摘要
以原子类型电性拓扑状态指数(ETSI)有效表征122个液态烃类物质的分子结构,并分别应用人工神经网络和多元线性回归方法,对这122种液态烃类物质的燃烧热进行关联和预测研究,建立应用电性拓扑状态指数预测烃类物质燃烧热的定量结构—性质相关性(QSPR)研究模型。应用人工神经网络和多元线性回归方法对训练集样本的预测平均相对误差分别为1.17%和0.95%,对测试集20种烃类物质的预测平均相对误差分别为1.49%和1.05%。实验结果表明,无论采用人工神经网络法还是多元线性回归法,燃烧热预测值与实验值一致性均令人满意。可见应用电性拓扑态指数法预测液态烃的燃烧热是可行的,为工程上提供了一种根据物质结构预测烃类物质燃烧热的新途径。
Atom-type electrotopelogical state indices (ETSI)were used to describe the structures of 122 liquid hydrocarbon molecules, while quantitative structure-property relationship (QSPR) models based on electrotopelogical state indices were developed to predict the combustion heats of these 122 liquid hydrocarbon through the artificial neural network and the multilinear regression approach, respectively. For the training set, the average relative deviations between the experimental and predicted values of the combustion heats were 1.17% by the artificial neural network approach and 0.95% by the multilinear regression analysis method, while for the testing set which contained 20 data, they were 1.49% and 1.05%, respectively. The results showed that the predicted values of the combustion heats were in good agreement with the experimental data whether by the artificial neural network approach or the multilinear regression analysis method. It was concluded that it is feasible to predict the combustion heats of liquid hydrocarbons by the eleetrotopologieal state indices method, which can provide a new way to predict the combustion heat of hydrocarbons based on molecular structures for chemical engineering.
出处
《天然气化工—C1化学与化工》
CAS
CSCD
北大核心
2008年第2期74-78,共5页
Natural Gas Chemical Industry
基金
国家自然科学基金(No.29936110)资助项目
新世纪优秀人才支持计划(NCET-05-0505)资助项目
关键词
电性拓扑状态指数
定量结构—性质相关性(QSPR)
液态烃类
燃烧热
预测
eleetrotopelogieal state index
quantitative structure-property relationship (QSPR)
liquid hydrocarbon
combustion heat
prediction