结合遗传算法的支持向量回归及在盐析分相法中的应用

发布时间：2018-01-29 14:50

本文关键词： 支持向量回归遗传算法参数优化盐析分相法　出处：《河北工业大学》2015年硕士论文　论文类型：学位论文

【摘要】：在化工分离过程的有关研究中,盐析分相法被公认为是一种有效的分离恒沸有机水溶液的方法,也是目前的一个研究热点。但该方法的理论尚不成熟,在其工艺的研发过程中存在很多困难。对此,本文构建了基于遗传算法优化参数的多维支持向量回归(GA-MSVR)模型,将其应用到盐析分相法的相关预测中,主要完成了以下工作:首先,作为一个跨学科的研究课题,通过了解化工分离领域的基础理论知识,重点研究了常用的恒沸有机水溶液的分离方法,针对盐析分相法缺乏实验数据的问题,采用计算机模拟求解的方法,对有机物-水-无机盐体系液-液相平衡的典型问题进行了分析,进而提出了应用多输出支持向量回归算法预测经历盐析效应之后有机物和水的含量。其次,研究了多维支持向量回归算法(MSVR),采用了遗传算法对MSVR的参数进行优化处理,以实现参数的自动选取,避免了参数选择的随机性和主观性,提出了GA-MSVR模型的构造性算法,并且利用多维函数映射的模拟数据进行了相关的实验设计和验证工作。再次,在分析了盐析分相法理论模型的基础上,形成了自己的设计思想,开发出以分离效果计算器为核心的应用程序,在不同的实验条件下,可方便求得经历盐析效应达到两相平衡时,有机相中水和有机物的含量,实现了分离效果的计算机仿真计算。最后,开发了基于GA-MSVR的盐析分相法预测软件,将GA-MSVR模型应用到分离恒沸有机水溶液的有关计算中,以期减少化学实验的盲目性,降低实验成本,为此问题的解决提供一种计算机模拟分析的途径。同时进行了相关的实验验证工作,取得了较好的实验结果。在与多个单输出支持向量回归算法预测的分离效果的比较中,由于GA-MSVR考虑到了多个输出之间的联系,具有更好的整体预测精度。
[Abstract]:In the related research of chemical separation process, salting-out phase separation method is recognized as an effective method for separating azeotropic organic aqueous solution, and it is also a research hotspot at present. However, the theory of this method is still immature. There are many difficulties in the process of research and development. In this paper, a multi-dimensional support vector regression model based on genetic algorithm (GA) is constructed. The main work of this paper is as follows: first, as an interdisciplinary research topic, through understanding the basic theoretical knowledge in the field of chemical separation. The separation methods of azeotropic organic aqueous solution were mainly studied. The method of computer simulation was used to solve the problem of lack of experimental data in salting-out phase method. The typical problems of liquid-liquid phase equilibrium in organic-water-inorganic salt system are analyzed, and the multi-output support vector regression algorithm is proposed to predict the content of organic matter and water after salting out. The multi-dimension support vector regression algorithm is studied in this paper. Genetic algorithm is used to optimize the parameters of MSVR so as to realize the automatic selection of parameters and avoid the randomness and subjectivity of parameter selection. The constructive algorithm of GA-MSVR model is proposed, and the related experimental design and verification work are carried out by using the simulation data of multidimensional function mapping. On the basis of analyzing the theoretical model of salting-out phase separation method, the author formed his own design idea, and developed an application program with the separation effect calculator as the core, under different experimental conditions. It is convenient to obtain the content of water and organic matter in organic phase when the salting-out effect reaches two-phase equilibrium and realize the computer simulation calculation of separation effect. Finally. A phase separation prediction software based on GA-MSVR was developed and the GA-MSVR model was applied to the calculation of separating azeotropic organic aqueous solution in order to reduce the blindness of chemical experiments. To reduce the cost of the experiment, this paper provides a way of computer simulation and analysis to solve the problem. At the same time, the related experimental verification work is carried out. Good experimental results have been obtained. In comparison with the prediction results of multiple single-output support vector regression algorithms, GA-MSVR takes into account the relationship between multiple outputs. Better overall prediction accuracy.
【学位授予单位】：河北工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TQ028;TP18

【参考文献】