基于振动信号SVM的管壳式换热器堵塞故障诊断方法研究

发布时间：2018-10-15 09:26

【摘要】：换热器是大型石化企业的一种典型化工设备,其典型故障是管道结垢、堵塞,严重时会引起泄漏,这些故障会影响换热器的换热效果,严重时甚至影响整套化工装置的安全运行。因此,研究换热器的状态监测系统十分必要。传统的化工设备状态监测系统,一般采用有线网络传输数据,因其布线困难、施工难度大等原因,对投产后的化工设备进行状态监测是个难题。无线传感器网络具有安装部署灵活,可扩展性和鲁棒性强等优点,使其可以解决有线监测系统的布线难题。迄今国内外针对换热器流体诱导振动原理和减振设计等方面研究较多,针对换热器故障监测的系统比较少,利用振动判断循环水换热器故障状态的研究则更少。本文以基于无线振动传感器的循环水换热器故障故障诊断为选题,主要研究内容有:(1)基于流固耦合的换热器堵塞故障模型研究采用有限元法建立输流直管不同结垢程度的故障模型,探讨结垢故障状态下输流直管的振动属性,研究流体流速和结垢程度对输流直管振动的影响程度。通过输流直管振动实验,验证该模型的可行性。通过建立流固耦合的换热器堵塞故障模型,研究不同堵塞管数、不同入口流速下换热器故障模型的振动属性,探讨流速和堵塞管数对换热器振动信号的影响。(2)基于振动信号的循环水换热器故障诊断方法研究通过分析换热器不同状态的振动信号,结合入口流速和离心泵转速等参数,提出一种基于换热器入口流速的振动信号时频域故障特征提取法,建立换热器时频故障特征数据库和基于整流平均值的支持向量机分类器(MAV-SVM)。研究换热器振动信号的五个经典无量纲参数,建立基于无量纲参数特征提取的支持向量机分类器(DPD-SVM),进一步构建基于整流平均值的六维DPD-SVM。通过研究SVM参数优化技术,提高DPD-SVM分类器在换热器故障诊断中的识别准确率。(3)换热器堵塞故障实验床及现场实验研究通过人工填充堵塞物的方法,建立不同堵塞管数的换热器实验模型,分析换热器各种堵塞故障状态下的振动信号,验证时频域故障特征提取法、MAV-SVM和DPD-SVM分类器的可行性。通过对石化生产现场不同状态三台换热器的振动信号采样和分析,研究生产现场换热器振动信号获取的可行性和现场环境噪声对数据采集的影响程度,并验证DPD-SVM方法对不同状态振动信号的可分性。(4)基于振动信号的换热器状态监测系统研究运用LabVIEW和Matlab混合编程技术,开发基于有线振动传感器和无线振动传感器的换热器信号分析系统,实现对振动信号的时域、频域参数和无量纲参数的分析。结合换热器典型故障的故障特征数据库和DPD-SVM分类器,开发基于实验室换热器模型的实时监控、故障诊断和报警系统。
[Abstract]:Heat exchanger is a kind of typical chemical equipment in large petrochemical enterprise. Its typical fault is pipe scaling, blockage and leakage when serious, which will affect the heat transfer effect of heat exchanger. It even affects the safe operation of the whole chemical plant when it is serious. Therefore, it is necessary to study the condition monitoring system of heat exchanger. The traditional state monitoring system of chemical equipment generally uses wired network to transmit data. Because of the difficulty of wiring and construction, it is a difficult problem to monitor the state of chemical equipment after putting into production. Wireless sensor network (WSN) has the advantages of flexible installation, scalability and robustness, which makes it possible to solve the wiring problem of wired monitoring system. Up to now, there are many researches on fluid induced vibration principle and vibration absorption design of heat exchangers at home and abroad, less on fault monitoring systems for heat exchangers, and even less on judging fault state of circulating water heat exchangers by vibration. In this paper, the fault diagnosis of circulating water heat exchanger based on wireless vibration sensor is selected. The main research contents are as follows: (1) based on the fluid-solid coupling model of heat exchanger clogging fault, the finite element method is used to establish the fault model of different scaling degree of the straight pipe, and the vibration properties of the straight pipe under the condition of scaling fault are discussed. The influence of flow velocity and scaling degree on the vibration of straight pipe was studied. The feasibility of the model is verified by the vibration experiment of straight pipe. The vibration properties of heat exchanger fault model with different number of blocked tubes and different inlet velocities are studied by establishing a fluid-solid coupling model of heat exchanger clogging fault. The effect of flow rate and the number of blocked tubes on the vibration signal of heat exchanger is discussed. (2) the fault diagnosis method of circulating water heat exchanger based on vibration signal is studied by analyzing the vibration signal of heat exchanger in different states, combining the parameters of inlet velocity and centrifugal pump speed, etc. A time-frequency fault feature extraction method for vibration signals based on inlet velocity of heat exchanger is proposed. The database of time-frequency fault features of heat exchangers and the support vector machine classifier (MAV-SVM) based on rectifier average are established. Five classical dimensionless parameters of vibration signal of heat exchanger are studied. Support vector machine classifier (DPD-SVM) based on feature extraction of dimensionless parameters is established, and a six-dimensional DPD-SVM. based on average rectifier is further constructed. In order to improve the accuracy of DPD-SVM classifier in heat exchanger fault diagnosis, the SVM parameter optimization technology is studied. (3) the method of manually filling the plug in heat exchanger clogging test bed and field experiment is studied. The experimental model of heat exchanger with different number of blocked tubes is established, and the vibration signals of heat exchanger under various clogging fault conditions are analyzed, and the feasibility of time-frequency domain fault feature extraction method, MAV-SVM and DPD-SVM classifier is verified. By sampling and analyzing the vibration signals of three heat exchangers in different states of petrochemical production field, the feasibility of obtaining vibration signals of heat exchangers in production site and the influence of environmental noise on data acquisition are studied. The separability of the DPD-SVM method for vibration signals in different states is verified. (4) the state monitoring system of heat exchangers based on vibration signals is studied using LabVIEW and Matlab mixed programming technology. The signal analysis system of heat exchanger based on wireline vibration sensor and wireless vibration sensor is developed to realize the analysis of vibration signal in time domain frequency domain and dimensionless parameter. A real-time monitoring fault diagnosis and alarm system based on laboratory heat exchanger model is developed based on the fault characteristic database of heat exchanger and DPD-SVM classifier.
【学位授予单位】：华南理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TQ051.5

【相似文献】