基于实时特征值的风机振动状态监测与数据挖掘的故障诊断研究

发布时间：2018-10-15 09:45

【摘要】：风机普遍应用于石油、化工、电力、冶金等行业,随着设备长时间的运行,发生故障的概率大大上升,便可能使其停产,会造成巨大的经济损失和安全隐患。因此,对风机设备进行状态监测与故障诊断具有极其重要的意义。本文详细介绍了国内外旋转机械状态监测和故障诊断技术的发展现状,发现了目前的研究现状是:在振动信号的监测上国内外都已有较成熟的方法和技术,但是在故障诊断上常用的频域方法由于计算量大,基本上仍然是离线式的。诊断结果的分析需要人工进行,各种应用方法的专业性强,不易完成实时计算和在线分析判断,缺少在线的、智能的方法和技术。随着就地监测设备成本的降低,对风机设备安装传感器进行运行状况监测变成了可能。通过采集风机设备全周期运行的时域特征参数,得到风机在不同运行状态下的特征参数变化,形成风机设备运行特征参数数据库,结合数据挖掘的方法,能够从海量的数据中建立基于时域特征值的故障诊断模型。这样做的优点是由于时域特征值是实时的、在线的,故障诊断模型也将是实时的、在线的,同时故障诊断的准确率也将随着数据库的扩充以及数据挖掘方法的改进而不断提高。本文为解决CAP1400安全壳再循环冷却风机无状态监测的问题,主要进行了以下工作:围绕振动的基本概念、旋转机械典型振动故障研究成果以及振动信号的特征提取和分析方法作了介绍,并且根据国际国内以及行业标准选取了部分时域的典型特征值作了具体的分析,为下一步的故障诊断方法建立了基础;其次,设计开发了风机振动状态监测平台软件部分,并且进行了风机运行状态监测试验,验证和完善了振动状态监测平台的功能,完成了风机特征参数的计算和数据存储,建立了风机振动特征值数据库;最后,依托于振动特征值数据库,对特征值参数进行了数据挖掘,建立了不同振动故障的特征值敏感等级分布表,确定了不同特征参数之间的潜在关系,评估预测了风机运行状态。在此基础上,开发了基于实时特征值数据挖掘的风机振动故障诊断模型;应用该振动故障诊断模型对实际故障进行在线分析和诊断。本论文的诊断方法,在试验台架上获得了有效的应用,表明了基于数据挖掘的诊断方法能够实时有效的完成振动故障的在线分析和不同故障原因的诊断,对实现机械行业广泛应用的风机、水泵等旋转机械的现场实时监测和振动智能远程诊断,具有工程应用价值和推广价值。
[Abstract]:Fan is widely used in petroleum, chemical, electric power, metallurgical and other industries. With the equipment running for a long time, the probability of failure is greatly increased, which may make it stop production, which will cause huge economic losses and safety risks. Therefore, the condition monitoring and fault diagnosis of fan equipment is of great significance. In this paper, the status quo of status monitoring and fault diagnosis of rotating machinery at home and abroad is introduced in detail. It is found that there are mature methods and techniques for vibration signal monitoring both at home and abroad. However, the frequency-domain method commonly used in fault diagnosis is still offline because of the large amount of calculation. The analysis of diagnosis results needs to be carried out manually, and all kinds of application methods are highly professional. It is difficult to complete real-time calculation and on-line analysis and judgment, and lacks on-line and intelligent methods and techniques. With the reduction of the cost of local monitoring equipment, it is possible to monitor the operating condition of fan equipment installation sensors. By collecting the time domain characteristic parameters of the fan equipment running in the whole period, the change of the fan characteristic parameters under different running conditions is obtained, and the database of the fan equipment running characteristic parameters is formed, and the method of data mining is combined. A fault diagnosis model based on time domain eigenvalue can be established from massive data. The advantage of this method is that the time domain eigenvalue is real-time and on-line, and the fault diagnosis model will be real-time and on-line. At the same time, the accuracy of fault diagnosis will be improved with the expansion of database and the improvement of data mining method. In order to solve the problem of stateless monitoring of CAP1400 containment recirculation cooling fan, the main work of this paper is as follows: around the basic concept of vibration, The research results of typical vibration faults of rotating machinery and the methods of feature extraction and analysis of vibration signals are introduced, and some typical characteristic values in time domain are selected according to international and domestic standards and industry standards for specific analysis. The foundation of the next fault diagnosis method is established. Secondly, the software of fan vibration condition monitoring platform is designed and developed, and the fan running condition monitoring test is carried out to verify and perfect the function of the vibration condition monitoring platform. The calculation and data storage of fan characteristic parameters are completed, and the database of fan vibration eigenvalue is established. Finally, based on the vibration characteristic value database, the data mining of the characteristic value parameter is carried out. The distribution table of eigenvalue sensitivity level of different vibration faults is established, the potential relationship between different characteristic parameters is determined, and the operating state of fan is evaluated and predicted. On this basis, a fault diagnosis model of fan vibration based on real-time eigenvalue data mining is developed, and the fault diagnosis model is used to analyze and diagnose the actual faults online. The method of diagnosis in this paper has been effectively applied on the test-bed, which shows that the diagnosis method based on data mining can effectively complete the on-line analysis of vibration faults and the diagnosis of different fault causes in real time. It has engineering application value and popularizing value to realize field real time monitoring and intelligent remote diagnosis of vibration of rotating machinery such as fan pump and so on which are widely used in mechanical industry.
【学位授予单位】：上海发电设备成套设计研究院
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH43;TP311.13

【参考文献】