基于X射线衍射特征提取的铜精矿原产地识别方法  

Identification of Copper Concentrate Origin Based on X-ray Diffraction Characteristics

作　　者：闵红 刘倩 王巧玲[3] 周海明 刘曙 邢彦军 MIN Hong;LIU Qian;WANG Qiao-Ling;ZHOU Hai-Ming;LIU Shu;XING Yan-Jun(Technical Center for Industrial Products and Raw Materials Inspection and Testing of Shanghai Customs District,Shanghai 200135;Donghua University,Shanghai 201620;Technology Center of Erlian Customs,Erlianhot 011100)

机构地区：[1]上海海关工业品与原材料检测技术中心,上海200135 [2]东华大学,上海201620 [3]二连海关技术中心,二连浩特011100

出　　处：《中国口岸科学技术》2023年第10期22-30,共9页China Port Science and Technology

基　　金：国家重点研发计划项目(2018YFF0215400)。

摘　　要：不同产地铜精矿在物相组成、元素含量上存在差异,开发入境铜精矿产地识别方法,为基于原产地验证的风险控制策略提供了可能,有助于铜精矿风险筛查和快速验放。本文建立了一种基于X射线衍射特征提取的铜精矿原产地识别方法。在X射线衍射谱图分类建模中,数据的冗余及共线性会严重影响模型的分类性能和稳健性,因此数据降维及特征提取是提高分类建模准确性的一种有效方法。本文采集了3个铜精矿主要来源国138批次样品的物相谱图,比较了主成分分析和随机森林特征重要性方法提取铜精矿X射线衍射光谱的特征数据,建立随机森林分类模型。结果表明,选取特征重要性前34个数据建立随机森林分类模型的准确率达94.28%,该方法与主成分载荷阈值相比,不仅有效地减少了特征输入变量的个数,而且可以达到较好的分类识别效果。X射线衍射分析技术具有分析速度快和稳定性好的优点,结合随机森林特征提取和分类算法,可以实现对铜精矿的原产地识别。Copper concentrates from different sources exhibit variations in their phase composition and elemental content.Developing a method for identifying the origin of imported copper concentrates provides a means for origin verification-based risk control strategies,aiding in the screening of copper concentrate quality and expeditious release by the customs.This study establishes an approach for origin identification of copper concentrates based on X-ray diffraction(XRD)characteristics.In XRD spectrum classification modeling,data redundancy and collinearity will significantly affect the classification performance and robustness of the model.Thus,data dimensionality reduction and feature extraction become effective techniques to improve the accuracy of classification modeling.This research collects XRD spectra of 138 batches of copper concentrates originating from three major source countries and compares principal component analysis(PCA)and random forest feature importance methods for extracting feature data from XRD spectra,ultimately constructing a random forest classification model.The results indicate that the random forest classification model built using the top 34 feature-importance-ranked variables achieves an accuracy of 94.28%.Compared with the principal component load threshold,this approach not only reduces the number of input feature variables effectively but also delivers a superior classification performance.XRD analysis technology has the advantages of fast analysis speed and good stability.By combining random forest feature extraction and classification algorithms,this approach facilitates the identification of the origin of copper concentrates.

关 键 词：铜精矿 X射线衍射光谱 特征提取 随机森林 产地识别 

分 类 号：TD926.3[矿业工程—选矿] O434.19[机械工程—光学工程]