基于反演的精馏过程动态扰动原因的诊断

发布时间：2018-08-09 21:11

【摘要】：精馏是石油化工生产过程中应用最广泛的操作之一,对精馏设备所作的微小扰动可能会引发事故,产生巨大的经济损失。为确保精馏装置安全、平稳运行,及时发现异常状态中参数的劣化趋势,需要及时识别异常原因,从源头上预防和控制精馏事故的发生。精馏过程中存在大量扰动,扰动原因难以确定,扰动量信息诊断精度难以提高。基于以上问题,本文将物理领域的反演思想应用到精馏扰动诊断领域,建立扰动反演模型,深入分析扰动原因。本文以单、双变量扰动为例,对精馏过程扰动原因反演问题进行了研究。考虑动态精馏过程的非线性与非稳定性,结合物料衡算和能量衡算等方程,运用机理建模的方法,建立了精馏塔的动态数学模型,并动态模拟出正常、异常样本。对应地,运用人工神经网络(ANN)、否定选择法(NSA)和支持向量机(SVM)方法确定出扰动类型,然后结合遗传算法(GA)建立了各扰动量与特征表示量的反演模型,通过运算获得扰动量的大小。考虑到上述诊断方法中人工提取特征所带来的复杂性和不确定性,本文运用深度学习(DL)对扰动类型进行识别,增强了识别过程的智能性。本文选取脱丙烷生产工艺为研究对象,建立了动态模拟仿真系统。同时也建立了扰动量的反演模型,诊断出了工艺中常见的扰动原因,对上述所提方法进行了验证和对比。结果表明,此方法能够快速定位扰动类型,准确得出扰动量,实现了动态精馏系统扰动原因的深入辨识。
[Abstract]:Distillation is one of the most widely used operations in the process of petrochemical production. Small disturbances to the distillation equipment may cause accidents and produce huge economic losses. In order to ensure the safety and smooth operation of the distillation unit, the deterioration trend of the parameters in the abnormal state is found in time. It is necessary to identify the abnormal causes in time and to prevent and control from the source. There are a lot of disturbances in the distillation process. The cause of disturbance is difficult to be determined and the accuracy of the disturbance information diagnosis is difficult to be improved. Based on the above problems, this paper applies the inverse thought in the physical field to the field of distillation disturbance diagnosis, establishes a disturbance inversion model, and analyzes the cause of disturbance. This paper takes the single and bivariate perturbation as an example. The problem of the inversion of the cause of the distillation process is studied. Considering the nonlinear and non stability of the dynamic distillation process, combining the equations of material balance and energy balance, the dynamic mathematical model of the distillation column is established by the method of mechanism modeling, and the normal and abnormal samples are simulated dynamically, and the artificial neural network (ANN) is used for the dynamic simulation. The NSA and support vector machine (SVM) method are used to determine the disturbance type, and then the inverse model of the disturbance momentum and the feature representation is established by combining the genetic algorithm (GA). The size of the disturbance is obtained by operation. The depth learning (D) is used in this paper. L) to identify the type of disturbance and enhance the intelligence of the identification process. In this paper, a dynamic simulation system is established by selecting the propane production process as the research object, and the inversion model of the disturbance is also established. The common disturbance causes in the process are diagnosed and the methods are verified and contrasted. The results show that this method is used. The method can locate the disturbance types quickly and get the disturbance quantities accurately, thus realizing the in-depth identification of the disturbance causes of the dynamic distillation system.
【学位授予单位】：青岛科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TQ028.31;TQ221.13

【参考文献】