Study Of Blended Acquisition Based Ferquency Domain Full Waveform Inversion

Oil and gas exploration has been gradually moved to ocean exploration from landwork due to the increasing difficulty of the resource survey. In this case, currentexploration technology cannot meet with the demand of resolution. Thus, we have tofurther the high-resolution seismic exploration techniques immediately.Seismic inversion, an important imaging technology, is always the keycomponent of seismic data processing. In the exploration and production industry,several methodologies have been developed to build velocity models of the earth.These methodologies can be mainly divided into three categories: traveltimetomography, migration velocity analysis and full waveform inversion. However, theformer two methods (traveltime tomography and migration velocity analysis) onlyemploy the traveltime information and the kinematic information respectively. A methodusing the whole information contained in the seismic data is needed to produce ahigh-resolution subsurface images. Luckily, FWI is just this kind of imaging method.Pratt and Tanratola proposed a method to implement full waveform inversion infrequency domain after Lines and Kelly gave a first attempt to FWI. And nowfrequency domain FWI has been a major part of the FWI research due to its’ highcomputation efficiency. In this paper, we detailed the problems exist in the frequencydomain FWI and a series of strategies have been proposed to solve them,such as applyL-BFGS optimization algorithm to FWI and employ multi-sources to generate seismicdata. The content of the paper can be described as following:(1) Describe the basic theory and implementation procedure of FWI and analysis theproblems FWI faced.(2)2D-9P finite difference scheme is used to formula the acoustic wave-equation andan iterative method is employed to solve Helmholtz operator as well. In order toincrease the computation efficiency, we proposed three different seismic sourcebased on the L-BFGS optimization algorithm for FWI in frequency domain.(3) Make a comparison between LSQR algorithm-based FWI and L-BFGSalgorithm-based FWI to outline the pros and cons of the L-BFGS algorithm.Meanwhile, we give out the reason why we chose L-BFGS algorithm to solve theobjective function.(4) According to the problems exist in FWI, we have done a lot of numericalexperiments to test the performances of the three different sources and search thepotential problems of the frequency domain FWI.(5) Conclude what we have done and give some advice to solve the problems wehaven’t overcome yet.