal are obtained by empirical mode decomposition. Because the characteristics of the signal are mainly concentrated in the first few components, the energy component Eigenvectors of the first seven IMF are calculated. Neural network and support vector machine are used to diagnose the working condition of rolling bearing, and the recognition effect of the two methods is compared.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;TH133.33
本文编号：2249143
[Abstract]:Rolling bearing is one of the key parts in rotating machinery, and it is also prone to malfunction. Fault diagnosis of rolling bearing is a subject of great concern in the field of engineering and technology at home and abroad. According to a large number of research facts, the most practical method to monitor and diagnose the status of rolling bearings is vibration signal analysis. The measured vibration signal is nonstationary and nonlinear. The empirical mode decomposition method is an adaptive decomposition method, which is especially suitable for the decomposition of nonlinear and non-stationary signals. In this paper, the fault mode and fault mechanism of rolling bearing are introduced, the characteristic frequency and natural frequency of bearing used in experiment are calculated, and the fault frequency is obtained by time-frequency analysis of vibration signal collected. By comparing the results of theoretical calculation and software calculation, the fault modes of bearings are preliminarily diagnosed, and the spectrum characteristics of each fault mode are obtained. In this paper, we mainly use the advantage of frequency band switching of wavelet to Denoise the collected vibration signal, and try to use the method of translation invariant to de-noise the collected signal, and get good results. Then the eigenmode function components of the denoised signal are obtained by empirical mode decomposition. Because the characteristics of the signal are mainly concentrated in the first few components, the energy component Eigenvectors of the first seven IMF are calculated. Neural network and support vector machine are used to diagnose the working condition of rolling bearing, and the recognition effect of the two methods is compared.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;TH133.33
【参考文献】
相关期刊论文 前2条
1 江虹,曾立波,胡继明;优化的BP神经网络分类器的设计与实现[J];计算机工程与应用;2001年05期
2 郭代飞,高振明,张坚强;利用小波门限法进行信号去噪[J];山东大学学报(自然科学版);2001年03期
相关硕士学位论文 前2条
1 张俊;基于小波分形和神经网络的滚动轴承故障诊断[D];北京交通大学;2009年
2 崔玉敏;基于振动信号非线性方法的轴承故障诊断研究[D];江苏大学;2010年
本文编号：2249143
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2249143.html
