| 基于卷积自编码网络的夏河—合作地区金矿定量预测 |
| 点此下载全文 |
| 引用本文：柳炳利,谢淼,孔韫辉,唐瑞,余正波,罗德江.2023.基于卷积自编码网络的夏河—合作地区金矿定量预测[J].地球学报,44(5):877-886. |
| DOI：10.3975/cagsb.2023.022801 |
| 摘要点击次数: 1576 |
| 全文下载次数: 723 |
|
| 基金项目:国家重点研发计划课题(编号: 2017YFC0601505);国家自然科学基金项目(编号: 42072322);四川省科技厅项目(编号: 2022NSFSC0510) |
|
| 中文摘要:甘肃夏河—合作地区属西秦岭成矿带, 区域地质构造复杂、矿产资源丰富。该地区已发现一定数量的金多金属矿床(点), 且在矿集区及周边地区仍存在良好的金多金属矿找矿潜力。本文以夏河—合作地区为研究区, 基于成分数据分析定量提取了成矿元素组合, 集成了以构造-地球化学异常为基础的多元信息综合找矿模型, 基于卷积自编码网络(Convolutional Auto-Encode, CAE)模型进行区域金矿产资源定量预测。结果表明CAE模型在该区的预测具有良好的性能(AUC=0.90), 以此为依据确定的7个预远景区值得进一步开展勘查工作。 |
| 中文关键词:深度学习 卷积自编码 成分数据分析 定量矿产预测 |
| |
| Quantitative Gold Resources Prediction in Xiahe–Hezuo Area Based on Convolutional Auto-Encode Network |
|
|
| Abstract:The Xiahe–Hezuo area of Gansu province has complex geological structures and abundant gold mineral resources, is an important metallogenic belt within the West Qinling Mountains. At present, a certain number of gold polymetallic deposits have been found in this area, and there is still a good prospecting potential of gold mineralization. Therefore, this paper uses the Xiahe–Hezuo area as a research area. Geochemical associations are quantitatively extracted by the method of compositional data analysis. Based on the geological structure and geochemical anomalies, multiple metallogenic information is integrated for building the mineral exploration model. And then, the Convolutional Auto-Encode network (CAE) method is used for regional gold resource prediction. Finally, 7 exploration targets are delineated. The result shows an excellent prediction performance (AUC=0.90), and the targets deserve further research. |
| keywords:deep learning convolutional auto-encode compositional data analysis quantitative mineral resources prediction |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
|
|
