Abstract
Elastic full waveform inversion (EFWI) is a powerful tool for estimating elastic models by reducing the misfit between multi-component seismic records and simulated data. However, when multiple parameters are updated simultaneously, the gradients of the loss function with respect to these parameters will be coupled together, the effect exacerbate the nonlinear problem. We propose a parametric EFWI method based on convolutional neural networks (CNN-EFWI). The parameters that need to be updated are the weights in the neural network rather than the elastic models. The convolutional kernel in the network can increase spatial correlations of elastic models, which can be regard as a regularization strategy to mitigate local minima issue. Furthermore, the representation also can mitigate the cross-talk between parameters due to the reconstruction of Frechét derivatives by neural networks. Both forward and backward processes are implemented using a time-domain finite-difference solver for elastic wave equation. Numerical examples on overthrust models, fluid saturated models and 2004 BP salt body models demonstrate that CNN-EFWI can partially mitigate the local minima problem and reduce the dependence of inversion on the initial models. Mini-batch configuration is used to speed up the update and achieve fast convergence. In addition, the inversion of noisy data further verifies the robustness of CNN-EFWI.
Similar content being viewed by others
Data availability
Data associated with this research are available and can be accessed via the following URL: https://github.com/guoketing/CNN-FWI.
References
Baumstein A (2014) Extended subspace method for attenuation of crosstalk in multi-parameter full wavefield inversion In: Proceedings 2014 SEG annual meeting 2014, All days: SEG-2014–0546.
Billette FJ, and Brandsberg-Dahl S (2005) The 2004 BP velocity benchmark In: Extended abstracts https://doi.org/10.3997/2214-4609-pdb.1.B035.
Bishop CM (2006) Pattern recognition and machine learning (information science and statistics). Springer, New York
Borisov D, Singh SC (2015) Three-dimensional elastic full waveform inversion in a marine environment using multicomponent ocean-bottom cables: a synthetic study. Geophys J Int 201(3):1215–1234. https://doi.org/10.1093/gji/ggv048
Brenders AJ, Albertin U and Mika J (2012) Comparison of 3-D time-and frequency-domain waveform inversion: Benefits and insights of a broadband, discrete-frequency strategy In: Proceedings 2012 SEG annual meeting, OnePetro.
Brossier R, Operto S, Virieux J (2009) Seismic imaging of complex onshore structures by 2D elastic frequency-domain full-waveform inversion. Geophysics 74(6):105–118. https://doi.org/10.1190/1.3215771
Chen F, Zong Z, Jiang M (2021) Seismic reflectivity and transmissivity parametrization with the effect of normal in situ stress. Geophys J Int 226(3):1599–1614. https://doi.org/10.1093/gji/ggab179
Chen F, Zong Z, Yang Y, Gu X (2022) Amplitude-variation-with-offset inversion using P- to S-wave velocity ratio and P-wave velocity. Geophysics 87(4):N63–N74. https://doi.org/10.1190/geo2021-0623.1
Choi Y, Shin C, Min D-J (2007) Frequency-domain elastic full-waveform inversion using the new pseudo-Hessian matrix: elastic Marmousi-2 synthetic test, SEG technical program expanded abstracts 2007. Soc Explor Geophys 65:1908–1912
Fabien-Ouellet G, Sarkar R (2020) Seismic velocity estimation: a deep recurrent neural-network approach. Geophysics 85(1):U21–U29. https://doi.org/10.1190/geo2018-0786.1
Forgues E, Lambaré G (1997) Parameterization study for acoustic and elastic ray plus born inversion. J Seism Explor 6(2–3):253–277
Freudenreich Y, Singh S (2000). Full waveform inversion for seismic data—Frequency versus time domain. https://doi.org/10.3997/2214-4609-pdb.28.C54
Innanen KA (2014) Seismic AVO and the inverse Hessian in precritical reflection full waveform inversion. Geophys J Int 199(2):717–734. https://doi.org/10.1093/gji/ggu291
Kennett BLN, Sambridge MS, Williamson PR (1988) Subspace methods for large inverse problems with multiple parameter classes. Geophys J Int 94(2):237–247. https://doi.org/10.1111/j.1365-246X.1988.tb05898.x
Kingma DP and Ba JJ (2014) Adam: a method for stochastic optimization. arXiv preprint arXiv:1412.6980
Komatitsch D, Martin R (2007) An unsplit convolutional perfectly matched layer improved at grazing incidence for the seismic wave equation. Geophysics 72(5):SM155–SM167. https://doi.org/10.1190/1.2757586
Lempitsky, V., A. Vedaldi, and D. Ulyanov. 2018, Deep image prior. paper read at 2018 IEEE/CVF conference on computer vision and pattern recognition. pp. 18–23.
