摘要
针对常见的10类固体废物开展X射线荧光(XRF)与X射线衍射(XRD)实验,通过非参数检验提取固体废物的XRF与XRD指纹特征,并采用归一化的方式结合两种指纹特征形成XRF-XRD耦合光谱指纹特征,以三种指纹特征为数据集训练机器学习分类模型,发现随机森林模型分类识别效果最好,验证集分类精确度、召回率、准确率分别达98.0%、97.5%、97.6%,证实耦合光谱指纹特征分类效果优于单一光谱指纹特征.形成了固体废物XRF-XRD多维指纹特征提取方法,为固体废物指纹特征识别奠定方法基础.
This study conducted X-ray fluorescence(XRF)and X-ray diffraction(XRD)experiments on 10 common types of solid waste.By employing non-parametric tests,the XRF and XRD fingerprint characteristics of solid waste were extracted.A normalized approach was adopted to combine the two fingerprint characteristics to form XRF-XRD coupled spectral fingerprint characteristics.A machine learning classification model was trained using the three fingerprint characteristics as a dataset.It was found that the random forest model exhibited the best classification recognition performance,with classification accuracy,recall rate,and precision rate in the validation set reaching 98.0%,97.5%,and 97.6%,respectively.This confirmed that the classification performance of coupled spectral fingerprint characteristics was superior to that of single spectral fingerprint characteristics.A method for extracting multi-dimensional fingerprint characteristics of solid waste using XRF-XRD was developed,laying a methodological foundation for the identification of fingerprint characteristics of solid waste.
作者
黄瑞潇
卢永琦
郑志敏
杨玉飞
杨金忠
黄启飞
HUANG Rui-xiao;LU Yong-qi;ZHENG Zhi-min;YANG Yu-fei;YANG Jin-zhong;HUANG Qi-fei(Research Institute of Solid Waste Management,State Key Laboratory of Environmental Criteria and Risk Assessment,Chinese Research Academy of Environment Sciences,Beijing 100012,China)
出处
《中国环境科学》
北大核心
2025年第10期5585-5595,共11页
China Environmental Science
基金
国家重点研发计划项目(2024YFC3906401)
黄河流域生态保护和高质量发展联合研究项目(2022-YRUC-01-0303)。
