基于定量构效关系预测烃类物质及其衍生物的过热极限温度  

Predicting the superheat limit temperature of hydrocarbons and derivatives of hydrocarbons based on the quantitative structure-property relation theory

作　　者：周璐璐[1] 蒋军成[1] 潘勇[1] 

机构地区：[1]南京工业大学安全科学与工程学院江苏省危险化学品本质安全与控制技术重点实验室,江苏南京210009

出　　处：《南京工业大学学报（自然科学版）》2017年第4期138-142,共5页Journal of Nanjing Tech University(Natural Science Edition)

基　　金：江苏省自然科学基金(BK20141457);江苏省普通高校研究生科研创新计划(KYZZ15_0221);江苏省"333高层次人才工程"项目(BRA2015313);江苏省高校自然科学基金(12KJA620001)

摘　　要：采用定量结构-性质相关性原理(QSPR)对烃类有机物及其衍生物的过热极限温度(SLT)进行研究,实现了通过分子结构预测过热极限温度的过程,通过遗传算法(GA)筛选出与过热极限温度关系最为密切的分子描述符,采用多元线性回归的方法建立了预测模型,模型的复相关系数为0.955,说明模型具有很强的线性相关性,对模型进行内部验证和外部验证,结果表明模型具有较强的稳定性和预测能力,并且对模型进行了应用域(AD)分析。The superheat limit temperature（ SLT）,maximum temperature above which a liquid will boil,is one of the most important hazard parameters for boiling liquid expanding vapor explosion and similar explosions.A quantitative structure-property relationship（ QSPR） model was proposed to predict the superheat limit temperature of hydrocarbons and derivatives of hydrocarbons from the only information of their molecular structures.Various kinds of molecular descriptors were calculated and the genetic algorithm（ GA） was used to select optimal subset of descriptors which had significant contribution to the superheat limit temperature.The multiple linear regression was employed to build the best model of six variables with the high fitting correlation coefficient of 0. 955. The model was used to check the stability and predictive capability,and its applicability domain（ AD） was also analyzed.

关 键 词：定量构效关系 过热极限温度 遗传算法 多元线性回归 应用域 

分 类 号：X937[环境科学与工程—安全科学]