基于支持向量机的滚动轴承故障诊断方法研究

发布时间：2018-01-09 08:11

本文关键词：基于支持向量机的滚动轴承故障诊断方法研究　出处：《江西理工大学》2012年硕士论文　论文类型：学位论文

【摘要】：轴承是旋转机械的核心部件与易损件。一般工业用途的旋转机械大多使用滚动轴承,在旋转机械中起着非常重要的作用,对其进行监测和故障诊断是学术界和产业界一直非常关注的课题。滚动轴承出现故障将可能导致设备的噪音和非正常振动,严重的将可能导致整台设备乃至整条生产线都不能正常运行。因此,对滚动轴承的故障诊断进行研究具有重要的理论意义和应用价值。本文结合江西省自然科学基金项目的研究内容,重点研究了基于小波包分析和支持向量机的滚动轴承故障诊断方法。借助设计的滚动轴承实验台以及数据采集系统,进行了滚动轴承模拟故障试验。在此基础上研究了小波分析理论,利用小波包分析完成了对数据采集系统采集的原始振动信号数据的降噪和故障特征提取,最后结合在少样本分类中具有明显优势的支持向量机(Support Vector Machine,简称SVM)理论,提出基于小波包分析和支持向量机相结合的故障诊断方法。本文的主要工作： 1、首先研究了滚动轴承的振动机理、失效形式和检测技术,分析了常见检测方法优点和不足基础,采用测量滚动轴承的勺振动信号数据进行轴承的故障诊断,进而研究了滚动轴承的振动机理和特征频率 2、分析了统计学习理论和支持向量机的基本思想,并将支持向量机引入到滚动轴承的故障诊断中,给出了支持现量机应用于滚动轴承故障诊断的基本步骤和方法。 3、以QPZ-I故障模拟试验平台为对象,设计了试验台振动数据采集系统,进行了滚动轴承内圈、外圈和滚动体的故障模拟,为滚动轴承的故障特征提取提供诊断数据。 4、以模拟的滚动轴承故障数据为诊断对象,研究了基于小波分析的信号降噪和运用小波包分解提取故障特征的方法,结合Matlab平台的优势,研究了基于小波包分解的滚动轴承信号特征提取方法,并进行实例进行了分析。 5、针对标准支持向量机不能直接用于解决故障诊断这种典型多值分类问题的不足,分析了支持向量机多值分类方法,采用二叉树多值分类算法构建分类器模型,结合小波包分解提取的振动故障特征向量,进行了滚动轴承的故障诊断,并探索了影响SVM分类精度的核参数的参数优化。仿真结果验证了支持向量用于滚动轴承故障诊断的正确性和有效性。综上所述,本文所研究的基于支持向量机的滚动轴承故障诊断方法是可行的,分析和诊断结果与实际吻合,能够满足滚动轴承故障诊断的要求,对滚动轴承的故障诊断具有一定的指导作用。
[Abstract]:The bearing is the key part of the rotating machinery and spare parts. General rotating machinery industrial uses most of the use of rolling bearings, plays a very important role in rotating machinery, the monitoring and fault diagnosis of the academia and industry have been very concerned about the issue. Rolling bearing failure may lead to equipment noise and non normal vibration, serious may cause the whole equipment and the whole production line can not run properly. Therefore, the research has important theoretical significance and application value for the fault diagnosis of rolling bearing.
This paper studies the content with the natural science foundation of Jiangxi Province, focuses on the research of wavelet packet analysis and support vector machine for rolling bearing fault diagnosis method based on the rolling bearing experimental design and data acquisition system for rolling bearing fault simulation test. Based on the theory of wavelet analysis, wavelet packet analysis the extraction of noise and fault characteristics of original vibration signal data acquisition system for data acquisition, finally combined with support vector machine has obvious advantages in small sample classification (the Support Vector Machine, referred to as SVM) theory, the fault diagnosis method of wavelet packet analysis and support vector machine based on the combination of the main work of this paper:
1, first study the vibration mechanism of rolling bearing, failure modes and detection technology, analyzes the advantages and disadvantages of common detection methods, the data measured by vibration signal of rolling bearing fault diagnosis of bearing, and then studied the vibration mechanism and characteristic frequency of rolling bearing
2, analysis of the basic theory of statistical learning theory and support vector machine, and the support is introduced into the fault diagnosis of rolling bearing in support vector machine, given the amount of machine used in rolling bearing fault diagnosis method and basic steps.
3, taking the QPZ-I fault simulation test platform as the object, we designed the test platform vibration data acquisition system, carried out the fault simulation of rolling bearing inner ring, outer ring and rolling element, and provided diagnostic data for rolling bearing's fault feature extraction.
4, the data simulation for rolling bearing fault diagnosis object, studies the signal denoising of wavelet analysis and wavelet packet decomposition method to extract fault features based on the combination of the advantages of Matlab platform, the wavelet packet method to extract signal of rolling bearing based on eigen decomposition, and the example is analyzed.
5, according to the standard support vector machine can not be directly used to solve the fault diagnosis of the typical multi valued classification problems, analyzes the classification method of multi valued support vector machine, using two binary tree multi value classification model classification algorithm based on wavelet packet decomposition to extract the fault feature of vibration vector, the fault diagnosis of rolling bearing, and to explore the effect of parameter optimization of kernel parameters for SVM classification accuracy. The simulation results verify the support vector for rolling bearing fault diagnosis is correct and effective.
To sum up, the support vector machine based rolling bearing fault diagnosis method is feasible, and the analysis and diagnosis results coincide with the reality, which can meet the requirements of rolling bearing fault diagnosis, and has a guiding role in the rolling bearing fault diagnosis.

【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3;TH133.33

【引证文献】