利用神经网络预测Flory-Huggins作用参数

Prediction of Flory-Huggins Interaction Coefficient with Neural Network

以理论溶剂化变色参数来定量表达有机分子结构 ,探讨 42个有机探针分子的结构及其与聚合物相互作用的Flory Huggins参数 χ12 之间的定量关系。采用改进的神经网络方法进行建模和预测 ,并与基于理论线性溶解能关系的结果进行了比较 。

The oretical solvatochomic parameters were used to describe the organic molecules, the quantitative relationships between these parameters and Flory Huggins coefficient χ 12 have been established by using neural networks which were trained by a modified backpropagation algorithm, predictions based on the trained neural networks have been compared with which obtained through TLSER(theoretical linear solvation energy relationship). The results show that neural networks have better prediction ability than TLSER.