遗传神经网络-分光光度法同时测定钨钼钛  

Genetic neural network-spectrophotometric determination of tungsten,molybdenum and titanium

作　　者：赵红霞[1] 李井会[1] 于洪梅[1] 陈国顺[1] 陈朝军[1] 张国民[2] 窦成[2] 陈平[1] 

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

出　　处：《冶金分析》2007年第4期18-21,共4页Metallurgical Analysis

基　　金：辽宁省教委资助项目(2004003)

摘　　要：建立了钨、钼、钛-5′-硝基水杨基荧光酮-CTMAB的同时测定新显色体系,用遗传神经网络解析钨、钼和钛的重叠光谱,分光光度法同时测定钨、钼、钛。在遗传神经网络中构造三组分解析的适应度函数,用作图法最后确定交叉概率和变异概率,并将遗传神经网络与小波神经网络的解析结果进行比较,表明遗传神经网络优于后者。将所建立的方法用于标准钢样中钨、钼、钛的测定,相对误差分别为1.83%,-0.09%和-1.92%。A new color system of tungsten（molybdenum, titanium） -5′-nitrosalicyl fluorone- CTMAB was developed, and the overlapped spectra of tungsten, molybdenum and titanium were resolved by genetic neural network in this paper. In the genetic neural network, a new fitness function was developed, crossover probability and mutation probability were determined by graphics method.A comparison of the analytical results between genetic neural network and wavelet neural network was made. Results showed that the genetic neural network was better than wavelet neural network. The method was used to determine the concentration of tungsten, molybdenum and titanium in standard steel sample, and the relative errors were 1.83 %, - 0.09 % and - 1.92 %, respectively.

关 键 词：钨 钼 钛 遗传神经网络 小波神经网络 5′-硝基水杨基荧光酮 

分 类 号：O657.3[理学—分析化学] TG115.33[理学—化学]