EDXRF与机器学习结合用于含重金属电镀污泥的快速分类识别  

Rapid Classification and Identification of Heavy Metal-Containing Electroplating Sludge by Combining EDXRF With Machine Learning

作　　者：李伟艳 滕婧 郑智慧 石靖靖[4] 石垚 李志宏 张晨牧 LI Wei-yan;TENG Jing;ZHENG Zhi-hui;SHI Jing-jing;SHI Yao;LI Zhi-hong;ZHANG Chen-mu(School of Electrical Engineering,Tongling University,Tongling 244061,China;School of Medicine and Healthcare,Guangxi Vocational and Technical College of Industry,Nanning 530001,China;School of Chemistry and Chemical Engineering,Chongqing University,Chongqing 401331,China;CAS Key Laboratory of Green Process and Engineering,National Engineering Research Center of Green Recycling for Strategic Metal Resources,Institute of Process Engineering,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]铜陵学院电气工程学院,安徽铜陵244061 [2]广西工业职业技术学院医药康养学院,广西南宁530001 [3]重庆大学化学化工学院,重庆401331 [4]中国科学院过程工程研究所绿色过程与工程重点实验室,战略金属资源绿色循环利用国家工程研究中心,北京100190

出　　处：《光谱学与光谱分析》2025年第5期1283-1289,共7页Spectroscopy and Spectral Analysis

基　　金：国家重点研发计划项目(2022YFC3901502);2024年铜陵学院人才科研启动基金项目(2024tlxyrc054);国家自然科学基金项目(52100215)资助。

摘　　要：电镀污泥等含重金属类危险废物的快速分类识别对区域生态环境质量监管至关重要。采用课题组自研发的能量色散型X射线荧光光谱仪(EDXRF),采集了东莞市100余家企业的八种不同类型电镀污泥的光谱信息。对谱图信息进行降噪和标准化处理后识别出关键分类因子,并将其作为输入变量;对不同机器学习模型进行训练比较,确定了含重金属电镀污泥X荧光快速分类识别的最佳方法体系。结果表明,铁、铜、镍、锌、铅、钙6种金属元素的特征谱线信号可作为电镀污泥的关键分类因子,尽管随机森林(RF)、支持向量机(SVM)、线性判别(LDA)三种机器学习模型经训练后均能实现X荧光光谱对电镀污泥的准确分类识别,但只有RF模型的准确度、精确度和灵敏度均可达到100%。机器学习与EDXRF技术的结合,能够有效解决传统含重金属类危险废物识别方法所存在的耗时长、时效性差等关键问题。在未来,这一结合在土壤重金属污染快速溯源以及重金属危险废物快速鉴别等生态环境监测管理领域具有广泛的应用前景。The rapid identification,classification,and pollution source tracing of hazardous wastes containing heavy metals is crucial to regional ecological and environmental quality supervision.This study used the energy-based X-ray fluorescence spectroscopy device(EDXRF)self-developed by the research group to collect spectral information of 8 different types of electroplating sludge from over 100 companies in Dongguan City.After spectral information noise reduction and data standardization,key classification factors were identified and used as input variables.The best method system for rapid X-fluorescence classification and identification of electroplating sludge containing heavy metals was determined through training and comparison of different machine models.The results show that the characteristic spectral line signals corresponding to the six metal elements of iron,copper,nickel,zinc,lead,and calcium can be used as a key factor to distinguish different types of electroplating sludge.Although random forest(RF),support vector machine(SVM),and linear discriminant(LDA)could achieve accurate classification and identification of electroplating sludge using X-ray fluorescence spectrum,only the RF model achieves 100%accuracy,precision,and sensitivity.The combination of machine learning and EDXRF technology can solve key problems such as the long,time-consuming,and poor timeliness of traditional chemical analysis methods for identifying hazardous wastes containing heavy metals.In the future,it will have broad application prospects in ecological environment monitoring and management such as rapid traceability of heavy metal pollution in soil and rapid identification of hazardous wastes containing heavy metals.

关 键 词：电镀污泥 能量色散型X射线荧光光谱仪 机器学习 快速分类识别 

分 类 号：O657.3[理学—分析化学]