水/二甲基乙酰胺/聚砜体系双结点线的神经网络预测

基于实验数据,采用人工神经网络对水/二甲基乙酰胺/聚砜体系的双结点线进行了预测. 设计并验证了具有两个输入神经元(体系温度和聚砜的质量分数)和两个输出神经元(水和二甲基乙酰胺的质量分数)的BP(Back Propagation)网络, 选取50组训练样本和39组检验样本分别对网络进行了训练与检验. 结果表明，对水与二甲基乙酰胺质量分数的预测平均相对误差分别为1.86%和0.10%. 采用该网络能较好地预测常压下、20~60℃温度范围内水/二甲基乙酰胺/聚砜体系的双结点线.

Prediction of Binodal Curve of H2O/DMAc/PSf System by ANN Method

Based on the literature data, a Back-Propagation (BP) network was designed for prediction of liquid liquid equilibrium properties of H2O/DMAc(N-N-dimethylacetamide)/PSf(polysulfone) system in terms of the ANN (Artificial Neural Network) theory. Two input neurons (temperature and mass fraction of PSf) and two output neurons (mass fractions of water and DMAc) have been chosen as variables of the network. The results show that predicted ARD (average relative deviation) values of water and DMAc concentrations are 1.86% and 0.10%, respectively, therefore the proposed ANN method in this work could be used to predict the binodal curve of H2O/DMAc/PSf system in the temperature range from 20 to 60oC. From this work, it can be concluded that the properly selected and trained network provides an effective method for prediction of the liquidliquid equilibrium properties of H2O/DMAc/PSf system, which could be used for guiding experimental researches in preparation of asymmetry polymer membrane.