流动轴承故障模式识别方法研究

发布时间：2018-11-27 14:45

【摘要】：近几年来,随着科学技术的不断进步与经济的不断发展,各行业中的生产装备都朝着大型化、精密化、复杂化、自动化的方向发展。作为应用最广泛的旋转类机械部件,滚动轴承直接决定并影响着整个系统的生产和运行状况。一方面,这些技术进步能够提升生产效率,为厂家带来可观的生产效益和丰厚的利润回报；另一方面,装备的大型化、复杂化、精密化及自动化也极大提高了装备的生产成本,一旦这些装备发生故障,就会造成巨大的经济损失和人员伤亡事故。因此,对滚动轴承故障模式识别技术展开研究,保证其正常运行,具有十分重要的意义。本研究在国家“十一五”科技支撑计划：“危险化学品生产安全保障关键技术研究”(项目编号：2006BAK01B01)的支持下完成的,主要研究工作如下：一、介绍了本课题的研究背景及目的,阐述了模式识别技术在国内外的研究现状及工程应用,列举了本研究的主要工作内容及创新点。二、介绍和研究了部分信号处理方法及特征选择和提取技术,主要包括快速傅里叶变换、循环统计理论、经验模态分解以及基于奇异值分解和主成分分析的特征提取方法。三、研究和改进了本论文中的两个重要模型,分别为经验模态分解过程中的局部均值模型及端点效应模型。在前人的研究基础上,提出了极值域均值和极值间均值相结合的局部均值模型,研究了端点效应处理方法,取得了一定的效果。四、提出了基于二阶循环统计量的奇异值分解模型的模式识别方法,并将其引入到滚动轴承故障状态识别中来。借助于CWRU轴承数据中心的滚动轴承不同工作状态数据,对该模型进行了实验验证,取得了较好的识别结果,可以值得深入研究和应用。五、提出了基于经验模态分解的主成分分析模型的模式识别方法,并将其引入到滚动轴承故障状态识别中,在CWRU轴承数据中心的试验数据支持下,对该理论模型进行了实验验证,结果表明识别精度较高,较好的完成了预期的目标。
[Abstract]:In recent years, with the continuous progress of science and technology and the development of economy, the production equipment in various industries has developed towards the direction of large-scale, precision, complexity and automation. As the most widely used rotating mechanical parts, rolling bearings directly determine and affect the production and operation of the whole system. On the one hand, these technological advances can improve the efficiency of production, bring considerable production benefits and rich profit returns for manufacturers; On the other hand, the large-scale, complex, precision and automation of the equipment also greatly increase the production cost of the equipment, once these equipment failure, will cause huge economic losses and casualties. Therefore, it is of great significance to study the fault pattern recognition technology of rolling bearing to ensure its normal operation. This study was completed under the support of the National "Eleventh Five-Year Plan" Science and Technology support Plan: "Research on key Technologies for Safety and Security of Hazardous Chemicals production" (Project No.: 2006BAK01B01). The main research work is as follows: 1. This paper introduces the research background and purpose of this subject, expounds the research status and engineering application of pattern recognition technology at home and abroad, and lists the main work contents and innovation points of this research. Secondly, some signal processing methods and feature selection and extraction techniques are introduced and studied, including fast Fourier transform, cyclic statistical theory, empirical mode decomposition and feature extraction based on singular value decomposition and principal component analysis. Thirdly, two important models in this paper are studied and improved, namely, the local mean model and the endpoint effect model in the process of empirical mode decomposition. On the basis of previous studies, a local mean model combining the mean of polar range and the mean between extreme values is proposed, and the method to deal with the endpoint effect is studied, and some results are obtained. Fourthly, a pattern recognition method for singular value decomposition (SVD) model based on second-order cyclic statistics is proposed and applied to the fault state recognition of rolling bearings. With the help of the different working state data of the rolling bearing in the CWRU bearing data center, the model is verified by experiments and good recognition results are obtained, which is worthy of further study and application. 5. A method of principal component analysis (PCA) based on empirical mode decomposition (EMD) is proposed, which is applied to the fault state recognition of rolling bearing, supported by the experimental data of CWRU bearing data center. The experimental results show that the recognition accuracy is high and the expected target is achieved.
【学位授予单位】：北京化工大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3

【参考文献】