基于BP神经网络的精对苯二甲酸回收系统水含量预测模型

采用电导率法测量精对苯二甲酸回收系统水含量,考察了电导率与水含量、金属离子浓度及温度的关系。以温度、金属离子浓度、电导率为输入变量,通过BP人工神经网络贝叶斯正则化算法建立水含量预测模型。优化后的BP神经网络模型结构为3-13-1,动量因子为0.75。使用优化的模型对水含量进行预测,测试集最大绝对相对偏差为4.36%,平均绝对相对偏差为0.96%,表明所建立的神经网络模型可较好地用于预测精对苯二甲酸回收系统的水含量。

Prediction model for moisture content of purified terephthalic acid recovery system based on BP neural network

The moisture content of purified terephthalic acid recovery system was measured by electrical conductivity technique.Effects of water concentration,metal ion concentration and temperature on electrical conductivity were investigated.Using temperature,metal ion concentration and electrical conductivity as input variables,a model was established to predict moisture content based on the bayesian regularization algorithm of BP neural network.The optimized structure for BP neural network model was fixed on 3-13-1,and the momentum constant was 0.75.The maximum absolute relative error and the mean absolute relative error for testing set were 4.36% and 0.96%,it indicated that the model could be used to estimate the water concentration of purified terephthalic acid recovery system.