神经网络法用于多元混酸同时测定

PRACTICAL SIMULTANEOUS DETERMINATION OF MULTICOMPONENT ACIDS BY USING NEURAL NETWORKS(NN)

利用多层神经网络误差反向传播算法处理酸碱电位滴定数据，求算出多元混合酸各组分的浓度．优化了神经网络的结构和参数，测定了三组分有机酸混合样品，结果良好，平均相对偏差ＲＳＤ≤４％．

Multilayer artificial neural network (NN) was systematically investigated as applied to titrimetic analysis. A modified backpropagation (MBP) was developed for training NN and was used to process the data obtained by acid-basic potential titrimetry and to estimate the concentrations of each acids in the multicomponent analytical system. The algorithm and the topological structure of NN were optimilized. The proposed method was applied to practical simultaneous determination of three organic acids in their mixture samples with good results. The average error and/or deviation is less than 4%.