An official website of the United States government
SUMMARY Full-waveform inversion (FWI) methods rely on accurate numerical simulation of wave propagation in the analysed medium. Acoustic or elastic wave equations are often used to model seismic wave propagation. These types of simulations do not account for intrinsic attenuation effects due to material anelasticity, and thus correction techniques have been utilized in practice to partially compensate the anelasticity. These techniques often only consider the waveform amplitude correction based on averaging of overall amplitude response over the entire data set, and ignore the phase correction. Viscoelastic wave equations account for the anelastic response in both waveform amplitude and phase, and are therefore a more suitable alternative. In this study, we present a novel 3-D Gauss–Newton viscoelastic FWI (3-D GN-VFWI) method. To address the main challenge of the Gauss–Newton optimization, we develop formulas to compute the Jacobian efficiently by the convolution of virtual sources and backward wavefields. The virtual sources are obtained by directly differentiating the viscoelastic wave equations with respect to model parameters. In order to resolve complex 3-D structures with reasonable computational effort, a homogeneous attenuation (Q factor) is used throughout the analysis to model the anelastic effects. Synthetic and field experiments are performed to demonstrate the utility of the method. The synthetic results clearly demonstrate the ability of the method in characterizing a challenging velocity profile, including voids and reverse velocity layers. The field experimental results show that method successfully characterizes the complex substructure with two voids and undulating limestone bedrock, which are confirmed by invasive tests. Compared to 3-D elastic FWI results, the presented viscoelastic method produces more accurate results regarding depths of the voids and bedrock. This study suggests that the improvement of imaging accuracy would warrant the widespread use of viscoelastic wave equations in FWI problems. To our best knowledge, this is the first reported study on 3-D GN-VFWI at any scale. This study provides the new theory and formulation for the use of Gauss–Newton optimization on the 3-D viscoelastic problem.
more »
« less
Comparison of machine learning and electrical resistivity arrays to inverse modeling for locating and characterizing subsurface targets
This study evaluates the performance of multiple machine learning (ML) algorithms and electrical resistivity (ER) arrays for inversion with comparison to a conventional Gauss-Newton numerical inversion method. Four different ML models and four arrays were used for the estimation of only six variables for locating and characterizing hypothetical subsurface targets. The combination of dipole-dipole with Multilayer Perceptron Neural Network (MLP-NN) had the highest accuracy. Evaluation showed that both MLP-NN and Gauss-Newton methods performed well for estimating the matrix resistivity while target resistivity accuracy was lower, and MLP-NN produced sharper contrast at target boundaries for the field and hypothetical data. Both methods exhibited comparable target characterization performance, whereas MLP-NN had increased accuracy compared to Gauss-Newton in prediction of target width and height, which was attributed to numerical smoothing present in the Gauss-Newton approach. MLP-NN was also applied to a field dataset acquired at U.S. DOE Hanford site.
