摘要
熊耳山矿集区是中国重要的金成矿带,但成矿地质条件复杂,传统地球化学异常提取方法难以有效捕捉潜在的复杂地质信息,且缺乏对数据全局特征的充分利用。本文引入具有多头注意力机制的Transformer模型,以捕捉地球化学数据中的长程关联,提升异常提取的精度和稳定性。基于1∶50000水系沉积物地球化学数据,利用Transformer模型重构地球化学背景,并通过计算原始与重构元素浓度的欧氏距离实现金矿相关的异常信息提取。模型评价结果AUC值达0.876,提取的异常与已知金矿点分布、控矿构造展布一致,表明识别的异常可靠性高。基于识别的化探异常,本文圈定了熊耳山地区金成矿潜力区6处,可供后续金矿勘查部署参考。
The Xiong'ershan ore concentration area represents a significant gold metallogenic belt within China.However,the metallogenic geological conditions are complex.Traditional geochemical anomaly extraction methodologies struggle to effectively capture the underlying complex geological information and lack the comprehensive utilization of the global characteristics inherent in the data.In this study,the transformer model,equipped with a multihead attention mechanism,is introduced.This mechanism enables the capture of long-range correlations within the geochemical data,thereby enhancing the precision and stability of anomaly extraction.Leveraging the 1∶50000 stream sediment geochemical data,the transformer model is employed to reconstruct the geochemical background.Subsequently,the geochemical anomaly information related to gold deposits is extracted by computing the Euclidean distance between the original and reconstructed element concentrations.The AUC value of the model reaches 0.876,and the extracted geochemical anomalies are consistent with the distribution of known gold deposits and the distribution of ore-controlling structures,indicating that the identified anomalies have high reliability.According to the results of geochemical anomaly identification,this paper delineates six gold metallogenic potential areas in the Xiong'ershan area,which can serve as a valuable reference for the subsequent deployment of gold exploration efforts.
作者
刘文毅
梁楠鑫
杜虹
张娅
喻姝研
辛梓豪
李永峰
张宇
刘中杰
毛先成
LIU Wenyi;LIANG Nanxin;DU Hong;ZHANG Ya;YU Shuyan;XIN Zihao;LI Yongfeng;ZHANG Yu;LIU Zhongjie;MAO Xiancheng(Henan Academy of Geology,Zhengzhou 450016,Henan,China;The Geological sub center of Henan Data and Application Center of Earth Observation System with High Resolution,Zhengzhou 450016,Henan,China;Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring Ministry of Education,School of Geosciences and Info-Physics,Central South University,Changsha 410083,Hunan,China)
出处
《矿产勘查》
2025年第8期2057-2066,共10页
Mineral Exploration
基金
河南省地质研究院2024年度地质科技攻关项目(2024-903-XM05)资助。
关键词
地球化学勘查
异常识别
机器学习
金矿
熊耳山
geochemistry exploration
anomaly identification
machine learning
gold deposit
Xiong'ershan district
