Deep Learning Guiding First-Arrival Traveltime Tomography
Ziang Li,Xiaofeng Jia,Jie Zhang
DOI: https://doi.org/10.1190/segam2019-3215632.1
2019-01-01
Abstract:PreviousNext No AccessSEG Technical Program Expanded Abstracts 2019Deep learning guiding first-arrival traveltime tomographyAuthors: Ziang LiXiaofeng JiaJie ZhangZiang LiLaboratory of Seismology and Physics of Earth’s Interior, School of Earth and Space Sciences, University of Science and Technology of China, Hefei 230026, ChinaSearch for more papers by this author, Xiaofeng JiaLaboratory of Seismology and Physics of Earth’s Interior, School of Earth and Space Sciences, University of Science and Technology of China, Hefei 230026, ChinaSearch for more papers by this author, and Jie ZhangLaboratory of Seismology and Physics of Earth’s Interior, School of Earth and Space Sciences, University of Science and Technology of China, Hefei 230026, ChinaSearch for more papers by this authorhttps://doi.org/10.1190/segam2019-3215632.1 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail AbstractThe first-arrival traveltime tomography is applied to produce a robust near-surface velocity model. The solutions of first-arrival traveltime tomography are ill-posed and the Tikhonov regularization is commonly employed to stabilize the inverse problem. However, the spatial resolution of the model solutions may be decreased due to the smooth effect of the regularization. In this study, we develop a constrained first-arrival traveltime tomography based on the Generative Adversarial Network (GAN), a machine learning approach. The GAN is applied to generate a prior model in each iteration of the traveltime tomography. We make the assumption that both the GAN model and the solutions of traveltime tomography must be sensing the same cross-gradients relationship. A modified crossgradients function is incorporated as a constraint in a nonlinear least squares problem formulation, which is solved using a Gauss–Newton method. The performance of the new method is tested with both synthetics and a real dataset acquired from Daqing, China. The results show the new method has significant improvement in model resolution compared with the conventional first-arrival traveltime tomography.Presentation Date: Tuesday, September 17, 2019Session Start Time: 1:50 PMPresentation Time: 1:50 PMLocation: Poster Station 2Presentation Type: PosterKeywords: machine learning, traveltime, near surfacePermalink: https://doi.org/10.1190/segam2019-3215632.1FiguresReferencesRelatedDetailsCited bySegNet-based first-break picking via seismic waveform classification directly from shot gathers with sparsely distributed tracesPetroleum Science, Vol. 19, No. 1Machine learning-enabled traveltime inversion based on the horizontal source-location perturbationIsa Eren Yildirim, Tariq Alkhalifah, and Ertugrul Umut Yildirim23 November 2021 | GEOPHYSICS, Vol. 87, No. 1 SEG Technical Program Expanded Abstracts 2019ISSN (print):1052-3812 ISSN (online):1949-4645Copyright: 2019 Pages: 5407 publication data© 2019 Published in electronic format with permission by the Society of Exploration GeophysicistsPublisher:Society of Exploration Geophysicists HistoryPublished Online: 10 Aug 2019 CITATION INFORMATION Ziang Li, Xiaofeng Jia, and Jie Zhang, (2019), "Deep learning guiding first-arrival traveltime tomography," SEG Technical Program Expanded Abstracts : 2513-2517. https://doi.org/10.1190/segam2019-3215632.1 Plain-Language Summary Keywordsmachine learningtraveltimenear surfacePDF DownloadLoading ...