VLBP神经网络在能量色散X荧光定量分析中的应用  

Application of VLBP Neural Network in Energy Dispersion X Fluorescence Quantitative Analysis

作　　者：何庆驹 葛良全[1] 李飞[1] 卢恒 温自强[1] 覃建强 HE Qing-ju;GE Liang-quan;LI Fei;LU Heng;WEN Zi-qiang;QIN Jian-qiang(Chengdu University of Technology,Chengdu 610059,China;SuiNing Environmental Monitoring Center,SuiNing Sichuan 629000,China)

机构地区：[1]成都理工大学核技术与自动化学院,成都610059 [2]四川遂宁环境监测中心,四川遂宁629000

出　　处：《核电子学与探测技术》2020年第4期610-615,共6页Nuclear Electronics & Detection Technology

基　　金：国家重点研发计划项目(2017YFC0602105);国家自然科学基金航空(41774190)资助。

摘　　要：基于基础BP神经网络,将VLBP神经网络应用在能量色散X荧光分析中。并分别采用基础BP、VLBP两个模型,对同一批实测铅锌矿样本进行预测,证明VLBP算法在定量分析中的优势。然后通过使用VLBP算法对铅锌矿石样品的Zn元素含量进行了预测,并与样品化学分析值对比。结果表明:预测值与化学分析值的相对误差小于5%。从样本中选取特征峰计数超过训练范围的样品进行预测,预测值与参考值的相对误差小于5%,可作为一个新型有效的方法应用在地质样品元素定量分析:领域。Based on the hasic BP neural network,the VLBP neural network is applied in energy dispersive X-ray fluorescence analysis.The basic BP and VLBP models were used to predict the same batch of measured lead-zinc samples,which proved the advantage of VLBP algorithm in quantitative analysis.The Zn element content of the lead-zinc ore sample was predicted by using the VLBP algorithm and compared with the chemical analysis value of the sample.The results show that the relative error between the predicted value and the chemical analysis value is less than 5%.The sample with the characteristic peak count exceeding the training range is vSelected for prediction.The relative error between the predicted value and the reference value is less than 5%,which can be used as a new and effective method in the field of quantitative analysis of geological vSample elements.

关 键 词：EDXRF VLBP神经网络 定量分析 

分 类 号：TL99[核科学技术—核技术及应用]