遗传神经网络用于分光光度法同时测定钢中镧和铈  

APPLICATION OF GENETIC NEURAL NETWORK TO THE SIMULTANEOUS SPECTROPHOTOMETRIC DETERMINATION OF LANTHANUM AND CERIUM IN STEEL

作　　者：彭丹[1] 李井会[1] 于洪梅[1] 吴秀红[1] 陈国顺[1] 赵红霞[1] 张国民[2] 窦成[2] 

机构地区：[1]鞍山科技大学化工学院,鞍山114044 [2]鞍山钢铁集团公司中心实验室,鞍山114021

出　　处：《理化检验（化学分册）》2006年第1期9-11,15,共4页Physical Testing and Chemical Analysis(Part B:Chemical Analysis)

基　　金：辽宁省教委资助项目(2004003)

摘　　要：应用遗传算法自适应概率搜索神经网络中隐含层结点数、学习速率、动量因子,无需大量训练样本,就可以使三者达到最优匹配,优化网络结构和参数,建立一种遗传神经网络算法,解决了BP神经网络过拟合问题。在镧(铈)-间磺酸基偶氮氯膦新的同时测定显色体系中,应用遗传神经网络同时测定钢中镧和铈。镧和铈配合物的表观摩尔吸光系数分别为5.03×104和7.26×104L.mol-1.cm-1。A new chemometric method, the genetic neural network （GA-ANN） was proposed to optimize the structure and parameters of BP-ANN （artificial neural network with back-propagation algorithm）. The self-adaptive probability of genetic algorithm was applied to search for the node number in the hidden layer, the learning rate （η） and the momentum factor （α）. Optimal matching of the triplicate was achieved without the necessity of large quantity of training samples. The drawbacks of BP-ANN were overcome by GA-ANN. The optimized GA-ANN was successfully applied to the simultaneous determination of lanthanum and cerium in steel by CPA-mS- spectrophotometry. The apparent molar absorptivity for La and Ce were found to be 5.03×10^4 and 7.26×10^4L·mol^-1·cm^-1 respectively.

关 键 词：分光光度法 遗传神经网络 间磺酸基偶氮氯膦 镧 铈 铜样 

分 类 号：O657.31[理学—分析化学]