| 基于深度学习的山东大尹格庄金矿床深部三维预测模型 |
| |
| 引用本文： | 邓浩,郑扬,陈进,魏运凤,毛先成.基于深度学习的山东大尹格庄金矿床深部三维预测模型[J].地球学报,2020,41(2):157-165. |
| |
| 作者姓名： | 邓浩 郑扬 陈进 魏运凤 毛先成 |
| |
| 作者单位： | 中南大学地球科学与信息物理学院;中南大学有色金属成矿预测与地质环境监测教育部重点实验室,中南大学地球科学与信息物理学院,中南大学地球科学与信息物理学院;中南大学有色金属成矿预测与地质环境监测教育部重点实验室,中南大学地球科学与信息物理学院,中南大学地球科学与信息物理学院;中南大学有色金属成矿预测与地质环境监测教育部重点实验室 |
| |
| 基金项目： | 国家重点研发计划“深地资源勘查开采”重点专项课题(编号: 2017YFC0601503);国家自然科学基金项目(编号: 41972309; 41772349) |
| |
| 摘 要： | 在隐伏矿体三维预测中,预测模型的准确性在很大程度上取决于找矿指标对矿化富集部位的指示性。然而,找矿指标容易受到找矿概念模型可靠性和成矿信息提取有效性限制,从而影响预测的准确性。论文以山东大尹格庄金矿隐伏矿体三维预测为例,基于深度学习方法,构建矿床深部隐伏矿体三维预测模型,旨在利用深度网络模型,学习获得对矿化具有显著指示性的找矿指标,提升三维预测的准确性。该方法将三维地质模型及其形态特征转换为适合卷积网络二维图像,采用卷积神经网络实现找矿指标的自动提取,并构建三维地质模型到矿化富集地段的定量关联。利用该方法建立了大尹格庄金矿的三维预测模型,经与几种人工建立找矿指标预测模型的对比分析,表明基于深度学习的预测模型较大地提升了预测准确性。
|
| 关 键 词： | 三维预测模型 大尹格庄矿区 卷积神经网络 特征提取 |
Deep Learning-based 3D Prediction Model for the Dayingezhuang Gold Deposit, Shandong Province |
| |
| DENG Hao,ZHENG Yang,CHEN Jin,WEI Yun-feng and MAO Xian-cheng.Deep Learning-based 3D Prediction Model for the Dayingezhuang Gold Deposit, Shandong Province[J].Acta Geoscientia Sinica,2020,41(2):157-165. |
| |
| Authors: | DENG Hao ZHENG Yang CHEN Jin WEI Yun-feng and MAO Xian-cheng |
| |
| Institution: | School of Geosciences and Info-physics, Central South University;Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring (Ministry of Education), Central South University,School of Geosciences and Info-physics, Central South University,School of Geosciences and Info-physics, Central South University;Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring (Ministry of Education), Central South University,School of Geosciences and Info-physics, Central South University and School of Geosciences and Info-physics, Central South University;Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring (Ministry of Education), Central South University |
| |
| Abstract: | In the three-dimensional (3D) prediction of buried orebodies, the performance of the prediction model depends majorly on the indication of predictor variables to the mineralized enrichment site. The establishment of predictor variables depends on the reliability of metallogenic conceptual model and the effectiveness of ore-prospecting information extraction, which, however is easily hindered by geological experience and ore-forming information extraction method. With the 3D prediction of buried orebodies in the Danyingezhuang gold deposit in Shandong Province as an example, this paper propose a prediction modeling method based on deep learning, with the purpose of using deep networks to obtain significant predictor variables and improve the accuracy of 3D prediction driven by geological data. In this method, the 3D geological model and its shape features are transformed into two-dimensional images suitable for the convolutional networks. The convolutional neural network is adopted to realize the automatic extraction of ore-prospecting variables, and the quantitative correlation between the 3D geological model and the mineralized enrichment areas is thus constructed. By using this method, the 3D prediction model in the Danyingezhuang gold deposit was established. The comparison between the proposed mothod and several traditional modeling methods shows that the prediction model based on deep learning greatly improves the prediction accuracy. |
| |
| Keywords: | 3D predictive modeling Dayingezhuang gold deposit convolutional neural networks feature extraction |
| 本文献已被 CNKI 维普 等数据库收录！ |
| 点击此处可从《地球学报》浏览原始摘要信息 |
| 点击此处可从《地球学报》下载免费的PDF全文 |
|
