Article
Geochemistry & Geophysics
Guochen Wu, Zhanyuan Liang, Xiaoyu Zhang, Lingyun Yang
Summary: Elastic full-waveform inversion (EFWI) is a method for recovering high-resolution model parameters, but the coupling of P- and S-waves introduces nonlinearity and parameter crosstalk in EFWI. We propose an EFWI approach with unconverted-wave adjoint propagators to address these issues and achieve satisfactory results.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Engineering, Biomedical
Abdelrahman M. Elmeliegy, Murthy N. Guddati
Summary: This paper presents a methodology for inverting 2D elasticity maps from measurements on a single line, with the ultimate goal of reconstructing 3D elasticity maps from ultrasound particle velocity measurements. The inversion approach is based on gradient optimization and uses full-wave simulation to accurately capture shear wave propagation and scattering in soft tissue. The proposed correlation-based cost functional shows better convexity and convergence properties compared to traditional least-squares functional, making it more robust against noisy measurements and other errors. The results demonstrate the effectiveness of the method in characterizing homogeneous inclusions and the entire region of interest. Significance: This new framework for shear wave elastography has the potential to accurately map shear modulus using ultrasound elastography data from standard clinical scanners.
PHYSICS IN MEDICINE AND BIOLOGY
(2023)
Article
Geochemistry & Geophysics
Leon Diekmann, Ivan Vasconcelos, Tristan van Leeuwen
Summary: Full waveform inversion and least-squares reverse time migration are commonly used for seismic wave imaging, relying on the Born approximation to compute gradients and update models. We propose using the Marchenko integral to obtain an alternative linear equation that includes all orders of scattering. This new linearization strategy, although relying on the quality of the Marchenko-derived Green's functions, produces slightly better inverted models than the single-scattering approximation.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2023)
Article
Geochemistry & Geophysics
Dirk Philip van Herwaarden, Michael Afanasiev, Solvi Thrastarson, Andreas Fichtner
Summary: We propose a new approach to full-waveform inversion that allows for continuous assimilation of growing data volumes without the need to reinvert all data. Specifically designed for seismological applications, our method utilizes a dynamic mini-batch stochastic L-BFGS to sequentially add new data while maintaining convergence and consistency in model fit measurement.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2021)
Article
Geochemistry & Geophysics
Scott D. Keating, Kristopher A. Innanen
Summary: Prior knowledge in seismic inversion can improve the accuracy of models compared to using seismic data alone. This study proposes an optimization strategy for full waveform inversion (FWI) that incorporates global regularization information and allows models to 'tunnel' between basins to honor prior information.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2022)
Article
Engineering, Multidisciplinary
Tim Buerchner, Philipp Kopp, Stefan Kollmannsberger, Ernst Rank
Summary: Full Waveform Inversion (FWI) is a successful method for reconstructing material models from measured wave signals. It is particularly successful in reconstructing smoothly varying material deviations, but not fully suitable for detecting sharp defects with low material contrast in non-destructive testing (NDT).
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Computer Science, Interdisciplinary Applications
A. Chang, L. Gross, S. Horning
Summary: This paper presents a geostatistical inversion approach called Random Mixing (RM) for deterministic full waveform inversion (FWI). RM uses linear combinations of spatial random fields to constrain the velocity field during inversion, allowing quantification of estimation uncertainty. The algorithm is implemented using the finite element method and Message Passing Interface for parallelization.
COMPUTERS & GEOSCIENCES
(2022)
Article
Geochemistry & Geophysics
Xin Zhang, Angus Lomas, Muhong Zhou, York Zheng, Andrew Curtis
Summary: Seismic full-waveform inversion (FWI) provides high resolution images of the subsurface by solving a highly non-linear and non-unique inverse problem using Bayesian inference. This study applies three variational inference methods (ADVI, SVGD, and sSVGD) to a 3-D FWI problem and compares their performance. The results show that ADVI is the most computationally efficient but underestimates uncertainty, while sSVGD provides the most accurate results at intermediate computational cost.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2023)
Article
Geochemistry & Geophysics
Zeyu Zhao, Mrinal K. Sen, Bertrand Denel, Dong Sun, Paul Williamson
Summary: A hybrid optimization framework is proposed for full waveform inversion (FWI) problems, which incorporates derivative information into the model update rule and improves the convergence speed to tackle the local minima issue of non-linear inverse problems.
