基于改进的遗传算法设计萃取精馏萃取剂

Extractants design based on modified genetic algorithm

将Mod.UNIFAC基团贡献法与改进的遗传算法相结合,进行萃取精馏萃取剂的分子设计。采用了一种简单易行的新编码方案,使合成的分子形式简单,便于链接数据库计算各种性质。提出了跨代精英选择策略,错位交叉算子,滞后变异算子,使设计出的分子结构稳定且形式多样,有效地避免了遗传算法不易得到全局最优解和容易提前收敛的问题。给出了分子设计的主要流程和方法,并编制了相应的计算机程序,将其应用于丙酮-甲醇体系,得到了水和乙二醇等萃取剂。结果表明,文中提出的方法对相对挥发度的预测精度高于文献值,同时检验了该方法的可靠性。

A novel strategy for designing extractants by combining Mod. UNIFAC group contribution method with modified genetic algorithm was presented. In order to simplify the synthetical molecular form and link easily to the concatenated data set for calculating molecules＇ physicochemical properties, a new code scheme was established. For the sake of avoiding the disadvantages of genetic algorithm with converging early and losing easily globally optimal results, cross generatent elitist selection, misplaced crossover and hysteresis mutation were improved. Because of the improvement of genetic algorithm, the molecules＇ output structures were stable and the forms were diversified. The methodology and design process for the molecular design were given. The computer program was developed. The feasibility and rationality of this method were verified by designing solvents for the extractive distillation processes of acetone-methanol. The results show that the prediction precision of relative volatility of this work is better than the document reported.