摘要
运用神经网络与分光光度结合的方法,提出了同时分析六氟化铀中Mo、Ti和W3种杂质元素的技术。训练网络确定的最佳训练参数为学习速率为0.60,动量项为0.85。检测集测定的六氟化铀中Mo、Ti和W3种元素的回收率均在90%～110%之间,方法精密度分别为3.3%、3.9%、4.3%,对实际样品进行测定,并与ICP-AES法结果相比较,两者间相对偏差均小于10%。
A method of combination neural networks tool and spectrophotometry method simultaneously determining three impurities molybdenum,titanium and tungsten in uranium hexafluoride is established in this paper.The best parameters are selected by training the artificial neural networks:learning rate is 0.60,momentum is 0.85.By estimating the inspecting set,the recoveries are from 90% to 110% and the precisions are 3.3%,3.9% and 4.3% respectively for molybdenum,titanium and tungsten.By determining the samples of uranium hexafluoride and comparing with ICP-AES,the relative deviations are lower than 10%.
出处
《新技术新工艺》
2008年第12期113-116,共4页
New Technology & New Process
关键词
神经网络
分光光度法
同时测定
neural network
spectrophotometry
simultaneous determination
