萃取溶剂选择的计算机辅助分子设计中的数学方法

萃取精馏是分离沸点相近或具有恒沸组成混合物的一种重要方法。选择最优的萃取剂是提高生产能力和降低能耗的根本途径。萃取剂选择的计算机辅助分子设计（CAMD）是基于性质估算的方法即通过CAMD来生成一组具有期望性质的分子，然后再对候选分子进行筛选。目前，CAMD法的研究主要可分为生成一验证、数学优化和组合优化。生成一验证法通常使用启发式的策略，是基于知识的，无法克服基团的组合爆炸问题，也不能保证结果的最优性；数学优化方法，包括混合整数非线性规划（MINLP）和混合整数线性规划（MILP），可用于表达非线性或线性的结构．性质关系，但准确表达结构-性质关系还有困难；组合优化法是使用遗传算法或模拟退火算法、禁忌搜索等的方法，理论上可克服以上问题。本文评述了以上三种方法在萃取精馏溶剂选择领域中的用途、原理及其工业化所面临的主要问题。

null

Extractive distillation is an important and useful technique for separating close boiling and azeotropic liquid mixtures.The selection of appropriate solvents for extractive distillation is the key to improve the production capacity and reduce the energy consumption.Computer-aided molecular design(CAMD) is on the basis of property estimate method,generating a set of solvent molecular structures with desirable properties and screening the candidate molecules.There are three main CAMD approaches: generation-and-test,mathematical optimization,and combinatorial optimizations.The generation-and-test usually implements heuristics or knowledge-based methods to combat combinatorial explosion of group combinations,can't guarantee the quality of the results.The mathematical optimization approach,which includes MINLP(mixed-integer nonlinear programming) and MILP(mixed-integer linear programming) problems,has been applied to the nonlinear or linear structure-property correlations,but exact expression of the structure-property correlations may be very difficult or even impossible.Thus,the third approach,combinatorial optimization such as genetic algorithm(GA) and simulated annealing(SA),is implemented in CAMD and can easily handle combinatorial complexity of the molecular design.In this paper,the principle of three approaches is depicted,and a comparison among three approaches is made.Furthermore,the major encounter problems in the industrialization of the selection of solvents are also discussed.