首页|激光诱导击穿光谱结合随机森林的铌钽矿中铀钍元素定量分析

激光诱导击穿光谱结合随机森林的铌钽矿中铀钍元素定量分析

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
铌钽矿石中富含的铀(U)和钍(Th)等放射性元素含量的分析测定是铌钽矿开发过程中的重要环节,对于辐射安全评估和环境保护具有重要的意义.本文基于激光诱导击穿光谱(LIBS)技术结合随机森林(RF)算法,在实验室条件下开展了铌钽矿石中放射性元素U、Th含量的定量分析工作.利用传统高纯锗(HPGe)γ能谱仪对11个铌钽矿石样品进行U、Th含量的测定,并设为LIBS的参考值.其中的9个样品设置为训练集,用于RF的建模;余下2个样品设为测试集用于模型验证.采用全谱面积归一化后的LIBS光谱及其相应含量作为RF算法的数据集,首先通过五折交叉验证对RF的决策树数量参数进行了优化,再结合重要性得分阈值设定探讨了 LIBS光谱的波长特征的筛选对于预测结果的影响.结果显示,结合波长特征筛选,以优化的决策树参数建立RF模型,训练集中U和Th均方根误差(RMSEC)分别为48和313 μg/g;对于2#、6 #样品中U和Th含量的预测,测试集均方根误差(RMSEV)分别为141、209 µg/g和750、914 μg/g;两种元素多次测量预测值的相对标准偏差(RSD)均在7％以内,且多次测量平均值与γ谱仪测量值相比,两种元素的预测相对误差(RE)均在8％以内.上述结果表明LIBS技术结合RF算法可以有效地实现铌钽矿石中放射性元素U和Th的定量分析,为铌钽矿石的精准开采与评估提供参考.
Quantitative analysis of uranium and thorium in niobium-tantalum ore by laser-induced breakdown spectroscopy combined with random forest
The analysis and determination of radioactive elements including uranium(U)and thorium(Th)in niobium-tantalum ores is a key step in the exploitation process of niobium-tantalum ore,and it has im-portant significance for the radiation safety assessment and environmental protection.In this study,the quantitative analysis of radioactive U and Th in niobium-tantalum ore was investigated in the laboratory based on laser-induced breakdown spectroscopy(LIBS)combined with random forest(RF)algorithm.The contents of U and Th in eleven niobium-tantalum ore samples were determined using the traditional high purity germanium(HPGe)γ spectrometer,which were used as the reference values for LIBS.Nine of the samples were set as the training set for RF modeling,while the remaining two samples were used as the test set for model validation.The LIBS spectra after full spectral area normalization and the corresponding contents were used as the dataset for the RF algorithm.Firstly,the number of decision trees parameter of RF was optimized through five-fold cross-validation.Then,the influence of wavelength feature selection of LIBS spectra on the prediction results was explored by setting the importance score threshold.The results showed that,by combining the wavelength feature selection and optimizing the decision tree parameters,the root mean square error of calibration(RMSEC)of U and Th in the training set was 48 μg/g and 313μg/g,respectively.For the prediction of U and Th contents in 2# and 6 # samples,the root mean square error of validation(RMSEV)of test set was 141 μg/g and 209 μg/g for U,and 750 μg/g and 914 μg/g for Th,respectively.The relative standard deviations(RSD)of multiple measurement results for both ele-ments were both within 7％.Moreover,the average value of multiple measurements was compared with the measurements by the γ spectrometer,and the relative error(RE)of prediction for two elements was within 8％.These results indicated that the combination of LIBS technology with the RF algorithm could effectively achieve the quantitative analysis of radioactive U and Th in niobium-tantalum ores,which pro-vided a reference for the accurate mining and evaluation of niobium-tantalum ores.

laser-induced breakdown spectroscopy(LIBS)niobium-tantalum oreuraniumthoriumquantitative analysisrandom forest(RF)

彭玲玲、谢绍荣、孟祥厅、焦宝宝、刘林、刘小亮

展开 >

东华理工大学核技术应用创新联合实验室,江西南昌 330013

江西省地质局实验测试大队,江西南昌 330013

先进核能技术设计与安全教育部重点实验室,湖南衡阳 421001

激光诱导击穿光谱(LIBS) 铌钽矿 铀 钍 定量分析 随机森林(RF)

国家自然科学基金青年基金国家自然科学基金青年基金江西省自然科学基金先进核能技术设计与安全教育部重点实验室项目放射性地质与勘探技术国防重点学科实验室开放基金东华理工大学博士启动基金东华理工大学博士启动基金

123650211200503720224BAB201021KLANEDS2023142022RGET19DHBK2019153DHBK2019158

2024

10.13228/j.boyuan.issn1000-7571.012496

冶金分析
中国钢研科技集团有限公司（钢铁研究总院） 中国金属学会

冶金分析

CSTPCD北大核心
影响因子：1.124
ISSN：1000-7571
年,卷(期)：2024.44(5)