A Hybrid Optimization Framework for Seismic Full Waveform Inversion. Issue 8 (14th August 2022)

• Record Type: Journal Article
• Authors: Zhao, Zeyu; Sen, Mrinal K.; Denel, Bertrand; Sun, Dong; Williamson, Paul
• Is Part Of: Journal of geophysical research. Volume 127:Issue 8(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-08-14
• Subjects: Global optimization -- inverse problem -- full waveform inversion; Geomagnetism -- Periodicals; Geochemistry -- Periodicals; Geophysics -- Periodicals; Earth sciences -- Periodicals; 551.1
• DOI: 10.1029/2022JB024483
• Languages: English
• ISSNs: 2169-9313

Abstract

Abstract: A hybrid optimization framework is proposed for full waveform inversion (FWI) problems by incorporating derivative information into the model update rule of a global optimization method called Very Fast Simulated Annealing (VFSA). The proposed optimization framework tackles the local minima issue of non‐linear inverse problems. Additionally, it can converge to the close neighborhood of the global solution from different starting points with an improved convergence speed compared to traditional global optimization methods. Applied to large‐scale FWI problems, the proposed framework greatly relaxes the issue of the dependence of FWI on starting models. Given a proper tuning and sufficient number of iterations, hybrid optimization based FWI can render a good background model, which is a good approximation to the ground truth, even with uninformative prior constraints and poor starting models. The output of hybrid optimization based FWI can be used as the starting model for a subsequent local optimization based FWI run to improve the spatial resolution of the result. The proposed hybrid optimization framework is very general, and can be applied to other linear and non‐linear problems that need optimization loops. Plain Language Summary: Seismic full‐waveform inversion (FWI) is a powerful tool for reconstructing high‐resolution subsurface models. A common practice is to implement FWI with local optimization methods. In order to achieve a geologically meaningful result, an accurate background model is often required by FWI based on local optimization methods. Building such accurate background model, however, can involve long and intensive pre‐processing, and sometimes can't even be achievable due to the limitations of data and methodologies. Global optimization methods relax the requirement of an accurate starting model, but they can be computationally prohibitive in high dimensional FWI problems. We combine advantages of local optimization and global optimization methods and propose a new class of optimization framework named hybrid optimization framework. The new framework can efficiently reduce the objective function and simultaneously explore possible model parameters. The method enables FWI to start with uninformative models and prior bounds, which may allow to dispense with the pre‐processing stage of building an accurate background starting model, making FWI much more automatic. The new method is very general, and can be applied to other geophysical inverse problems that require linear or non‐linear optimization loops. Key Points: We combine a global optimization method with gradient to introduce a hybrid optimization framework for geophysical inverse problems The method is able to efficiently reduce the objective function and explore possible solutions The method enables inversion to start with uninformative prior, and greatly relaxes the dependence of the inversion on starting models