基于支持向量回归的轴承故障定量诊断方法研究
发布时间：2018-10-24 14:35
【摘要】：机械设备运行条件复杂、环境恶劣的工况下，其核心零部件和重要机械结构会不可避免地发生不同程度的故障。机械设备一旦出现故障，可能会带来巨大的经济损失和人员伤亡。然而机械设备故障的演变存在一个由轻微到严重的发展过程。因而准确及时识别运行过程中萌生和演变过程，对保障机械设备安全运行、避免经济损失和灾难性事故意义重大。故障定量诊断方法是一种能够有效的确定故障演变历程和故障大小的机械故障诊断方法。本文以轴承为对象，针对故障状态特征与故障大小之间的高度非线性、故障样本少等特点，提出采用支持向量回归机方法，建立轴承故障严重程度判断模型和故障大小定量诊断研究方案。 首先，，进行了轴承常见的故障失效形式分析以及滚动轴承故障的运动学研究，并介绍了本文的轴承故障模拟实验系统，在此系统上进行了不同程度的轴承外圈故障试验。 然后，系统的阐述了统计学习理论，介绍了基于结构风险最小化原则的支持向量机理论，并进一步提出将支持向量分类机和支持向量回归机用于故障特征分类和回归描述，为故障诊断模型的建立提供了理论基础。研究了轴承振动信号故障特征的提取方法，提取了相关的特征如峰值、方差、峭度等，并且分析了这些特征与故障大小之间的关系，为故障诊断模型的建立提供了数据基础。 最后根据所提取的特征运用支持向量回归机分别建立了轴承故障程度的分类模型和故障大小的定量诊断模型，分类模型用于定量评价故障程度，定量诊断模型用于确定故障大小。方法在训练集和测试集上的效果验证了该方法的有效性。 本文基于轴承振动信号特征，采用支持向量回归机建立了轴承故障诊断模型，包括了故障程度分类模型和故障大小定量诊断模型，为轴承故障程度分类和定量诊断确立了有效的方法。 本论文依托于江苏省自然科学基金项目(批准号: BK2010225)：“基于瞬态振动特征辨识的轴承局部故障定量诊断研究”。
[Abstract]:When the operating conditions of machinery and equipment are complex and the environment is bad, the failure of its core parts and important mechanical structures will inevitably occur in varying degrees. Once the mechanical equipment breaks down, it may bring huge economic losses and casualties. However, the evolution of mechanical failure has a slight to serious development process. Therefore, it is of great significance to accurately and timely identify the process of initiation and evolution in the operation process to ensure the safe operation of mechanical equipment and to avoid economic losses and catastrophic accidents. The quantitative fault diagnosis method is a kind of mechanical fault diagnosis method which can effectively determine the fault evolution history and fault size. This paper takes the bearing as the object, aiming at the high nonlinearity between the fault state characteristic and the fault size, and so on, and puts forward the support vector regression machine method. The model of bearing fault severity and the research scheme of fault size quantitative diagnosis are established. Firstly, the common failure forms of bearing and the kinematics of rolling bearing fault are analyzed, and the bearing fault simulation experiment system is introduced in this paper, and the bearing outer ring fault test is carried out on the system. Then, the statistical learning theory is systematically expounded, and the support vector machine theory based on structural risk minimization principle is introduced. Furthermore, support vector classification machine and support vector regression machine are proposed for fault feature classification and regression description. It provides a theoretical basis for the establishment of fault diagnosis model. The fault feature extraction method of bearing vibration signal is studied, and the correlation features such as peak value, variance and kurtosis are extracted, and the relationship between these features and fault size is analyzed, which provides a data basis for the establishment of fault diagnosis model. Finally, according to the extracted features, the classification model of bearing fault degree and the quantitative diagnosis model of fault size are established by using support vector regression machine, and the classification model is used to quantitatively evaluate the fault degree. The quantitative diagnosis model is used to determine the fault size. The effectiveness of the method is verified on the training set and test set. Based on the characteristics of bearing vibration signal, the bearing fault diagnosis model is established by using support vector regression machine, which includes the classification model of fault degree and the quantitative diagnosis model of fault size. An effective method for classification and quantitative diagnosis of bearing fault degree is established. This paper is based on the project of Jiangsu Provincial Natural Science Foundation (Grant No.: BK2010225): "quantitative diagnosis of bearing local faults based on transient vibration feature identification".
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
[Abstract]:When the operating conditions of machinery and equipment are complex and the environment is bad, the failure of its core parts and important mechanical structures will inevitably occur in varying degrees. Once the mechanical equipment breaks down, it may bring huge economic losses and casualties. However, the evolution of mechanical failure has a slight to serious development process. Therefore, it is of great significance to accurately and timely identify the process of initiation and evolution in the operation process to ensure the safe operation of mechanical equipment and to avoid economic losses and catastrophic accidents. The quantitative fault diagnosis method is a kind of mechanical fault diagnosis method which can effectively determine the fault evolution history and fault size. This paper takes the bearing as the object, aiming at the high nonlinearity between the fault state characteristic and the fault size, and so on, and puts forward the support vector regression machine method. The model of bearing fault severity and the research scheme of fault size quantitative diagnosis are established. Firstly, the common failure forms of bearing and the kinematics of rolling bearing fault are analyzed, and the bearing fault simulation experiment system is introduced in this paper, and the bearing outer ring fault test is carried out on the system. Then, the statistical learning theory is systematically expounded, and the support vector machine theory based on structural risk minimization principle is introduced. Furthermore, support vector classification machine and support vector regression machine are proposed for fault feature classification and regression description. It provides a theoretical basis for the establishment of fault diagnosis model. The fault feature extraction method of bearing vibration signal is studied, and the correlation features such as peak value, variance and kurtosis are extracted, and the relationship between these features and fault size is analyzed, which provides a data basis for the establishment of fault diagnosis model. Finally, according to the extracted features, the classification model of bearing fault degree and the quantitative diagnosis model of fault size are established by using support vector regression machine, and the classification model is used to quantitatively evaluate the fault degree. The quantitative diagnosis model is used to determine the fault size. The effectiveness of the method is verified on the training set and test set. Based on the characteristics of bearing vibration signal, the bearing fault diagnosis model is established by using support vector regression machine, which includes the classification model of fault degree and the quantitative diagnosis model of fault size. An effective method for classification and quantitative diagnosis of bearing fault degree is established. This paper is based on the project of Jiangsu Provincial Natural Science Foundation (Grant No.: BK2010225): "quantitative diagnosis of bearing local faults based on transient vibration feature identification".
【学位授予单位】：苏州大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3
【参考文献】
相关期刊论文 前10条
1 魏传锋,李运泽,王浚,于涛;航天器热故障诊断专家系统推理机的设计[J];北京航空航天大学学报;2005年01期
2 高彦宇,杨扬,陈飞;基于融合特征和LS-SVM的脱机手写体汉字识别[J];北京科技大学学报;2005年04期
3 翟永杰,王东风,韩璞;基于多类支持向量机的汽轮发电机组故障诊断[J];动力工程;2003年05期
4 高学,金连文,尹俊勋,黄建成;一种基于支持向量机的手写汉字识别方法[J];电子学报;2002年05期
5 翟永杰,毛继s
本文编号：2291687
本文链接：https://www.wllwen.com/kejilunwen/jixiegongcheng/2291687.html
