含卤代烃二元共沸物共沸特性的QSPR模型  

QSPR model for azeotropic properties of binary azeotropes containing halogenated hydrocarbons

作　　者：周倩 诸林[1] 李松照 吕利平[2] 张春花 郝强 ZHOU Qian;ZHU Lin;LI Song-zhao;LYU Li-ping;ZHANG Chun-hua;HAO Qiang(School of Chemistry and Chemical Engineering,Southwest Petroleum University,Chengdu 610500,Sichuan Province,China;School of Chemistry and Chemical Engineering,Yangtze Normal University,Chongqing 408100,China)

机构地区：[1]西南石油大学化学化工学院,四川成都610500 [2]长江师范学院化学化工学院,重庆408100

出　　处：《化学工程》2021年第11期47-52,65,共7页Chemical Engineering(China)

摘　　要：为了安全、经济地获取含卤代烃二元共沸物共沸特性数据,提出基于定量结构-性质关系(QSPR)原理预测共沸特性。选用239种含卤代烃的二元共沸物作为研究对象,利用遗传函数逼近算法(GFA)筛选分子描述符并分别构建共沸温度和共沸组成模型。根据拟合能力挑选最佳模型,然后采用内部验证、外部验证、y-随机验证及应用域分析对最佳模型稳定性及预测精度进行评价。最后,对比了最佳模型与同类QSPR模型以及基于UNIFAC基团贡献法模型。结果表明:最佳共沸温度模型由5个描述符组成,最佳共沸组成模型由8个描述符组成。经过对最佳模型的分析与验证,得出所建模型具有良好的稳定性和预测性。In order to safely and economically predict the azeotropic characteristics data of binary azeotropes,a method based on the quantitative structure-property relationship was proposed.239 binary azeotropes containing halogenated hydrocarbons as research samples,the genetic function approximation(GFA)method was used in the screening molecular descriptors process and constructed azeotropic temperature and azeotropic composition models.The best models that were selected according to the fitting ability were evaluated by internal validation,external validation,y-scrambling and application domain analysis for stability and prediction accuracy assessment.At last,statistical parameters of the best models were compared with the similar QSPR model and the UNIFAC group contribution method model.The results show that the both best models of azeotropic temperature and azeotropic composition are composed of 5 descriptors and 8 descriptors,respectively.The established model has good stability and predict ability through the analysis and validation of the best model.

关 键 词：卤代烃 共沸物 定量构效 遗传算法 预测 

分 类 号：TQ015[化学工程]