Implicit Seismic Full Waveform Inversion With Deep Neural Representation. Issue 3 (27th February 2023)

• Record Type: Journal Article
• Authors: Sun, Jian; Innanen, Kristopher; Zhang, Tianze; Trad, Daniel
• Is Part Of: Journal of geophysical research. Volume 128:Issue 3(2023)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2023-02-27
• Subjects: seismic inversion -- implicit representation -- deep neural network -- full waveform inversion -- initialization -- uncertainty analysis; Geomagnetism -- Periodicals; Geochemistry -- Periodicals; Geophysics -- Periodicals; Earth sciences -- Periodicals; 551.1
• DOI: 10.1029/2022JB025964
• Languages: English
• ISSNs: 2169-9313

Abstract

Abstract: Full waveform inversion (FWI) is arguably the current state‐of‐the‐art amongst methodologies for imaging subsurface structures and physical parameters with seismic data; however, important challenges are faced in its implementation and use. Keys amongst these are (a) building a suitable initial model, from which a local minimum is unlikely to be reached, and (b) availability of tools for evaluation of uncertainty. An algorithm we refer to as implicit full waveform inversion (IFWI), designed using continuously and implicitly defined deep neural representations, appears in principle to address both of these issues. We observe in IFWI, with its random initialization and deep learning optimization, improved convergence relative to standard FWI model initialization and optimization. Models close to the global minimum, capturing relatively high‐resolution subsurface structures, are obtained. In addition, uncertainty analysis, though not solved in IFWI, is meaningfully addressed by approximating Bayesian inference with the addition of dropout neurons. Numerical experimentation with a range of 2D geological models is suggestive that IFWI exhibits a strong capacity for generalization, and is likely well‐suited for multi‐scale joint geophysical inversion. Plain Language Summary: We propose implicit full waveform inversion (IFWI) by replacing the grid‐based subsurface parameters with a continuous neural network representation. Compared to the conventional full waveform inversion (FWI), this simple reparameterization allows IFWI to start from a random initial model by benefiting from the frequency principle of deep learning optimization. The uncertainty of inversion results using IFWI can be easily performed by adopting the Bayesian neural network, or adding dropout neurons into the Multilayer Perceptron. In addition, one can use a single network to represent physical parameters at different scales, thus accommodating multi‐scale and multi‐physics problems. Synthetic examples demonstrate that IFWI is able to produce a high‐resolution image of subsurface with fine structures, and has strong generalization ability and a certain degree of robustness. Key Points: We propose an implicit full waveform inversion (IFWI) algorithm using continuously and implicitly defined deep neural representations IFWI has the ability to capture high‐resolution heterogeneities of subsurface with random initialization IFWI allows uncertainty analysis without additional calculations and exhibits a strong capacity for generalization, robustness