Infinte Dilution Activity Coefficient Predictions Via Artificial Neural Network

Wang Xiuli, Sun Wei, Yan Yan, Song Haihua^*
(School of Chemical Engineering, Tianjin University, Tianjin 300072)

Abstract The infinite dilution activity coefficients of solvents are calculated by using BP neural network and molecular connectivity index, and the average relative error is within 9.25%. The prediction with artificial neural network is more accurate than that with UNIFAC model.
Key words Activity coefficient, Molecular connectivity indice, Artificial neural network

用人工神经网络预测无限稀释活度系数

王秀丽 孙 伟 闫 岩 宋海华^*
(天津大学化工学院 天津 300072)

摘要 用BP神经网络，以分子连接性指数等作为输入参数，对各种溶剂的无限稀释活度系数进行预测，平均相对误差约9.25%。与UNIFAC法比较，人工神经网络方法要优于UNIFAC法。
关键词 活度系数 分子连接性指数 人工神经网络

王秀丽 女，31岁，博士生，现从事化工分离研究。 *联系人
2002-05-13收稿，2002-09-05修回