The Research On Surface-oriented Extended Waveform Inversion

Using the seismic amplitude and phase,full waveform inversion can estimate more accurate subsurface parameters to minimize the difference between synthetic seismic data and real data.It can provide an accurate velocity for pre-stack time migration and pre-stack depth migration and provide reliable velocity data.However,when the collected seismic data is lack of low frequency and long offset information,the conventional full waveform inversion method can only recover the short wavelength component of the subsurface model.In this thesis,our research is on the basis of acoustic FWI in time-space domain.We use the adjoint gradient method for the parallel FWI process,model test prove that conventional FWI can achieve the inversion results with high accuracy when the initial model is quite accurate,on the contrary,if the initial is lack of low frequency information,the inversion result is turned out to be bad.In this paper,we propose a method to improve the long wavelength updates of full waveform inversion which uses gradient separation and recombination.This method decomposes the FWI gradient into a tomographic and a migration term,the tomographic term mainly updates the long wavelengths of the model parameters,the migration term mainly updates the short wavelengths of the model parameters.We use poynting vectors to decompose wave fields and separate gradient.After recombining these two terms,we can achieve a new gradient with a higher weight on the tomographic component.Model test shows that this tomographic waveform inversion using wave decomposition method can accelerate converge rate and improve the inversion result.At last,our paper do the research on the extended waveform inversion,the extended modeling concept links migration velocity analysis to the waveform inversion,on the basis of the concept we built the extended waveform inversion objective function,make best use of the advantages and bypass the disadvantages.There are two types of extended modeling,surface-oriented and depth oriented,our research is surface-oriented extended waveform inversion which has high compute rate.Model test improve that our method can achieve the right result while the conventional method cannot converge.