| 大数据挖掘与智能预测找矿靶区实验研究——卷积神经网络模型的应用 |
| |
| 引用本文： | 刘艳鹏,朱立新,周永章.大数据挖掘与智能预测找矿靶区实验研究——卷积神经网络模型的应用[J].大地构造与成矿学,2020(2):192-202. |
| |
| 作者姓名： | 刘艳鹏 朱立新 周永章 |
| |
| 作者单位： | 东华理工大学核资源与环境国家重点实验室;广东省地质过程与矿产资源探查重点实验室;中山大学地球环境与地球资源研究中心;中国地质调查局 |
| |
| 基金项目： | 中国地质调查局地质调查项目(DD20190305);国家重点研发计划重点专项(2016YFC0600506)联合资助. |
| |
| 摘 要： | 成矿预测需要通过一定的规则集合将专家观点、地质背景、成矿类型等因素进行综合考虑。但由于受到人类实际计算能力的生物条件限制,影响找矿预测成果的最大因素是找矿者的经验知识。随着大数据时代的到来,成矿预测可充分利用数学计算,即以特定规则对成矿系统进行计算,以概率表示成矿前景。依靠计算机的超级运算能力,结合机器学习的方法技术,可以对地质大数据进行成矿预测特征学习,实现对众多地质变量与矿体相关性之间的验证,从而进行预测。本文以安徽东至兆吉口铅锌矿床为例,示范如何通过机器学习的卷积神经网络方法,学习元素Zn在地表的分布特征与矿体在地下空间就位的耦合关系,并圈定靶区。经过450次训练后,得到了准确率95%,损失率14%的CNN模型,并成功实现智能圈定3块找矿靶区。这种神经网络模型可能表达了矿体在地下就位时元素在地表分布的响应,可以用来进行找矿勘查并圈定靶区。
|
| 关 键 词： | 大数据 成矿预测 卷积神经网络 机器学习 地球化学 兆吉口 |
Experimental Research on Big Data Mining and Intelligent Prediction of Prospecting Target Area--Application of Convolutional Neural Network Model |
| |
| LIU Yanpeng,ZHU Lixin,ZHOU Yongzhang.Experimental Research on Big Data Mining and Intelligent Prediction of Prospecting Target Area--Application of Convolutional Neural Network Model[J].Geotectonica et Metallogenia,2020(2):192-202. |
| |
| Authors: | LIU Yanpeng ZHU Lixin ZHOU Yongzhang |
| |
| Institution: | (State Key Laboratory of Nuclear Resources and Environment,East China University of Technology,Nanchang 330013,Jiangxi,China;Guangdong Provincial Key Laboratory of Geological Processes and Mineral Resource Survey,Guangzhou 510275,Guangdong,China;Centre for Earth Environment and Resources,Sun Yat-sen University,Guangzhou 510275,Guangdong,China;China Geological Survey,Beijing 100037,China) |
| |
| Abstract: | Metallogenic prediction needs to consider factors such as expert opinion,geological background,and metallogenic types under a comprehensive set of rules.However,due to the limitation of the biological conditions of human’s actual computing ability,the biggest factor affecting the prospecting results of metallogenic prediction is the experience and knowledge of prospectors.With the advent of the era of big data,the metallogenic prediction can be regarded as mathematical calculation,that is,the metallogenic system is calculated according to specific rules,and the result is the metallogenic prospect expressed by probability.Relying on the computer’s super-computing ability and machine learning methods and techniques,the computer can learn the characteristics of metallogenic prediction from big geological data,and realize the one-to-one verification of the correlation between different geological variables and ore bodies,so as to make predictions.In this paper,a case study of the Zhaojikou Pb-Zn ore deposit,Anhui province,is carried out to demonstrate how to use the convolutional neural network to learn the coupling relationship between the surficial distribution characteristics of Zn and the position of the ore body in the depth,and finally delineate the target area.After 450 trainings,a CNN model with 95%accuracy and 14%loss rate was obtained,and achieved intelligent delineation of three target areas.This neural network model may express the response of element distribution on the surface when the ore body was in deep underground,and can be used for ore mineral prospecting and delineation of target areas. |
| |
| Keywords: | big data metallogenic prediction convolutional neural network machine learning geochemistry Zhaojikou |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