JOURNAL OF GEOPHYSICAL RESEARCH-SOLID EARTH
(2022)
Review
Geochemistry & Geophysics
Yudi Pan, Lingli Gao
Summary: Shallow-seismic full-waveform inversion (FWI) is an effective method for accurate reconstruction of near-surface models. However, it suffers from ill-posedness and high computational cost. The recently proposed random-objective waveform inversion (ROWI) method shows better efficiency and robustness than FWI.
SURVEYS IN GEOPHYSICS
(2023)
Article
Geochemistry & Geophysics
Da Li, Michael P. Lamoureux, Wenyuan Liao
Summary: This paper introduces several improvements to the Full Waveform Inversion (FWI) method and demonstrates them with numerical examples. The improvements include using an unbalanced optimal transport distance, constructing a mixed L1/Wasserstein distance, and utilizing entropy regularization and convolutional scaling algorithms to improve efficiency.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2022)
Article
Mathematics, Applied
Chenglong Bao, Lingyun Qiu, Rongqian Wang
Summary: Full waveform inversion is a powerful tool for high-resolution subsurface parameter reconstruction, but it usually requires a good initial model. This study focuses on the impact of source wavelets on the optimization problem and proposes a decomposition scheme. Numerical experiments show that our approach improves the gradient quality in subsequent FWI and provides better inversion performance.
SIAM JOURNAL ON SCIENTIFIC COMPUTING
(2023)
Article
Engineering, Multidisciplinary
Tim Buerchner, Philipp Kopp, Stefan Kollmannsberger, Ernst Rank
Summary: Full waveform inversion (FWI) is an iterative process used to identify parameters of a physical object. It has been successfully applied in seismic imaging and other fields. This paper extends previous research by proposing the use of isogeometric finite cell analysis (IGA-FCM) as the wave field solver and comparing consistent and lumped mass matrix discretization. An adaptive multi-resolution algorithm is also proposed to refine the material grid.
COMPUTER METHODS IN APPLIED MECHANICS AND ENGINEERING
(2023)
Article
Geochemistry & Geophysics
Donggeon Kim, Jongha Hwang, Dong-Joo Min, Ju-Won Oh, Tariq Alkhalifah
Summary: Full waveform inversion (FWI) is a highly non-linear optimization problem that aims to reconstruct high-resolution subsurface structures. In this study, a two-step strategy is proposed to separate a given model into reflectivity and background velocity models and alternately update them using diffraction-angle filtering (DAF) based scale-separation technique, which effectively controls the wavenumber components and reconstructs low-wavenumber structures from the reflected waves.
GEOPHYSICAL JOURNAL INTERNATIONAL
(2022)
Article
Computer Science, Interdisciplinary Applications
Mehran Eslaminia, Abdelrahman M. Elmeliegy, Murthy N. Guddati
Summary: This paper proposes an efficient method for accurately approximating the gradient and the Hessian operator in large-scale problems for full-waveform inversion (FWI). The method utilizes a double-sweeping solver to divide the domain into smaller slabs and sequentially solve the wavefields. By approximating continuity conditions, the long-range coupling between subdomains is relaxed, thus enabling sequential solution. The proposed method, incorporated into an inexact Gauss-Newton approach, computes the gradient and the Hessian vector multiplication more efficiently. Numerical experiments demonstrate that the convergence of FWI is not degraded when using the double-sweeping approximation. Compared to standard FWI, the proposed method is computationally cheaper, making it more efficient.
JOURNAL OF COMPUTATIONAL PHYSICS
(2022)