Study On Converted Wave Imaging For 3D Elastic Reverse-time Migration

Multicomponent seismic techniques can provide better solution for complex reservoir exploration.Converted wave imaging plays significant role in revealing the potential of multicomponent seismic.Elastic reverse time migration(ERTM)is specifically designed for multicomponent data imaging and shows theoretical advantages in converted wave imaging.However,ERTM can be affected by many aspects of the processing flow.In this paper,the whole processing flow for multicomponent data is reexamined and researches are carried out in the aspects of data processing,wavefield extrapolation,imaging condition and computation efficiency.New methods are developed to enhance converted wave imaging.ERTM directly takes the multicomponent data as input.However,the multicomponent are traditionally processed as separated P-and S-wave data,which changes the mathematical appearances as well as the amplitude,phase and travel-time relations among the P-and S-waves.As a result,the data is no more applicable to ERTM.In this paper,the concept of multicomponent joint processing is proposed.Accordingly,a multicomponent joint datuming method is developed as a combination of multicomponent data decomposition,multicomponent P-and S-wave wave equation datuming and multicomponent data reconstruction.This technique makes sure that the data remains multicomponent after the processing and the converted wave are well preserved during the datuming.Then the multicomponent data is propagated into the subsurface to rebuild the wavefield in ERTM,during which numerical dispersion can be serious since the S-wave velocities are typically much lower than the P-wave velocities.To solve this problem,a pseudo-analytical method based on the first-order elastic wave equation is proposed in this paper.And its efficient implementation in the time-space domain is achieved by the pseudo-analytical finite difference(PAFD)method.In order to rapidly obtain the velocity-dependent FD coefficients,the model-adaptive predicted scheme for PAFD is proposed.Numerical tests and dispersion analysis prove that the method is more accurate than the traditional staggered grid finite difference method.After the subsurface wavefield is built,the imaging condition has to be applied to get the converted images.In this paper,the S-wave is regarded as the superposition of the SV-and SH-waves.Based on the analysis of the vector relations among the S-waves and the raypath plane determined by propagations of the incident and reflected waves,scalar SV-and SH-waves are separated from the S-wave obtained with Helmholtz decomposition.Then scalar imaging conditions are implemented to generate scalar SV-P,SV-SV and SH-SH images.Those images have correct polarity distribution and can reveal the properties of the corresponding reflection patterns.Despite of the advantages of ERTM,it is impossible to implement before the computational efficiency becomes acceptable.The disk storage and the amount of computation are two typical concerns.In this paper,wavefield reconstruction technique is applied to significantly reduce the amount of the disk storage at the cost of an extra wavefield extrapolation.And high performance computing based on CPU and GPU is applied to effectively speedup the extrapolation.Since the hard disk I/O is less efficient than the wavefield computing,the total time consumption can be reduced.In summary,the research in this paper makes significant contribution to the ERTM theory.And it will promote the application of the converted wave imaging based on ERTM.