基于全矢谱的旋转机械故障特征提取研究

发布时间：2018-01-03 10:11

本文关键词：基于全矢谱的旋转机械故障特征提取研究　出处：《郑州大学》2011年硕士论文　论文类型：学位论文

【摘要】：大型旋转类机械往往是企业的关键性咽喉设备,它们以转子及其它回转部件作为工作的主体。在石油、化工、冶金、发电等大、中型企业中,旋转设备约占80%的比例,包括压缩机、鼓风机、汽轮机、发电机、轧钢机等。作为企业的核心设备,一旦发生故障,将给企业甚至人们的生命财产带来难以估量的损失和伤害。所以大型旋转类设备运行状态的监测及其故障的及时诊断和解决也是科技工作者愈来愈关心的问题。旋转类机械运行状态的监测及其故障诊断依据是被诊断对象所表征的一切有用的信息,比如振动、噪声、转速、温度、压力、流量等。旋转类机械的振动信号中蕴含了大量的信息,可以帮助人们监测设备的运行状态及判断故障的类型。故障特征提取就是对系统的动态信号预处理后得到的信息进行分析和处理,提取与系统状态有关的数据,再对得到的数据进行处理和分析,提取其中与系统状态相关性较大的敏感特征。有效特征向量的提取是故障诊断中的关键环节,也是能否及时、正确的做出故障诊断的关键因素。针对传统单通道信息采集的不完整及实时性差等问题,本文将全矢谱技术分别与粗集理论、小波分析方法结合起来,提出了基于全矢谱技术的旋转类机械故障特征提取方法。全矢谱分析技术基于旋转机械同源信息融合,它是矢量谱分析及其一系列扩展分析方法的统称。它可以融合转子一个截面上的两个或三个通道的信息并对这些信息进行组合,它不仅弥补了传统单通道分析信息不足及不完整等缺点,而且具有分辨率高、三维分析可行性、高分辨率下指示转子在各回转频率下的振动强度和方位及与传统分析方法的兼容性等特点。粗糙集理论是由波兰的Z.Pawlak教授于上世纪80年代提出,是一种能够分析处理不精确、不一致、不完整信息与知识的数学工具,它的基本思想是通过数据库分类归纳形成概念和规则,通过等价关系的分类对目标的近似实现知识发现。通常用来作为数据约简的工具,它在消除冗余信息等方面有良好效果。小波分析是傅里叶分析的一个自然延伸,小波分析技术是由Morlet在1984年首先提出的,它克服了传统傅里叶变换只考虑正弦振动的能量而没有考虑其他振动方式的能量的缺点,对输入信号的要求较低,具有灵敏度高,克服噪声能力强等优点。小波变换具有良好的时频局部化特性和对信号自适应变焦、多分辨率分析的能力,可以将信号在不同尺度上展开.提取各个频带上的特性的同时也保留了频带相应的各个尺度上的时频特性。用小波分析技术对故障特征进行提取更为有效。本文探讨了矢谱理论的基本原理及算法,并将全矢谱技术分别与粗糙集理论和小波分析技术相结合,提出了基于全矢谱技术-粗集理论在旋转机械频谱特征提取中的应用及对小波-包络分析和全矢小波分析在滚动轴承故障特征提取中应用并对两种方法进行对比研究,编制Matlab程序及进行相关实验验证其功能。
[Abstract]:Large rotating machinery is often the key equipments in the enterprise, they are the rotor and other rotating parts as the main body of work. In the petroleum, chemical, metallurgy, power generation and other large and medium-sized enterprises, accounting for about 80% of the proportion of rotating equipment, including compressor, blower, steam turbine, generator, rolling machines as the core. The equipment of enterprises, once the fault will bring immeasurable loss and damage to the enterprise and even the life and property of people. So the monitoring of large rotating equipment operating status and fault diagnosis and solution is more and more concerned about the problems of science and technology workers.
Monitoring and fault diagnosis for rotating machinery running condition is diagnosed all the useful information, object representation such as vibration, noise, speed, temperature, pressure, flow and so on. The vibration signal contains a lot of information, can help people to monitor the operational status of equipment and determine the fault type. Analysis and treatment of fault feature extraction is the system dynamic signal after pretreatment of information extraction, data related to the system state, then the processing and analysis of the data obtained from the system state and correlated sensitive features. The feature vector is the key of fault diagnosis, but also timely, correctly make fault diagnosis of the key factors.
Aiming at the problems of incomplete information and poor real-time performance of traditional single channel information collection, the full vector spectrum technology is combined with rough set theory and wavelet analysis method respectively. A feature extraction method of rotating machinery based on full vector spectrum technology is proposed.
The full vector spectrum analysis technology of rotating machinery based on homologous information fusion, it is referred to as the vector spectrum analysis and patulous analysis methods. It can be two or three channel fusion rotor to a section of the information and the combination of these information, it not only makes up for the traditional single channel analysis and the lack of information is not complete shortcomings, but also has high resolution, three-dimensional feasibility analysis, compatibility features of instruction of the rotor's vibration intensity and range and the traditional analysis method of high resolution.
Rough set theory is put forward by Professor Z.Pawlak of Poland in 80s, is an analysis to deal with imprecise, inconsistent, incomplete mathematical tools of information and knowledge, it is the basic idea of the database classification form concepts and rules, similar to realize knowledge discovery of target classification by equivalence relations. Usually used as the tools of data reduction, it has good effect in eliminating redundant information. Wavelet analysis is an extension of Fourier analysis, wavelet analysis is first proposed by Morlet in 1984, which overcomes the disadvantages of traditional Fourier transform only considering sinusoidal vibration energy without considering other modes of vibration energy faults, low requirement on the input signal, which has high sensitivity, strong ability to overcome the advantages of noise. Wavelet transform has good time-frequency localization characteristics and the signal from Adapting to the capability of zoom and multiresolution analysis, we can expand the signal on different scales, extract the characteristics of each frequency band, and also retain the time-frequency characteristics of the corresponding scales at different frequencies. It is more effective to extract fault features with wavelet analysis technology.
This paper discusses the basic principle and algorithm of vector spectrum theory, and the vector spectrum theory and wavelet analysis technology respectively with the combination of rough sets, puts forward the application of the full vector spectrum technology of rough set theory in rotating machinery spectrum in the feature extraction and wavelet envelope analysis and full vector wavelet analysis and comparative study the two methods used in rolling bearing fault feature extraction based on Matlab program and the related experiments to verify its function.

【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3

【引证文献】