Mavko G, Mukerji T, Dvorkin J (2020) The rock physics handbook. Cambridge University Press
Mora P (1987) Nonlinear two-dimensional elastic inversion of multioffset seismic data 52(9):1211–1228. https://doi.org/10.1190/1.1442384
Nihei KT, Li X (2007) Frequency response modelling of seismic waves using finite difference time domain with phase sensitive detection (TD—PSD). Geophys J Int 169(3):1069–1078. https://doi.org/10.1111/j.1365-246X.2006.03262
Operto S, Gholami Y, Prieux V, Ribodetti A, Brossier R, Metivier L, Virieux J (2013) A guided tour of multiparameter full-waveform inversion with multicomponent data: from theory to practice. Lead Edge 32(9):1040–1054. https://doi.org/10.1190/tle32091040.1
Pratt RG, Shin C, Hick GJ (1998) Gauss-Newton and full Newton methods in frequency–space seismic waveform inversion. Geophys J Int 133(2):341–362. https://doi.org/10.1046/j.1365-246X.1998.00498.x
Prieux V, Brossier R, Operto S, Virieux J (2013) Multiparameter full waveform inversion of multicomponent ocean-bottom-cable data from the Valhall field Part 2: imaging compressive-wave and shear-wave velocities. Geophys J Int 194(3):1665–1681. https://doi.org/10.1093/gji/ggt178
Ren Z, Liu Y (2016) A hierarchical elastic full-waveform inversion scheme based on wavefield separation and the multistep-length approach. Geophysics 81(3):R99–R123. https://doi.org/10.1190/geo2015-0431.1
Richardson A (2018) Seismic full-waveform inversion using deep learning tools and techniques. arXiv preprint arXiv:.07232.
Sears TJ, Singh SC, Barton PJ (2008) Elastic full waveform inversion of multi-component OBC seismic data. Geophys. Prospect 56(6):843–862. https://doi.org/10.1111/j.1365-2478.2008.00692.x
Shin C, Jang S, Min DJ (2001) Improved amplitude preservation for prestack depth migration by inverse scattering theory. Geophy Prospect 49(5):592–606. https://doi.org/10.1046/j.1365-2478.2001.00279.x
Shipp RM, Singh SC (2002) Two-dimensional full wavefield inversion of wide-aperture marine seismic streamer data. Geophys J Int 151(2):325–344. https://doi.org/10.1046/j.1365-246X.2002.01645.x
Sirgue L, Etgen JT, Albertin U (2008) 3D frequency domain waveform inversion using time domain finite difference methods. Google Pat. https://doi.org/10.3997/2214-4609.20147683
Sun J, Niu Z, Innanen KA, Li J, Trad DO (2020) A theory-guided deep-learning formulation and optimization of seismic waveform inversionTheory-guided DL and seismic inversion. Geophysics 85(2):R87–R99
Sun J, Innanen KA, Huang C (2021) Physics-guided deep learning for seismic inversion with hybrid training and uncertainty analysis. Geophysics 86(3):R303–R317. https://doi.org/10.1190/geo2020-0312.1
Tarantola A (1986) A strategy for nonlinear elastic inversion of seismic reflection data. Geophysics 51(10):1893–1903. https://doi.org/10.1190/1.1442046
Wang TF, Cheng JB (2017) Elastic full waveform inversion based on mode decomposition: the approach and mechanism. Geophys J Int 209(2):606–622. https://doi.org/10.1093/gji/ggx038
Wang W, McMechan GA, Ma J (2021) Elastic isotropic and anisotropic full-waveform inversions using automatic differentiation for gradient calculations in a framework of recurrent neural networks. Geophysics 86(6):795–810. https://doi.org/10.1190/geo2020-0542.1
Wu Y, Lin Y (2019) InversionNet: an efficient and accurate data-driven full waveform inversion. IEEE Trans Comput Imaging 6:419–433
