基于大数据的黄土覆盖区下伏多金属矿产资源定量预测高精度模型  

High precision model for quantitative prediction of underlying polymetallic mineral resources in loess covered area based on big data

作　　者：张悦秋 牛琳 居维伟 刘小胡 肖娥 ZHANG Yueqiu;NIU Lin;JU Weiwei;LIU Xiaohu;XIAO E(Jiangsu Geological Bureau,Nanjing 210001,China)

机构地区：[1]江苏省地质局,南京210001

出　　处：《矿冶》2023年第5期8-14,共7页Mining And Metallurgy

摘　　要：为降低黄土覆盖区下伏多金属矿的开采难度,提出基于大数据的黄土覆盖区下伏多金属矿产资源定量预测高精度模型,合理预测隐伏矿区的金属矿产资源分布情况。构建矿产资源定量预测模型,向模型中输入二维GIS图层与三维实体地质模型,从输入的海量数据中提取成矿有利信息,利用主成分分析算法将多类型的成矿有利信息并行处理为统一格式,并按照数据类别依次存放在相应的分布式数据库中,通过非负矩阵分解算法向数据库中的数据增加非负约束条件,实现矿产资源数据分解,利用基于NMF算法的模糊C均值分类器对该数据进行分类,获得最佳矿产资源样本数据,完成黄土覆盖区下伏多金属矿产资源定量预测。经试验验证:该模型可精准预测黄土覆盖区下伏多金属矿区每个测点的金属量与金属密度,还可按照总金属量划分靶区规模,并能够预测出每个测点下的金属矿产储量以及成矿有利度。In order to reduce the difficulty of mining polymetallic ore under loess overlying area,a quantitative prediction model of polymetallic mineral resources under loess overlying area based on big data was proposed to reasonably predict the distribution of metal mineral resources in hidden mining area.Build the model of quantitative prediction of mineral resources,input 2D GIS layer with 3D geological model to the model,extract entity mineralization favorable information from input of huge amounts of data,use principal component analysis algorithm to multi-type mineralization favorable information for the parallel processing of unified format,and according to the type of data stored in the corresponding distributed database in turn,a non-negative matrix factorization algorithm is used to add non-negative constraints to the data in the database to realize the decomposition of mineral resources data.The fuzzy C-means classifier based on NMF algorithm is used to classify the data,obtain the best sample data of mineral resources,and complete the quantitative prediction of polymetallic mineral resources under loess overlying area.The experimental results show that the model can accurately predict the metal content and metal density of each measuring point in the polymetallic mining area under the loess overlying area,and can also divide the scale of the target area according to the total metal content,and predict the metal mineral reserves and metallogenic favorable degree under each measuring point.

关 键 词：大数据 黄土覆盖区 金属矿产资源 定量预测 成矿有利信息 非负矩阵分解 

分 类 号：P612[天文地球—矿床学] P628[天文地球—地质学]