首页 | 本学科首页   官方微博 | 高级检索  
     

构建三维深度监督网络的断层检测方法
引用本文:王静,张军华,芦凤明,孟瑞刚,王作乾,常健强.构建三维深度监督网络的断层检测方法[J].石油地球物理勘探,2021,56(5):947-957.
作者姓名:王静  张军华  芦凤明  孟瑞刚  王作乾  常健强
作者单位:1. 中国石油大学(华东)地球科学与技术学院, 山东青岛 266580;2. 中国石油大港油田公司勘探开发研究院, 天津 300280
基金项目:本项研究受国家自然科学基金项目“含油气盆地低级序断裂的发育规律及其与高级序断裂的成因联系——以渤海湾盆地济阳坳陷为例”(42072169)和大港油田横向项目“自来屯开发区孔店组高精度油藏描述油藏地球物理关键技术研究”(DGYT-2020-JS-50413)联合资助。
摘    要:地震资料人工解释断层往往具不确定性。随着计算机和人工智能的发展,深度学习技术越来越多地应用于地球物理领域,多种基于卷积神经网络的算法也广泛地应用于断层识别。为此,结合三维U-Net和深度残差网络,引入多层深度监督的机制,构建了一种基于三维深度监督网络的断层检测方法。残差模块的引入能够简化网络的学习目标,降低训练难度,而多层的深度监督能够为网络提供更多的反馈,减轻训练过程中潜在的梯度消失,使解码器子网络能够学习到不同尺度的断层语义信息,可进一步提高断层识别的准确性。理论模型测试和实际地震资料的应用表明,该方法可以有效识别断层位置;与常规U-Net网络相比,减少了小断层的漏识别和错误识别;识别的大断层连续性好,断层细节更丰富,明显提高了断层识别的准确性。

关 键 词:地震资料解释  三维U-Net网络  残差模块  多层深度监督  断层识别  
收稿时间:2021-04-01

Research on fault detection method based on 3D deeply supervised network
Wang Jing,Zhang Junhua,Lu Fengming,Meng Ruigang,Wang Zuoqian,Chang Jianqiang.Research on fault detection method based on 3D deeply supervised network[J].Oil Geophysical Prospecting,2021,56(5):947-957.
Authors:Wang Jing  Zhang Junhua  Lu Fengming  Meng Ruigang  Wang Zuoqian  Chang Jianqiang
Affiliation:1. School of Geosciences, China University of Petroleum(East China), Qingdao, Shandong 266580, China;2. Research Institute of Exploration & Development, Dagang Oilfield Company, PetroChina, Tianjin 300280, China
Abstract:Manual interpretation of seismic faults is often uncertain. Amid the progress in computers and artificial intelligence, deep learning technologies are increasingly used in the field of geophysics, and a variety of algorithms based on the convolutional neural network are widely applied to fault recognition. In this paper, we propose a fault detection method based on the 3D deeply supervised network by combining 3D U-Net and the deep residual network and introducing the mechanism of multi-layer deep supervision. The residual block can simplify the learning goal of the network and reduce the difficulty of training, and the multi-level deep supervision provides more feedback to the network and alleviates the potential vanishing gradient problem during training, which enables the decoder sub-network to take advantage of multi-scale information and further improve the accuracy of fault detection. The theoretical model test and seismic raw data have proved that the 3D deeply supervised network can correctly identify the fault location; compared with the conventional U-Net, it reduces missed and misidentified small faults; the details of faults are more abundant and the accuracy of fault detection can be significantly improved.
Keywords:seismic data interpretation  3D U-Net  residual block  multi-level deep supervision  fault detection  
点击此处可从《石油地球物理勘探》浏览原始摘要信息
点击此处可从《石油地球物理勘探》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号