人工神经网络应用于维生素B族4组分同时测定

Application of artificial neural network to simultaneous determination of four B-group vitamins

应用人工神经网络原理 ,采用 Levenberg- Marquardt BP算法 ,对于吸收光谱严重重叠的维生素B1,B2 ,B6和烟酰胺 4组分同时进行含量测定 .四者的平均回收率分别为 10 0 % ,98% ,96 % ,10 0 % .该方法用于复合维生素 B片剂中 B1,B2 ,B6和烟酰胺的同时测定 ,获得较满意的结果 .实验表明 ,该方法与现有的算法相比具有训练速度快、预测结果准确度高等特点 ,有望能成为多组分分析的有效方法之一 .

By means of artificial neural network and the Levenberg Marquardt back propagation train algorithm, the multicomponents of vitamin B 1,B 2,B 6 and nicotinamide were determined simultaneously, though their spectra are overlapped. The mean recovery of vitamin B 1,B 2,B 6 and nicotinamide were 100%,98%,96% and 100% respectively.The method has been applied to the determination of vitamin B compound tablets with satisfactory results. The results were better than those of other networks in terms of training speed and the accuracy. In conclusion ,the new network spectrophotometry is a good choice for resolving multicomponent.