基于X射线荧光光谱(XRF)-Pb同位素与深度学习的铜精矿产地溯源  

Origin Tracing of Copper Concentrates Based on X-ray Fluorescence Spectrometry(XRF)-Pb Isotopes and Deep Learning Techniques

作　　者：赵伟 杨涛[2] 唐晨 封亚辉 郑建明 ZHAO Wei;YANG Tao;TANG Chen;FENG Yahui;ZHENG Jianming(Industrial Product Testing Center,Nanjing Customs,Nanjing,Jiangsu 210019,China;Nanjing University,Nanjing,Jiangsu 210046,China)

机构地区：[1]南京海关工业产品检测中心,南京210019 [2]南京大学,南京210046

出　　处：《中国无机分析化学》2025年第5期667-675,共9页Chinese Journal of Inorganic Analytical Chemistry

基　　金：海关总署科研项目(2020HK25);南京海关科技项目(2022KJ08)。

摘　　要：铜精矿产地溯源研究在保障资源供应链安全、优化冶炼工艺控制、履行ESG监管责任、打击贸易欺诈行为、推动矿业数字化转型等方面具有重要意义。选取了3个铜精矿的典型产地,每个产地在一定的时间跨度内随机选取了8个样品。首先,通过15种元素的X射线荧光光谱(XRF)分析和3种铅稳定同位素分析手段获取关键指标信息。再次,采用Python Scikit-learn对数据集进行分析,建立分类预测模型并评价预测准确率。结果表明,3个产地的铜精矿检测结果存在一定的差异,对原始数据和标准化处理后的数据分别建模评估,数据标准化处理后的模型预测准确率有着显著提高,使用同位素和微量元素标准化处理后的数据集建立的KNN或SVM模型中,测试集样品能够实现100%正确分类。利用分析化学与机器学习的技术交叉优势,其标准化处理-特征选择-模型验证的技术闭环,为矿产资源数字护照体系建设提供了可复用的技术框架。为铜精矿供应链贸易合规监管提供支撑,为冶炼工艺参数优化提供数据支持,为类似矿物分析提供方法参考以及推动传统矿业向数据驱动模式转型。Research on the origin tracing of copper concentrate is crucial for ensuring the security of the resource supply chain,optimizing smelting process control,fulfilling ESG regulatory obligations,combating trade fraud,and driving the digital transformation of the mining industry.In this study,three representative production areas of copper concentrate were chosen,and 8 samples were randomly collected from each area over a specific time period.Initially,key indices were derived through X-ray fluorescence spectrometry(XRF) analysis of 15 elements and three lead stable isotope analysis methods.Subsequently,Python Scikit-learn was utilized to analyze the dataset,construct a classification prediction model,and assess the prediction accuracy.The results indicate distinct differences in the testing of copper concentrates from the three production areas.Separate modeling and evaluation were conducted on the original data and the standardized data.The prediction accuracy of the model was notably enhanced after data standardization.Under various model parameters,the overall discrimination accuracy was high,reaching an ideal level for classification discrimination.In the KNN or SVM models constructed using the isotope and trace element normalized datasets,the test set samples were classified with 100% accuracy.The application of this method supports the compliant supervision of copper concentrate supply chain trade,offers data support for optimizing smelting process parameters,provides a methodological reference for the analysis of similar minerals,and promotes the transformation of the traditional mining industry towards a data-driven model.Its uniqueness lies in integrating the cross-cutting advantages of analytical chemistry and machine learning technologies.The technical loop of standardized processing-feature selection-model validation provides a reusable technical framework for building the digital passport system for mineral resources.

关 键 词：铜精矿 产地溯源 深度学习 X射线荧光光谱 铅稳定同位素 

分 类 号：O657.34[理学—分析化学]