摘要
基于BP(Back-Propagation)神经网络模型,编写了通用的计算程序,对30℃时H2O-[Bmim]BF4-Na2CO3离子液体双水相体系的液液平衡数据进行了关联,经过反复试算,得到最佳的神经网络拓扑结构为{3,10,3},采用该结构关联富含水相中三种组分的质量分数,平均相对误差分别为0.065%、2.218%和0.781%,最大相对误差为3.916%;优于文献中的Othmer-Tobias经验方程+溶解度曲线方程的关联精度,该方程得到的平均相对误差分别为0.09%、2.77%和1.73%,最大相对误差为4.52%。较为成功地应用神经网络模型处理了含离子液体的双水相体系的液液平衡数据。
A reasonable data model for the ionic liquid-based aqueous two-phase system of H2O-[Bmim]BF4-Na2CO3 was introduced.Based on the artificial neural network model using back propagation algorithm,the general programme was compiled to correlate liquid-liquid equilibrium data of this system.The perfect topology{3,10,3} for structure of the network was gained by trial computations with the programme.The quality fraction of three components in the aqueous phase were correlated by using this structure.The average relative errors were 0.065%,2.218%and 0.781%respectively,as well as the maximum relative error was 3.916%.The results show that the artificial neural network model has better accuracy compared with the othmer-tobias and solubility equations mentioned in the document,whose average relative error are 0.09%,2.77% and 1.73%respectively as well as whose maximum relative error is 4.52%.
出处
《当代化工》
CAS
2011年第7期748-750,共3页
Contemporary Chemical Industry
关键词
离子液体
双水相
液液平衡
BP神经网络
Ionic liquid
Aqueous two-phase system
Liquid-liquid phase equilibrium
BP neural network