基于Transformer的地震数据断层识别  

Faults identification of seismic data based on Transformer

作　　者：武庭润 高建虎[3] 常德宽 王海龙[3] 陶辉飞 李沐阳 WU Tingrun;GAO Jianhu;CHANG Dekuan;WANG Hailong;TAO Huifei;LI Muyang(Northwest Institute of Eco-Environment and Resources,Chinese Academy of Sciences,Lanzhou,Gansu 730000,China;College of Earth and Planetary Sciences,University of Chinese Academy of Sciences,Beijing 101400,China;PetroChina Research Institute of Petroleum Exploration&Development-Northwest,Lanzhou,Gansu 730000,China)

机构地区：[1]中国科学院西北生态环境资源研究院,甘肃兰州730000 [2]中国科学院大学地球与行星科学学院,北京101400 [3]中国石油勘探开发研究院西北分院,甘肃兰州730000

出　　处：《石油地球物理勘探》2024年第6期1217-1224,共8页Oil Geophysical Prospecting

基　　金：甘肃省科技重大专项“陇东地区天然气藏地球物理预测关键技术研发及应用”(23ZDGA004);中国石油集团科学技术研究与技术开发项目“多物理场融合及储层成像方法研究”(2023ZZ0505)联合资助。

摘　　要：利用地震资料识别断层在油气勘探中有着重要的作用。目前,机器学习和深度学习技术提高了断层识别的精度和效率,但断裂预测结果仍难以满足生产需求。为此,提出基于Transformer的地震断层识别方法,即3D SwinTrans-U-Net。该网络由Swin Transformer模块、卷积模块组成。其中,Swin Transformer模块可以利用Transformer的注意力机制提取全局信息,并将计算全局注意力转变为计算窗口的注意力,从而比Transformer减少了计算复杂度;卷积模块具有归纳偏置的特性,避免了Swin Transformer存在弱归纳偏置的缺陷;最后,利用U-Net结构,结合Swin Transformer层与卷积层,融合深层与浅层的信息并提取相关特征,充分学习全局性和局部依赖性信息,在保证断层识别精度的基础上提高了计算效率,实现端到端的地震断层学习。模型数据和实际数据测试均表明,3D SwinTrans-U-Net网络能进一步提升断层识别精度。Faults identification of seismic data plays an important role in oil and gas exploration.At present,the application of machine learning and deep learning techniques has enhanced faults identification precision and efficacy.However,the fracture prediction outcomes remain challenging to meet the production needs.Therefore,a Transformer-based seismic fault identification method,namely 3D SwinTrans-U-Net,is proposed.The network consists of Swin Transformer module and convolution module.Among the aforementioned modules,the Swin Transformer module employs the attention mechanism of the Transformer to extract global information,and transforms global attention computing into window attention computing,resulting in less computational complexity compared to the Transformer.The convolution module,with the property of inductive bias,avoids the Swin Transformer’s defect of weak inductive bias.Finally,the U-Net structure is utilized to combine the Swin Transformer and the convolutional layer.Thus,the structure can achieve deep and shallow information fusion,relevant feature extraction,as well as full learning of global and local dependency information.All these improve computational efficiency while ensuring fault identification accuracy,and enable end-to-end seismic fault deep learning.Synthetic and field seismic data tests have proven that the 3D SwinTrans-U-Net network can further improve the accuracy of fault identification.

关 键 词：深度学习 3D SwinTrans-U-Net 断层识别 TRANSFORMER Swin Transformer 卷积 

分 类 号：P631[天文地球—地质矿产勘探]