反向传播人工神经网络分光光度法同时测定环境水样中的苯酚、间苯二酚和间氨基酚

Spectrophotometric Simultaneous Determination of Phenol,Resorcinol and m-Aminophenol in Environmental Wastewater by BP-ANN

本文应用人工神经网络原理 ,采用误差反向传播算法 ,对环境水样中的苯酚、间苯二酚和间氨基酚可以用分光光度法不经分离进行了同时测定。三种酚类的平均回收率分别为 98 0 % ,99 6 %和 99 7%。实验证明 ,反向传播 人工神经网络方法应用在本体系中进行结果校正 ,结果令人满意。

A spectrophotometric system for simultaneous determination of phenol, resorcinol, and m-aminophenol has been studied. A back-propagation artificial neural network (BP-ANN) method has been used in this system and the net parameters we chose in this training programme were the most applicable ones (node = 10, eta = 0.25, alpha = 0. 75 and the time of training was about 5,000) so that we could obtain the satisfied results. The experimental conditions were optimized by a chemometric method called General Revolving Combination Design to describe the relationship between the absorbance responses and the quantity of potassium ferric cyanide, 4-amino antipyrine and the pH values. The data obtained from the BP-ANN method were compared with the Kalman filtering method that was applied successfully to the linear system and the results showed that the BP-ANN method is better than the KF method because there are many nonlinear factors, in this experimental system. The average relative errors of KF and the BP-ANN methods were about 19.67% and 3.74% respectively. The average recoveries of each component ranged from 98.0% to 99.%. The approach has been applied to the analysis of the three components in environmental wastewater.