人工神经网络光度法同时测定饮料中三种色素  

Simultaneous determination of three kinds of coloring matter in beverage by ANN spectropotometry

作　　者：王志有[1] 于洪梅[1] 吴秀红[1] 戴洁莹[1] 

机构地区：[1]鞍山科技大学化学工程学院,辽宁鞍山114044

出　　处：《鞍山科技大学学报》2005年第3期187-190,共4页Journal of Anshan University of Science and Technology

基　　金：辽宁省教育厅资助项目(2004003)。

摘　　要：用BP人工神经网络解析了日落黄、柠檬黄和胭脂红的吸收光谱,提出同时测定这三种色素的计算分析方法;并对饮料中的日落黄、柠檬黄和胭脂红进行了同时测定。日落黄、柠檬黄和胭脂红回收率分别为99 61%、103 45%和98 50%;使用了改进的BP算法,避免了神经网络学习过程中可能产生的麻痹现象;提出了目标向量的简单变换方法,与偏最小二乘法进行了比较。结果表明,BP神经网络法与偏最小二乘法一样能获得满意的分析结果。The spectra of sunset yellow,tartrazine and ponceau 4R were analyzed by ANN-BP algorithm,and the computation and analyzing method of simultaneously determinating three kinds of coloring matter was proposed.And sunset yellow,tartrazine and ponceau 4R in beverage were simultaneously determinated,the mean recoveries of sunset yellow,tartrazine and ponceau 4R were 99.61%,103.45% and 98.50%,respectively.The paralysis in the procedure of ANN training was avoided by the improved BP algorithm and a simple transformation method for target vectors was put forward.Compared the results from ANN-BP spectrophotomeytry with that from partial least square spectrophotomeytry,it shows that BP neural network can obtain satisfactory analyzed result just like partial peast squares do.

关 键 词：人工神经网络 偏最小二乘法 分光光度法 日落黄 柠檬黄 胭脂红 

分 类 号：O657.32[理学—分析化学]