- 428 views
- 1225 downloads
Elastic Least-squares Reverse Time Migration and Elastic Gauss-Newton Full-waveform Inversion
-
- Author / Creator
- Chen, Ke
-
With the fast development of high-performance computing resources, imaging and inversion techniques in the exploration geophysics community are moving from simplified methods to more complex methods that honour as far as possible the physics of wave propagation. Multiparameter imaging and inversion based on two-way wave equation operators are becoming viable for estimating subsurface structures and media properties.
In this thesis, I present new time-domain methods for linearized and nonlinear inversion of elastic wavefields. First, I derive the elastic Born and reverse time migration (RTM) operators for the first-order velocity-stress isotropic elastic wave-equation system. I develop an elastic least-squares reverse-time migration (LSRTM) method with the elastic Born and RTM as forward and adjoint operators. I adopt the conjugate gradient least-squares (CGLS) algorithm to solve the least-squares optimization problem that only requires the action of forward (elastic Born) and adjoint operator (elastic RTM) applied on the fly to vectors. In this case, the Hessian operator of the problem is implicitly inverted via a matrix-free algorithm. The proposed elastic LSRTM can suppress the ubiquitous multiparameter crosstalk artifacts that arise in seismic elastic inversion.
I then draw connections between waveform linearized imaging and full-waveform inversion techniques. I point out that the elastic LSRTM can be viewed as one iteration of the elastic Gauss-Newton full-waveform inversion (FWI) algorithm. This framework offers considerable freedom to design and apply Gauss-Newton FWI algorithms to elastic wavefields. I develop a matrix-free elastic Gauss-Newton FWI method based on the elastic LSRTM code. It consists of two loops of iterations: outer Gauss-Newton nonlinear iterations and inner CGLS linear iterations. The Hessian in each Gauss-Newton iteration is inverted in a matrix-free form that only requires the forward (elastic Born), and adjoint Frechet derivative operator (elastic RTM) applied to vectors.
Numerical tests are utilized to demonstrate the ability of the proposed inversion techniques to effectively perform multiparameter seismic inversion in both the linear and non-linear regimes.
-
- Graduation date
- Fall 2018
-
- Type of Item
- Thesis
-
- Degree
- Doctor of Philosophy
-
- License
- Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.