基于稀土元素的软玉产地溯源研究  

Study on the origin traceability of nephrite based on rare earth elements

作　　者：崔中良 郭心雨 王嘉宇 郭钢阳 CUI Zhongliang;GUO Xinyu;WANG Jiayu;GUO Gangyang(Jiangxi Institute of Applied Science and Technology,Nanchang Jiangxi 330100,China;Henan Third Geological Exploration Institute Co.,Ltd.,Zhengzhou Henan 450014,China)

机构地区：[1]江西应用科技学院,江西南昌330100 [2]河南省第三地质勘查院有限公司,河南郑州450014

出　　处：《化工矿物与加工》2024年第5期30-41,共12页Industrial Minerals & Processing

基　　金：国家岩矿化石标本资源库(NCSTI-RMF202301)。

摘　　要：近年来,软玉产地溯源研究受到了国内外学者的广泛关注,其中稀土元素在开展软玉产地溯源方面潜力巨大。基于我国青海、广西、贵州、辽宁、新疆5个产地软玉的14种稀土元素,建立了软玉产地溯源判别模型,研究结果表明:Fisher判别模型、多元Logistic判别模型、MLP神经网络判别模型,回代检验判别准确率分别为89.7%、100.0%、100.0%;PCA-多项式判别模型、PCA-Fisher判别模型、PCA-多元Logistic判别模型、PCA-MLP神经网络判别模型,回代检验判别准确率分别为51.5%、42.6%、48.5%、72.1%;最优判别模型为多元Logistic判别模型和MLP神经网络判别模型。In recent years,study on the origin traceability of nephrite has received widespread attention from scholars at home and abroad,with rare earth elements having great potential in carrying out the origin traceability of nephrite.The traceability discrimination model for nephrite production was established based on 14 rare earth elements from 5 production areas including Qinghai,Guangxi,Guizhou,Liaoning and Xinjiang.The results showed that the Fisher discriminant model,multivariate logistic discriminant model and MLP neural network discriminant model had the returned validation accuracy of 89.7%,100.0%and 100.0%,respectively;The returned validation accuracy of PCA-polynomial discriminant model,PCA-Fisher discriminant model,PCA-multivariate logistic discriminant model,and PCA-MLP neural network discriminant model were 51.5%,42.6%,48.5%and 72.1%,respectively;The optimal discriminant models are the multivariate logistic discriminant model and the MLP neural network discriminant model.

关 键 词：软玉 稀土元素 产地溯源 主成分分析 多元Logistic FISHER判别 

分 类 号：P617[天文地球—矿床学]