摘要
重力异常反演通过地表重力数据推测地下异常体的密度分布,是地球物理勘探中的重要工具,广泛应用于油田、矿床、地质结构和地下工事探测等领域。传统的重力反演方法面临计算复杂、分辨率低以及反演结果依赖于先验信息等问题。深度学习重力异常反演技术能够不依赖于初始模型或先验信息,在提高反演精度和减少计算时间等方面表现出明显优势。文中回顾了传统重力异常正、反演方法的发展及其局限性,总结了深度学习重力反演方法的研究现状,重点从数据准备、网络模型、网络优化和网络验证等四个方面介绍了不同重力反演问题的改进和创新,并比较了不同重力反演方法在美国路易斯安那州Vinton盐丘和墨西哥圣尼古拉斯矿床实测数据上的应用效果,其中多任务框架CDUNet在Vinton盐丘数据上取得了最准确的反演深度值,3D U-Net++网络在圣尼古拉斯矿床数据上获得了比U-Net网络更清晰且准确的反演结果。
The gravity anomaly inversion,which infers the density distribution of subsurface anomalies from sur-face gravity data,is an essential tool in geophysical exploration and is widely applied in fields such as oilfields,mineral deposits,geological structures,and underground works detection.Traditional gravity inversion methods face challenges of complex computation,low resolution,and dependence on prior information for inversion re-sults.However,deep learning-based gravity anomaly inversion techniques show significant advantages,par-ticularly in terms of improving inversion accuracy and reducing computation time,without the reliance on initial models or prior information.This paper reviews the development and limitations of traditional gravity anomaly forward and inversion methods and summarizes the current research on deep learning-based gravity inversion methods.Meanwhile,it introduces the improvements and innovations of different gravity inversion problems in four respects,including data preparation,network models,network optimization,and network validation.Ad-ditionally,the application effect of various gravity inversion methods on the measured data from Vinton Dome in Louisiana,the USA,and the San Nicolás ore deposit in Mexico.The multi-task framework CDUNet yields the most accurate inversion depth values on data of Vinton Dome,while the 3D U-Net++network obtains clearer and more accurate inversion results on the data of the San Nicolás ore deposit than the U-Net network.
作者
黄兴业
胡青青
邝文俊
万伏彬
范延松
徐馥芳
HUANG Xingye;HU Qingqing;KUANG Wenjun;WAN Fubin;FAN Yansong;XU Fufang(Advanced Interdisciplinary Technology Research Center,National Innovation Institute of Defense Technology,Beijing 100071,China)
出处
《石油地球物理勘探》
北大核心
2025年第4期1046-1058,共13页
Oil Geophysical Prospecting
基金
国家自然科学基金青年基金项目“原子干涉重力仪条纹对比度提高机理与技术研究”(11904408)资助。
关键词
重力异常反演
深度学习
数据驱动
网络模型
网络优化
gravity anomaly inversion
deep learning
data-driven
network model
network optimization
