摘要
基于分子距边矢量（ＭＤＥ），借助多元线性回归技术（ＭＬＲ）建立起描述卤代烃沸点变化规律的定量结构－性质相关关系（ＱＳＰＲ）模型，其复相关系数Ｒ＝０．９９３５，均方根误差ＲＭＳ＝１０．０７Ｋ。最后５次随机选取２５个化合物作预测集，以余下的６２个化合物作校正集建ＱＳＰＲ模型（Ｒ＝０．９９２８，ＲＭＳ＝１０．８８Ｋ），并有效地预测了沸点（Ｒ＝０．９９２５，ＲＭＳ＝１１．７８Ｋ）。结果表明，模型的预测能力良好。另外，还运用所建立的模型对３０个未知沸点的卤代甲、乙烷的沸点值进行了预测，根据预测出的沸点数据指出了ＣＦＣ１１，ＣＦＣ１２和ＣＦＣ１１３的一些可能的代用品。
An approach based on a molecular distance edge (MDE) vector is used to study the relationship between boiling point of 87 haloalkanes with one through two carbon atoms,halo mathane and halo ethane ,and their structure. The correlative coefficient and rooted mean squares of the QSPR model are R=0.9935 and RMS=10.07K ,respectively. Then 5 models are developed by using 62 haloalkanes selected as the training sets at random from the total 87 haloalkanes and their performances are demonstrated by employing the rest 25 samples as the testing sets. And good results are obtained with the mean values of R and RMS between the calculated Bp and observed Bp values being R=0.9928 and RMS=10.88K . All these results show that the predicted Bp values are in good accordance with the experimental data.Finally the QSPR model is applied in prediction of boiling points of 30 halo methane and haloethane samples whose Bp are unknown.Some possible alternatives for CFC 11 ,CFC 12 and CFC 113 can be found with the predicted boiling point values.
基金
国家自然科学基金
国家教委及机械部优秀人才专项基金
关键词
分子距边矢量
卤代甲烷
卤代乙烷
卤代烃
沸点
molecular distance edge (MDE) vector
haloalkane
boiling point (Bp)
mutiple linear regression (MLR)
quantative structure property relationship(QSPR)
