| 基于成矿条件数值模拟和支持向量机算法的深部成矿预测——以粤北凡口铅锌矿为例 |
| |
| 引用本文： | 王语,周永章,肖凡,王俊,王恺其,余晓彤.基于成矿条件数值模拟和支持向量机算法的深部成矿预测——以粤北凡口铅锌矿为例[J].大地构造与成矿学,2020(2):222-230. |
| |
| 作者姓名： | 王语 周永章 肖凡 王俊 王恺其 余晓彤 |
| |
| 作者单位： | 中山大学地球环境与地球资源研究中心;广东省地质过程与矿产资源探查重点实验室;中山大学地球科学与工程学院;南方海洋科学与工程广东省实验室（珠海） |
| |
| 基金项目： | 国家自然科学基金项目(41872245);国家重点研发计划重点专项(2016YFC0600506);中央高校基本科研业务费中山大学青年教师培养项目(17lgpy48)联合资助。 |
| |
| 摘 要： | 随着计算机科学和地质大数据技术的迅猛发展,数值模拟和机器学习已成为当今地学领域定量发展的重要前沿方向。数值模拟综合运用了研究区地质、构造、地球物理、地球化学等多源信息,将成矿条件与过程进行量化模拟分析,对研究成矿动力学演化过程及成矿响应有重要意义,可对已有成矿要素/信息在时空上进行扩展/外推,扩大了成矿预测信息的广度和深度,为解决深部成矿预测中获取深部信息难题提供了一种可能的有效途径。支持向量机是一种重要的机器学习分类算法,它具有简洁、方便、高效和计算结果较稳定等特点,在众多领域中得以成功应用,是成矿预测中多源信息提取与融合的一种可靠的技术手段。为了充分利用数值模拟与机器学习的优势,本文提出将计算机数值模拟方法和机器学习(即支持向量机算法)相结合来进行深部成矿预测的新方法。以粤北凡口超大型铅锌矿为例,首先,对凡口矿区勘探线剖面进行构造应力场模拟;进而,以已知钻孔数据作为训练集和测试集,运用支持向量机算法对模拟结果中的不同参量(也即模拟所得的成矿条件)进行训练学习;最后,建立相应的定量找矿预测模型对研究区(或剖面)外围和深部找矿进行预测评价。研究结果表明,本文所建立的预测模型精确度和召回率都较好,预测结果显示出了三个成矿可能较大的区域,说明数值模拟技术和机器学习算法结合应用的效果较好。这种新的成矿预测方法为深部找矿预测提供了一种可行的新思路和新途径,可以有效地拓展运用到其他矿区、其他类型矿床的深部找矿预测工作中。
|
| 关 键 词： | 数值模拟 机器学习 支持向量机 深部找矿 凡口铅锌矿 |
Numerical Metallogenic Modelling and Support Vector Machine Methods Applied to Predict Deep Mineralization:A Case Study from the Fankou Pb-Zn Ore Deposit in Northern Guangdong |
| |
| WANG Yu,ZHOU Yongzhang,XIAO Fan,WANG Jun,WANG Kaiqi,YU Xiaotong.Numerical Metallogenic Modelling and Support Vector Machine Methods Applied to Predict Deep Mineralization:A Case Study from the Fankou Pb-Zn Ore Deposit in Northern Guangdong[J].Geotectonica et Metallogenia,2020(2):222-230. |
| |
| Authors: | WANG Yu ZHOU Yongzhang XIAO Fan WANG Jun WANG Kaiqi YU Xiaotong |
| |
| Institution: | (Center for Earth Environment and Resource,Sun Yat-sen University,Guangzhou 510275,Guangdong,China;Guangdong Key Laboratory of Geological Process and Mineral Resources Exploration,Guangzhou 510275,Guangdong,China;School of Earth Sciences and Engineering,Sun Yat-sen University,Guangzhou 510275,Guangdong,China;Southern Marine Science and Engineering Guangdong Laboratory(Zhuhai),Zhuhai 519000,Guangdong,China) |
| |
| Abstract: | With the rapid development of technologies in both computer science and geological big data,numerical modelling and machine learning have become two important frontiers in quantitative geoscience.The numerical modelling method simulates and analyzes the metallogenic conditions and processes of a study area by utilizing multi-source geoscience datasets such as geological,tectonic,geophysical and geochemical data,which is useful for investigating the dynamic evolution of a mineralization system and its metallogenic response.It means that numerical modelling can be used to expand or/and inverse existing ore-forming features or/and conditions in time and space,which has extended metallogenic prediction information in both the range and depth.Thus,numerical modelling technique could be a plausible tool for obtaining deep geo-information in mineralization depth prediction,which has been considered as one of the most critical issues in deep ore exploration.In addition,the machine learning algorithms such as support vector machine(SVM)have been illustrated to be available means for multi-source information extraction and fusion in metallogenic prediction.The SVM method has been successfully used in various fields including mineralization prediction because of its advantages in simplicity,convenience and stability.The main purpose of this study is to develop and demonstrate a novel method that combines numerical modelling with machine learning(i.e.SVM algorithm)for metallogenic depth prediction.In this case study,we first simulated the tectonic stress field of No.–217 prospecting profile in the Fankou Pb-Zn ore deposit by using finite element numerical modelling technique.The numerical simulation results are the finite element node values of stresses and strains of Pb-Zn mineralization(i.e.the simulated metallogenic conditions).Then,we divided the nodes near to the drill holes randomly into the training and test datasets for studying and examining the machine learning algorithm of SVM.Finally,we established a quantitative prediction model to predict and evaluate mineralization depth of the study area(No.–217 prospecting profile).The prediction results showed that the SVM method performance well in both precision and recall rate.Three target areas with high potential for Pb-Zn mineralization,which are close to the known ore bodies,were identified and delineated.The inspiring outcomes of the numerical modelling and machine learning algorithm in this study showed that the novel method is applicable for deep prospecting prediction,and can be effectively extended to other ore fields and other types of ore deposits for prediction of mineralization at depth. |
| |
| Keywords: | numerical modelling machine learning support vector machine deep ore prospecting Fankou Pb-Zn ore deposit |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