Wu Y, McMechan GAJG (2019) Parametric convolutional neural network-domain full-waveform inversion. Geophysics 84(6):881–896
Xu K, McMechan GAJG (2014) 2D frequency-domain elastic full-waveform inversion using time-domain modeling and a multistep-length gradient approach. Geophysics 79(2):41–53
Yang F, Ma J (2019) Deep-learning inversion: a next-generation seismic velocity model building method. Geophysics 84(4):R583–R599. https://doi.org/10.1190/geo2018-0249.1
Zhang T, Sun J, Innanen KA and Trad D (2021) Numerical analysis of a deep learning formulation of elastic full waveform inversion with high order total variation regularization in different parameterization. arXiv preprint arXiv:2101.08924
Zhu W, Xu K, Darve E, Beroza G (2021) A general approach to seismic inversion with automatic differentiation: Computers. Comput Eosci 151:104751
Zhu W, Xu K, Darve E, Biondi B, Beroza GCJG (2022) Integrating deep neural networks with full-waveform inversion: Reparameterization, regularization, and uncertainty quantification. Geophysics 87(1):93–109
Acknowledgements
We would like to acknowledge the sponsorship of National Natural Science Foundation of China (42174139,41974119, 42030103) and Science Foundation from Innovation and Technology Support Program for Young Scientists in Colleges of Shandong province and Ministry of Science and Technology of China (2019RA2136) and support by the Marine S&T Fund of Shandong Province for Pilot National Laboratory for Marine Science and Technology(Qingdao) (Grant No.2021QNLM020001-6).(Forgues and Lambaré, 1997).
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no competing interests. No infringement of others' intellectual property rights.
Additional information
Edited by Prof. Junlun Li (ASSOCIATE EDITOR) / Prof. Gabriela Fernández Viejo (CO-EDITOR-IN-CHIEF).
Appendix I
Appendix I
In this appendix, we derive the gradient of cost function with respect to Lamé parameters (\(\lambda\) and \(\mu\)) and density (\(\rho\)). For simplicity, we ignore the non-reflecting PML regions. The sum of squared errors is used as the cost function:
where \(N_{t}\) is the number of time points. In the adjoint state method algorithms, the gradient for elastic model parameters \(\partial J/\partial \lambda\), \(\partial J/\partial \mu\) and \(\partial J/\partial \rho\) depend on the adjoint wavefields, which is calculated by chain rule. We obtain the adjoint equations:
where the wavefields with a tilde denote the adjoint wavefield. In other words, they are the partial derivatives of \(J\) with respect to corresponding wavefields. For example, \(\tilde{v}_{z}\) indicates \(\partial J/\partial v_{z}\). \({\Delta }t\) is discrete time interval. \(S_{x} (t)\) and \(S_{z} (t)\) are adjoint source.
As the adjoint wave fields propagate backward in time, the derivatives of the objective function with respect to the elastic parameters are calculated as:
Note that the system is solved backwards in time. Connect Eqs. (3) and (2), the gradients of the objective function \(J\) with respect to elastic models \(\lambda\), \(\mu\) and \(\rho\) are:
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Guo, K., Zong, Z., Yang, J. et al. Parametric elastic full waveform inversion with convolutional neural network. Acta Geophys. 72, 673–687 (2024). https://doi.org/10.1007/s11600-023-01123-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11600-023-01123-3