石化复杂系统故障诊断方法研究

发布时间：2018-04-09 09:32

本文选题：石化复杂系统　切入点：故障诊断　出处：《燕山大学》2015年硕士论文

【摘要】：石油化工生产是社会发展的基础产业和经济发展的支柱产业,它在现代化社会发展进程中起着十分重要的作用。随着石油化工过程的生产结构和规模日趋现代化、大型化和复杂化,生产事故发生的概率也逐渐增加。因而对石化生产过程进行有效的故障诊断来预防或避免事故的发生势在必行。为了进行有效的故障诊断,本文提出了混合故障诊断方法,其包括故障监测和故障诊断两个方面,以某石化公司的气体分馏装置中的脱异丁烷单元为应用实例进行研究验证。首先建立了基于主元分析(Principal Component Analysis,PCA)的故障监测模型。运用主元分析方法对设置的4种在线工况进行了故障监测,监测结果均与预先设置的工况一致,结果表明运用主元分析方法不仅可以极大的降低数据维数,简化计算,而且可以有效地进行在线监测,及时发现故障。然后建立了基于误差反馈神经(Back Propagation,BP)网络的诊断模型。结果表明运用BP神经网络构建故障诊断系统,只能判断出故障所属的故障类别,无法判断出具体的故障情况。采用DS(Dempster-Shafer evidence theory)证据理论对BP神经网络的诊断结果进行数据融合,结果表明运用DS证据理论可以在预知工况所属故障类别的前提下有效地诊断出故障的原因。但是BP-DS相结合的方法仍存在着诊断时间长,计算量大等不足之处。径向基(Radial Basis Function,RBF)神经网络虽然没有BP神经网络应用广泛,但是其分类能力、逼近能力、训练速度等特性全都优于BP网络,于是本文最后建立了基于RBF神经网络的诊断模型。结果表明在判别阈值为0.85时,单独运用RBF神经网络故障诊断方法可以在诊断精度稍低的情况下快速地判断出故障原因。
[Abstract]:Petrochemical production is the basic industry of social development and the pillar industry of economic development. It plays a very important role in the process of modern social development.With the production structure and scale of petrochemical process being modernized, large-scale and complicated, the probability of production accident is increasing gradually.Therefore, it is imperative to make effective fault diagnosis to prevent or avoid accidents in petrochemical production process.In order to carry out effective fault diagnosis, a hybrid fault diagnosis method is proposed in this paper, which includes two aspects: fault monitoring and fault diagnosis. The application of deisobutane unit in a gas fractionation unit of a petrochemical company is studied and verified.Firstly, a fault monitoring model based on Principal Component Analysis (PCA) is established.The method of principal component analysis (PCA) is used to monitor the four online working conditions, and the monitoring results are consistent with the pre-set conditions. The results show that the PCA method can not only greatly reduce the dimension of data and simplify the calculation.Moreover, it can effectively monitor on-line and find fault in time.Then a diagnosis model based on error feedback neural back propagation (BPN) network is established.The results show that using BP neural network to construct fault diagnosis system can only judge the fault category of the fault, but can not judge the specific fault situation.DS(Dempster-Shafer evidence the theory of evidence is used to fuse the diagnosis results of BP neural network. The results show that the DS evidence theory can be used to diagnose the causes of faults effectively on the premise of predicting the types of faults belonging to the working conditions.However, the BP-DS method still has some shortcomings, such as long diagnosis time and large amount of calculation.Although Radial Basis function neural network is not widely used in BP network, its classification ability, approximation ability and training speed are all better than BP neural network. Therefore, a diagnosis model based on RBF neural network is established in this paper.The results show that when the threshold value is 0. 85, the fault causes can be quickly determined by using the RBF neural network fault diagnosis method alone under the condition of lower diagnostic accuracy.
【学位授予单位】：燕山大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE96;TE65

【参考文献】