基于BP神经网络的膜蒸馏性能预测系统  

Prediction System of Membrane Distillation Performance Based on BP Neural Network

作　　者：杨超环 张诗曼 徐春燕 郝如意 YANG Chao-huan;ZHANG Shi-man;XU Chun-yan;HAO Ru-yi(College of Chemistry and Chemical Engineering,Hulunbuir University,Hulun Buir,Inner Mongolia 021000,China)

机构地区：[1]呼伦贝尔学院化学化工学院,内蒙古呼伦贝尔021000

出　　处：《浙江化工》2022年第12期34-37,46,共5页Zhejiang Chemical Industry

摘　　要：利用BP(反向传播)神经网络模型建立了气隙式膜蒸馏性能(通量和造水比)与热进料温度、冷进料温度和管膜比之间的映射关系,神经网络由输入层(三输入)、隐含层和输出层(二输出)构成,隐含层包含6个神经元。神经网络模型预测通量和造水比的决定系数分别为0.9981和0.9801,具有很高的预测精度。Pearson相关性分析显示热进料温度、管膜比与通量呈正相关,冷进料温度与通量呈负相关,热进料温度、冷进料温度及管膜比对造水比均有正向影响,其中热进料温度对造水比的正向影响最大。The mapping relationship between the performance of air gap membrane distillation(permeate flux and gained output ratio)and hot feed temperature,cold feed temperature and dense fiber-membrane ratio is established by using the BP(back-propagation)neural network model.The neural network is composed of input layer with three inputs,output layer with two outputs and hidden layer with six neurons.The determination coefficients of permeate flux and gained output ratio predicted by the neural network model were 0.9981 and 0.9801 respectively,with high prediction accuracy.Pearson correlation analysis showed that hot feed temperature and dense fiber-membrane ratio were positively related to permeate flux,while cold feed temperature was negatively correlated with permeate flux.Meanwhile,hot feed temperature,cold feed temperature and dense fiber-membrane ratio were positively correlated with gained output ratio,among which hot feed temperature had the greatest positive effect on gained output ratio.

关 键 词：气隙式膜蒸馏 BP神经网络 预测 

分 类 号：TQ028.8[化学工程] TP183[自动化与计算机技术—控制理论与控制工程]