纯物质汽化热的人工神经网络算法

Calculation for Heat of Vaporization of Pure Compounds by Artificial Neural Network

１前言纯物质汽化热是重要的基础化工数据，其测定、关联、预测和理论研究相当活跃。在诸多预测模型中最有代表性的当属基团贡献模型，但多采用基团的简单加和或加权组合，基团贡献值由实验数据回归，通用性差，难以区别同分异构体，预测结果不尽人意。人工神经网络具有基...

According to the artificial neural network theory and by the research for organic molecular structure, this article proposes that the contribution of group to molecular property can be reflected by the joint action of neuron of artificial neural network, weight factor and calculation law. And a new prediction model of artificial neural network is set up for the heat of vaporization of pure compounds at normal boiling point and 298.15K. Through the test of more than 400 kinds of compounds, it can be showed that this model is superior to the residual function model and group contribution model which are thought better than others till now, this model is also applicable for organic compounds with multiple groups.