more »
« less
- PAR ID:
- 10566033
- Publisher / Repository:
- Elsevier
- Date Published:
- Journal Name:
- Journal of Applied Geophysics
- Volume:
- 229
- Issue:
- C
- ISSN:
- 0926-9851
- Page Range / eLocation ID:
- 105493
- Subject(s) / Keyword(s):
- Geophysics Electrical resistivity Neural networks Machine learning Random forests Boosting
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Sinkhole collapse may result in significant property damage and even loss of life. Early detection of sinkhole attributes (buried voids, raveling zones) is critical to limit the cost of remediation. One of the most promising ways to obtain subsurface imaging is 3D seismic full-waveform inversion. For demonstration, a recently developed 3D Gauss-Newton full-waveform inversion (3D GN-FWI) method is used to detect buried voids, raveling soils, and characterize variable subsurface soil/rock layering. It is based on a finite-difference solution of 3D elastic wave equations and Gauss-Newton optimization. The method is tested first on a data set constructed from the numerical simulation of a challenging synthetic model and subsequently on field data collected from two separate test sites in Florida. For the field tests, receivers and sources are placed in uniform 2D surface grids to acquire the seismic wavefields, which then are inverted to extract the 3D subsurface velocity structures. The inverted synthetic results suggest that the approach is viable for detecting voids and characterizing layering. The field seismic results reveal that the 3D waveform analysis identified a known manmade void (plastic culvert), unknown natural voids, raveling, as well as laterally variable soil/rock layering including rock pinnacles. The results are confirmed later by standard penetration tests, including depth to bedrock, two buried voids, and a raveling soil zone. Our study provides insight into the application of the 3D seismic FWI technique as a powerful tool in detecting shallow voids and other localized subsurface features.more » « less
-
In this paper we propose a quasi-Newton algorithm for the celebrated nonnegative matrix factorization (NMF) problem. The proposed algorithm falls into the general framework of Gauss-Newton and Levenberg-Marquardt methods. However, these methods were not able to handle constraints, which is present in NMF. One of the key contributions in this paper is to apply alternating direction method of multipliers (ADMM) to obtain the iterative update from this Gauss-Newton-like algorithm. Furthermore, we carefully study the structure of the Jacobian Gramian matrix given by the Gauss-Newton updates, and designed a way of exactly inverting the matrix with complexity $$\cO(mnk)$$, which is a significant reduction compared to the naive implementation of complexity $$\cO((m+n)^3k^3)$$. The resulting algorithm, which we call NLS-ADMM, enjoys fast convergence rate brought by the quasi-Newton algorithmic framework, while maintaining low per-iteration complexity similar to that of alternating algorithms. Numerical experiments on synthetic data confirms the efficiency of our proposed algorithm.more » « less
-
In this paper we propose a quasi-Newton algorithm for the celebrated nonnegative matrix factorization (NMF) problem. The proposed algorithm falls into the general framework of Gauss-Newton and Levenberg-Marquardt methods. However, these methods were not able to handle constraints, which is present in NMF. One of the key contributions in this paper is to apply alternating direction method of multipliers (ADMM) to obtain the iterative update from this Gauss-Newton-like algorithm. Furthermore, we carefully study the structure of the Jacobian Gramian matrix given by the Gauss-Newton updates, and designed a way of exactly inverting the matrix with complexity $$\cO(mnk)$$, which is a significant reduction compared to the naive implementation of complexity $$\cO((m+n)^3k^3)$$. The resulting algorithm, which we call NLS-ADMM, enjoys fast convergence rate brought by the quasi-Newton algorithmic framework, while maintaining low per-iteration complexity similar to that of alternating algorithms. Numerical experiments on synthetic data confirms the efficiency of our proposed algorithm. \end{abstract}more » « less
-
We have developed a memory and operation-count efficient 2.5D inversion algorithm of electrical resistivity (ER) data that can handle fine discretization domains imposed by other geophysical (e.g, ground penetrating radar or seismic) data. Due to numerical stability criteria and available computational memory, joint inversion of different types of geophysical data can impose different grid discretization constraints on the model parameters. Our algorithm enables the ER data sensitivities to be directly joined with other geophysical data without the need of interpolating or coarsening the discretization. We have used the adjoint method directly in the discretized Maxwell’s steady state equation to compute the data sensitivity to the conductivity. In doing so, we make no finite-difference approximation on the Jacobian of the data and avoid the need to store large and dense matrices. Rather, we exploit matrix-vector multiplication of sparse matrices and find successful convergence using gradient descent for our inversion routine without having to resort to the Hessian of the objective function. By assuming a 2.5D subsurface, we are able to linearly reduce memory requirements when compared to a 3D gradient descent inversion, and by a power of two when compared to storing a 2D Hessian. Moreover, our method linearly outperforms operation counts when compared with 3D Gauss-Newton conjugate-gradient schemes, which scales cubically in our favor with respect to the thickness of the 3D domain. We physically appraise the domain of the recovered conductivity using a cutoff of the electric current density present in our survey. We evaluate two case studies to assess the validity of our algorithm. First, on a 2.5D synthetic example, and then on field data acquired in a controlled alluvial aquifer, where we were able to match the recovered conductivity to borehole observations.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1016/j.jappgeo.2024.105493
