Study On Joint Velocity Modeling With First Arrival Traveltime Tomography And Waveform Inversion

As the key technology of seismic exploration,seismic velocity modeling is not only related to the quality of subsurface structure migration,but also affects the reliability of seismic data interpretation.The methods of seismic velocity modeling mainly include tomography,migration velocity analysis,stack velocity analysis and full waveform inversion.The first arrival tomography only uses the travel-time information of seismic waves to invert the shallow velocity,which has a better velocity modeling effect on background velocity field with longwavelength component.The full waveform inversion method can reconstruct the physical property parameter field of the subsurface using the full wave field information of the seismic wave,and theoretically obtains the highest resolution underground structure imaging result.The conventional first arrival tomography techniques are computationally expensive in three-dimensional situations due to ray tracing and solving nonlinear equations of chromatography.A method proposed from this thesis can transforms the traditional first arrival tomography into the nonlinear optimization problem by solving the extreme value of the objective functional.Wherein the first arrival can be calculated by the fast sweeping method,and the adjoint state method is used to obtain the inversion gradient.The first arrival tomography can be performed in an iterative inversion manner,and the computational efficiency and memory occupancy in the three-dimensional case are superior to the conventional methods.The full waveform inversion has a perfect mathematical and physical theoretical basis.It uses the full wavefield information to obtain a high-precision,high-resolution subsurface property parameter model,but FWI is difficult to promote in practical applications due to the strong nonlinearity,the high dependence on the initial model and the low frequency information in the data,and the high computational cost.The frequency-multi-scale inversion strategy can realize the frequency-by-band inversion from low frequency to high frequency,and can obtain good inversion results even in the case of poor initial model.In order to speed up the computational efficiency of full-waveform inversion,especially 3D full-waveform inversion,we use the polar coding strategy,combined with the GPU acceleration algorithm,and adopts the multi-scale multiscale inversion strategy of 3D multi-source to reduce the computational efficiency,which can reduce the cycle jump phenomenon and ensure higher computational efficiency.In addition,this thesis proposes a joint-inversion method jointed the advantage of first arrival tomography and full-waveform inversion.It can reduce the dependence of the full waveform inversion on the initial model and weak the cycle jump phenomenon of waveform inversion by establishing a weighted form of joint inversion objective function and adding the gradient of first arrival tomography to the full waveform inversion.Numerical test show that even in the case where the initial model has lower accuracy,the long-wavelength component and the short-wavelength component of the velocity model can still be reconstructed at the same time,but the method has a large improvement effect on the region covered by the radiation.In addition,the algorithm can also be combined with other non-seismic shallow exploration methods,and it has great application prospects in mineral surveys and engineering geophysical exploration.