Extended model space full waveform inversion via minimization of virtual scattering source

초록

Basically, FWI is implemented by minimizing an objective function consisting of data residuals. However, many variants of conventional FWI are suggested to overcome or mitigate cycle-skipping problem by extending model space. In these methods, the most critical issue is to derive a convenient way to compute gradient of objective function and apply gradient-descent optimization. In this study, we introduce an inversion method which uses the inverse scattering concept. We calculate virtual scattering sources by solving a least-squares problem consisting of numerical Green's functions and scattered seismic wavefields. The velocity model is updated by the gradient which can be obtained by minimizing the norm of the virtual scattering sources. Also, we could apply the full Newton method with simple calculations because the hessian has only diagonal components. © 81st EAGE Conference and Exhibition 2019. All rights reserved.