基于机器学习的磷灰石微量元素判别岩石类型  

Rock type discrimination by using trace elements of apatite based on the machine learning

作　　者：韩凤歌 冷成彪 陈加杰 占义勇 HAN Feng-ge;LENG Cheng-biao;CHEN Jia-jie;ZHAN Yi-yong(State Key Laboratory of Nuclear Resources and Environment,East China University of Technology,Nanchang 330013,China;School of Science,East China University of Technology,Nanchang 330013,China;School of Earth Sciences,East China University of Technology,Nanchang 330013,China)

机构地区：[1]东华理工大学核资源与环境国家重点实验室,南昌330013 [2]东华理工大学理学院,南昌330013 [3]东华理工大学地球科学学院,南昌330013

出　　处：《矿物岩石地球化学通报》2024年第3期607-620,共14页Bulletin of Mineralogy, Petrology and Geochemistry

基　　金：国家重点研发计划项目(2023YFC2906801);江西省国家级高层次人才创新创业项目(K20230004)。

摘　　要：近几年机器学习在地学领域发展迅猛。本文致力于利用机器学习方法研究岩浆成因的磷灰石微量元素对于岩浆岩不同岩石类型能否进行准确的判别。磷灰石是一种重要的矿物,在地质勘探和资源开发中有广泛的应用。然而,不同岩石类型中磷灰石微量元素的含量差异较大,本文收集了来自68篇文献的3720条岩浆成因的磷灰石微量元素数据信息,提出了基于随机森林(Random Forest,RF)和人工神经网络(Artificial Neural Network,ANN)2种机器学习的快速判别方法。结果表明,随机森林模型的分类精度达到了93.7%,其中Lu、La/Yb、MnO、Ce是对模型影响最大的4项指标;人工神经网络模型分类精度为89.7%,略低于随机森林模型,其中Lu、Gd/Yb、Ce、MnO是对模型影响最大的4项指标。本研究基于机器学习方法利用磷灰石微量元素实现了对岩浆岩岩石类型的判别,其成果不仅为磷灰石勘探提供了一种高效手段,也为机器学习在地质学领域的应用提供了实证案例。这种基于机器学习的判别方法有望为将来岩石类型鉴别和资源勘探领域的研究提供新的思路和方法。The machine learning techniques have been applied rapidly to deal with issues/problems in the field of Earth Sciences in recent years.This article aims to employ machine learning methods to investigate whether trace elements of magmatic apatites can be used to accurately discriminate different types of magmatic rocks.Apatite is a significant mineral with extensive applications in geological exploration and development of mineral resources.However,contents of trace elements of apatites in different types of rocks are relative significantly different.In this paper,we have collected data of trace elements of 3720 magmatic apatite samples from 68 literatures and have introduced two machine learning-based rapid rock type discrimination methods including the Random Forest(RF)and Artificial Neural Network(ANN).The results show that the rock type classification accuracy of the Random Forest model is 93.7%,with contents of Lu,MnO,and Ce,and La/Yb values being four indicators that had the greatest impact on the model;The rock type classification accuracy of the Artificial Neural Network model is 89.7%,which is slightly lower than that of the Random Forest model,with contents of Lu,Ce,and MnO,and Gd/Yb values being four indicators that had the greatest impact on the model.This study has discriminated types of magmatic rocks by using machine learning techniques for dealing with data of trace elements of apatites.Its achievements have not only offered an efficient approach for mineral exploration using data of apatites,but also demonstrated the practical application of machine learning techniques in geology.This machine learning-based rock type discrimination method is expected to provide new ideas and methods for future researches in fields of rock type identification and mineral resource exploration.

关 键 词：随机森林算法 人工神经网络算法 磷灰石 微量元素 岩石类型 

分 类 号：P578.92[天文地球—矿物学]