纳滤膜对脂肪族及杂环有机物的截留特性

分别选用NF90、NF270和NF-型号的纳滤膜测定了20种脂肪族及杂环有机物的截留率®,并分析了纳滤膜截留效果的影响因素.结果表明,脂肪族及杂环有机物的R受到分子结构和膜特性的影响:对于同分异构体,分枝结构越多,R越高;环状有机物与分子量相近的直链有机物相比,R明显偏高;孔径越小的纳滤膜R越高.利用遗传算法(GA)结合偏最小二乘回归(PLS)和人工神经网络法(ANN)建立了脂肪族及杂环物质的R与其结构的定量关系模型,2种方法建立的模型相关系数可分别达到0.8809和0.9944,通过2种模型进一步分析了R的影响规律,并对几种物质的R进行了有效的预测.从预测结果来看,GA-ANN模型的预测精度要好于GA-PLS模型.

Rejection characteristics of aliphatic and heterocyclic organisms by nanofiltration membranes

Nanofiltration membrane NF90, NF270 and NF- modes were selected separately to determine the rejection rates of 20 kinds of aliphatic and heterocyclic organism (AHO); and the influence factors of nanofiltration rejection effect were analyzed. The rejection rates of AHO were influenced by the molecular structures and characteristics of membranes. On the isomers, the more the branch structures, the higher the rejection rates; the cyclic organism compared with the straight chain organism of similar molecular weights, the rejection rates were higher obviously; the smaller the pore radius of the membrane, the higher the rejection rate. The quantitative structure-property relationships (QSPR) model with the rejection rate of AHO was established, utilizing least square regression (PLS) and artificial neural networks (ANN) combined with genetic algorithm (GA); the correlation coefficients of model established by these two techniques could reach 0.8809 and 0.9944 respectively. Through these two models, the influence rule of the rejection rates was further analyzed and the rejection rates of certain AHO were predicted effectively. According to the predicted results, the precision in predicting by GA-ANN model was better than that by GA-PLS.