Automated Time-Window Selection Based on Machine Learning for Full-Waveform Inversion

Yangkang Chen,Judith Hill,Wenjie Lei,Matthieu Lefebvre,Jeroen Tromp,Ebru Bozdag,Dimitri Komatitsch

DOI: https://doi.org/10.1190/segam2017-17734162.1

2017-01-01

Abstract:PreviousNext No AccessSEG Technical Program Expanded Abstracts 2017Automated time-window selection based on machine learning for full-waveform inversionAuthors: Yangkang ChenJudith HillWenjie LeiMatthieu LefebvreJeroen TrompEbru BozdagDimitri KomatitschYangkang ChenOak Ridge National LaboratorySearch for more papers by this author, Judith HillOak Ridge National LaboratorySearch for more papers by this author, Wenjie LeiPrinceton UniversitySearch for more papers by this author, Matthieu LefebvrePrinceton UniversitySearch for more papers by this author, Jeroen TrompPrinceton UniversitySearch for more papers by this author, Ebru BozdagColorado School of MinesSearch for more papers by this author, and Dimitri KomatitschLMA, CNRS UPR 7051Search for more papers by this authorhttps://doi.org/10.1190/segam2017-17734162.1 SectionsAboutPDF/ePub ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InRedditEmail Abstract Due to increased computational capabilities afforded by modern and future computer architectures, the seismology community is demanding a more comprehensive understanding of full waveform information from recorded seismic data. Full waveform inversion seeks to match observed seismic data with synthesized seismograms by iteratively updating subsurface model parameters. Synthetic data are generated by solving the seismic wave equation using an effective and effcient numerical algorithm. In order to ensure inversion accuracy and stability, both synthesized and observed seismograms must be carefully pre-processed. More specifically, when synthetic and observed data have a large waveform mismatch during the initial iterations, waveforms should be carefully selected for calculating the misfit gradient in order to avoid instability. We introduce a fully automated algorithm based on machine learning (ML) to intelligently select time windows for calculating the misfit between observed and synthetic seismograms. The training dataset can be prepared using time windows obtained based on the FLEXWIN method, in which selection parameters are finely tuned. Results show that automatically selected time windows are of sufficiently high quality compared with the benchmark FLEXWIN method. Presentation Date: Tuesday, September 26, 2017 Start Time: 11:25 AM Location: Exhibit Hall C, E-P Station 3 Presentation Type: EPOSTER Keywords: full-waveform inversion, inversion, 3DPermalink: https://doi.org/10.1190/segam2017-17734162.1FiguresReferencesRelatedDetailsCited byAdjoint Waveform Tomography of South America27 January 2022 | Journal of Geophysical Research: Solid Earth, Vol. 127, No. 2Time Window Selection of Seismic Signals for Waveform Inversion Based on Deep LearningIEEE Transactions on Geoscience and Remote Sensing, Vol. 60Research on cross‐well pseudorandom electromagnetic detection method and extraction of response characteristics27 June 2020 | Energy Science & Engineering, Vol. 8, No. 10Extrapolated full-waveform inversion with deep learningHongyu Sun and Laurent Demanet16 April 2020 | GEOPHYSICS, Vol. 85, No. 3Seismic wavefield imaging of Earth’s interior across scales9 December 2019 | Nature Reviews Earth & Environment, Vol. 1, No. 1Automatic Waveform Classification and Arrival Picking Based on Convolutional Neural Network25 July 2019 | Earth and Space Science, Vol. 6, No. 7Low frequency extrapolation with deep learningHongyu Sun and Laurent Demanet27 August 2018Automatic microseismic-event detection via supervised machine learningShan Qu, Eric Verschuur, and Yangkang Chen27 August 2018 SEG Technical Program Expanded Abstracts 2017ISSN (print):1052-3812 ISSN (online):1949-4645Copyright: 2017 Pages: 6093 publication data© 2017 Published in electronic format with permission by the Society of Exploration GeophysicistsPublisher:Society of Exploration Geophysicists HistoryPublished Online: 17 Aug 2017 CITATION INFORMATION Yangkang Chen, Judith Hill, Wenjie Lei, Matthieu Lefebvre, Jeroen Tromp, Ebru Bozdag, and Dimitri Komatitsch, (2017), "Automated time-window selection based on machine learning for full-waveform inversion," SEG Technical Program Expanded Abstracts : 1604-1609. https://doi.org/10.1190/segam2017-17734162.1 Plain-Language Summary Keywordsfull-waveform inversioninversion3DPDF DownloadLoading ...