The Research Of First-arrival Waveform Inversion In Time Domain

Velocity is one of the most important parameters in geophysical exploration processing and interpretation,especially important for the structural description of underground media and reservoir prediction.Parameter inversion and modeling of surface and underground double complex structures is one of the difficult points in seismic data processing.The first arrival travel-time tomographic has a good adaptability to the near-surface region where the longitudinal and lateral changes in velocity are severe.The use of different types of first arrival wave travel time information in the process of restoring the underground medium structure is one of the main means of near surface modeling.However,with the increase of exploration difficulty and the improvement of precision requirements,the accuracy of the first arrival travel-time tomographic inversion cannot meet the existing exploration requirements.Full waveform inversion can accurately obtain parameters such as velocity and density of the underground medium.However,it has strong nonlinearity and requires high accuracy of the initial model,so it is difficult to apply it to actual data processing.The first arrival wave is the information of the seismic wave field within a period of time after the first arrival,its degree of nonlinearity is relatively weaker than the full wave field information,and its inversion process is relatively stable.Therefore,this paper starts with the time domain FWI theory and conducts waveform inversion research on the first arrival wave.This article begins with the effective picking of the first arrival wave,combined with edge protection smoothing operator in signal denoising and edge detection,improved the traditional energy ratio method to pick up the first arrival wave and improve the pickup accuracy.After study of the basic theory of two-dimensional time domain waveform inversion,the finite difference scheme of the acoustic wave equation and the gradient formula of the objective function are derived.In this paper,the first arrival travel-time tomographic is used to provide a more accurate initial velocity model to improve the stability of the inversion.The first arrival travel-time tomographic uses the FMM algorithm to obtain the travel time,and the Rung-Kutta ray tracing to obtain the travel time and calculate the sensitivity matrix,thereby completing the first arrival travel-time tomographic inversion.On this basis,the first arrival waveform inversion is performed to improve the accuracy of the final result.In view of the problem of excessive storage in the conventional first arrival waveform inversion process,this paper introduces L-BFGS algorithm to improve the efficiency of inversion.The L-BFGS algorithm occupies less computer memory and does not need to store Hessian matrix in the inversion process.The Hessian matrix is updated by the previous m-step correction information and the initial positive definite matrix,which has fast convergence speed and high calculation accuracy.Aiming at the accuracy problem of the initial velocity model,the cross-correlation objective function is introduced to recover the low-frequency information of the underground medium,and the possibility that the initial-wave waveform inversion is trapped in the local minimum is reduced.The cross-correlation objective function is only sensitive to the similarity of simulated seismic records and observed seismic records,focusing on phase matching and having stronger linear properties.The first arrival travel-time tomographic inversion results are used as the initial velocity model.Iterative update is performed by adding a new objective function of cross-correlation,which can effectively improve the accuracy of the inversion results and the stability in the inversion process.From the model results,the optimized first arrival waveform inversion has higher computational efficiency and accuracy than the conventional method.