遗传神经网络-分光光度法同时测定钴锰铜镍  

Simultaneous determination of cobalt,manganese,copper and nickel by genetic neural network-spectrophotometry

作　　者：陈平[1] 李井会[1] 于洪梅[1] 董林[1] 褚宁[1] 张国民[2] 

机构地区：[1]辽宁科技大学,辽宁鞍山114044 [2]鞍山钢铁公司中心实验室,辽宁鞍山114021

出　　处：《冶金分析》2008年第1期31-34,共4页Metallurgical Analysis

基　　金：辽宁省教育厅资助项目(2004003)

摘　　要：建立了钴(锰、铜和镍)-间氯偶氮安替比林4组分同时测定的显色体系,并用于光度分析。4组分吸收光谱严重重叠,为解析重叠光谱,将BP神经网络的隐含层节点数、学习速率(η)和动量因子(α)二进制编码作为个体,再由个体组成群体,进行遗传操作,由神经网络输出结果的误差构造适应度函数,建立了遗传神经网络(GA-ANN-BP)算法,用于分光光度法不经分离同时测定钴、锰、铜和镍,合成样品的相对标准误差分别为0.68%,2.28%,2.47%,1.67%。方法用于标准样品的分析,分析结果相对误差小于5%,测定值与认定值相符。A new colour system of cobalt （manganese, copper, nickel）-m-chloroantipyrylazo was developed spectrophotometrically. The spectra of four components overlapped seriouly. In order to analyzed the overlapped spetra, the node numbers in hidden layer, learning rate （η） and momentum coefficient （α） of artificial neural network with back-propagation algorithm （BP-ANN） were transformed into binary code as individuals, and then the individuals formed into groups, and were operated by genetic operators. The fitness function was formed by error of neural network output results, and a genetic neural network （GA-BP-ANN） was developed. The method was applied to simultaneous determination of cobalt, manganese, copper and nickel in steel by spectrophotometry without separation. The relative standard deviation of synthetic samples was 0. 68%, 2.28%, 2.47%, 1.67%. The proposed method has been applied to the determination of CRMs with the relative errors less than 5 %, and the results are consistent with the certified ones.

关 键 词：遗传神经网络 分光光度法 间氯偶氮安替比林 钴 锰 铜 镍 

分 类 号：O657.3[理学—分析化学]