基于层次分解、主成分分析和高斯混合模型的火成岩岩性识别——以惠州26洼古潜山为例  

Identification of igneous rock lithology based on hierarchical decomposition,principal component analysis and Gaussian mixture model:Taking a case studyof buried hills in Huizhou 26 sub-sag

作　　者：高楚桥[1] 詹旺 赵彬[1] 程鑫财 GAO Chuqiao;ZHAN Wang;ZHAO Bin;CHENG Xincai(College of Geophysics and Petroleum Resources,Yangtze University,Wuhan 430100,Hubei)

机构地区：[1]长江大学地球物理与石油资源学院,湖北武汉430100

出　　处：《长江大学学报（自然科学版）》2024年第2期36-44,共9页Journal of Yangtze University(Natural Science Edition)

基　　金：国家科技重大专项“东海深层低渗致密天然气勘探开发技术”(2016ZX05027-002-002)。

摘　　要：火成岩油气成藏规律复杂,受到火山运动、强构造运动以及风化剥蚀等叠加影响,火成岩的化学成分和结构构造复杂多样,非均质性极强,采用常规岩性识别方法难以一次性将所有岩性准确识别。借鉴层次分解思路,以惠州26洼古潜山为例,提出了一种火成岩岩性测井识别分类方法:综合考虑火成岩地质分类原则和测井响应特征来确定岩性识别层级,基于这种层次性的分类原则,在每一层次定量优选岩性识别敏感参数,建立研究区岩性识别优选层级;在明确岩性识别优选层级的基础上,逐级逐次使用主成分分析(PCA)和高斯混合模型(GMM)对岩性进行判别并确定其计算函数,建立分级优选岩性识别模型,最终达到整体岩性区分的目的。研究结果表明,研究区辉绿岩和闪长岩识别正确率分别为87.31%和84.32%,未分级未优选辉绿岩和闪长岩识别正确率为60.45%和54.88%,分级未优选其岩性识别正确率为69.61%和67.04%,有效提高了研究区的复杂岩性识别精度。该方法的提出对提高火成岩岩性识别精度提供了一种思路,也为研究区古潜山火成岩岩性精确识别提供了参考依据。Igneous rocks have complex hydrocarbon accumulation laws.Due to the complexity and diversity of chemical composition and structure,and strong heterogeneity,it is difficult to accurately identify lithology by using conventional lithologic identification methods.Referring to the hierarchical decomposition method,taking a case study of buried hills in Huizhou 26 sub-sag,a classification principle of igneous rock lithology logging is proposed according to the hierarchical decomposition method.The classification level of lithology is determined by using the geological classification principle of igneous rock and logging response characteristics.Considering the geological classification principle of igneous rock and logging response characteristics,which can get the level of lithologic identification.Based on this principle,the sensitive parameters of lithologic identification are quantitatively optimized at each level to establish the optimal level of lithologic identification in the study area.On the basis of defining the preferred level of lithological identification,principal component analysis(PCA)and Gaussian mixture model(GMM)are used step by step to distinguish lithology and determine its calculation function.Through the establishment of a classification and optimization lithology identification model,the overall lithology can be finally distinguished.Using this method,the correct identification rates of diabase and diorite are 87.31%and 84.32%,while the correct identification rates of diabase and diorite without classification and optimization are 60.45%and 54.88%,and the correct identification rates of lithology without classification and optimization are 69.61%and 67.04%.The results show that the classification optimization method can effectively improve the identification accuracy of complex lithology in the study area.This method provides a new idea for improving the accuracy of igneous rock lithology identification,and also provides a reference for accurate identification of ancient buried hill igneous ro

关 键 词：火成岩 岩性识别 层次分解法 主成分分析 高斯混合模型 

分 类 号：P631.84[天文地球—地质矿产勘探] TE132.14[天文地球—地质学]