BP神经网络预测EDXRF中铁、钛元素含量  

Fe and Ti content Prediction in X-Ray Fluorescence Measurement Based on BP Neural Network

作　　者：何泽[1] 颜瑜成 刘祥和[1] 梁超[1] 雷洲阳 刘明哲[3] HE Ze YAN Yu- cheng LIU Xiang- he LIANG Chao LEI Zhou - yang LIU Ming - zhe(Chengdu University of Technology College of Engineering Technology, Leshan Sichuan 614000, China Southwest Institute of Physics, Chengdu 610225, China College of Applied Nuclear Technique in Geosciences Key Laboratory of Sichuan Province, Chengdu 610059, China)

机构地区：[1]成都理工大学工程技术学院,四川乐山614000 [2]西南物理研究院,成都610225 [3]成都理工大学四川省地学核技术重点实验室,成都610059

出　　处：《核电子学与探测技术》2016年第12期1192-1195,共4页Nuclear Electronics & Detection Technology

基　　金：国家自然科学基金(41274109);四川省青年科技创新团队(2015JTD0020)资助

摘　　要：为了更好地定量分析矿石样品中铁、钛元素的含量,应用EDXRF分析技术建立了一个基于BP神经网络的预测模型。将矿石样品在EDXRF光谱仪中测得的荧光强度计数作为BP神经网络模型的输入变量,对该模型进行训练和检测。实验结果表明:该BP神经网络预测模型能获得较精确的结果,预测铁含量结果的相对误差不大于2.4%;预测钛含量结果的最大相对误差不大于2.3%;可用于地质矿石样品元素含量的分析预测。In order to better quantify the content of iron and titanium of mineral samples, this paper builds a prediction model based on the BP neural networks as well as the energy - dispersive x - ray fluorescence （EDXRF） teehnology. The fluorescence strength counts that measured during the x - ray analysis are set as input for training and testing the model. The experiment results show that this neural network is able to obtain a precise result, the biggest error for iron is 2.4%, and the smallest is 0.33% ; 2.3% and 0.24% for titanium respectively. This experiment is easy to operate and have a significant effect, which provides an easy and high efficiency method for prediction analysis in the mineral samples at mining and manufacture industry.

关 键 词：铁 钛元素含量 预测模型 EDXRF BP神经网络 

分 类 号：O657.34[理学—分析化学]