Study On VSP Reverse-time Migration Based On Acoustic And Elastic Wave Equations With Finite-difference Method

Surface seismic exploration technology generally does not provide comprehensive,reliable information for salt and flank structure.However,vertical seismic profile(VSP)technology has an important role in imaging for structure beside wells and salt flank.Reverse time migration(RTM),based on two-way wave equation,could handle with any complex velocity without angle limit,and it is an imaging method with high accuracy.The research on the VSP RTM based on the finite-difference(FD)includes imaging by up-going wavefield or down-going wavefield,FD methods,absorbing boundary condition(ABC),P-wave and S-wave decompsosition,amplitude-preserving,storage strategy,computational efficiency and so on.The main achievements are as follows:Firstly,VSP RTM by direct wave,reflected wave and multiples are discussed.The dissertation analyzes VSP RTM by single source and single receiver seismic data with direct wave,primary,first-order multiple,second-order multiple and higher-order multiples.The VSP RTM by up-going wavefield,down-going wavefield and complete wavefield data except the direct wave(full-wavefield)are compared on three synthetic examples.The results show that full-wavefiled contributes to the imaging.Compared to the conventional migration on VSP data using only up-going primary wavefield,the VSP RTM using full-wavefield could enhance the illumination and enlarge the migration scope.Secondly,optimal acoustic regular grid(RG)FD by least squares(LS)based on 3D time-space dispersion relationship is proposed.The 3D acoustic time-space RGFD based on the Taylor series expansion(TE)is compared with the one based on LS optimization.Their comparisons on dispersion and numerical modeling imply that the LS-based FD leads to less dispersion.In addition,the adaptive variable-length spatial operators are introduced to the 3D FD method which indicates that the adaptive FD could improve the efficiency without decreasing accuracy.Thirdly,the 2D and 3D hybrid ABC based on GPU is developed.The hybrid ABC divides the computational area into inner area,transition area and boundary.The wavefield in transition area is calculated by one-way wave equation and two-way wave equation.Two technical problems on GPU for the conventional hybrid ABC in calculating the transition area by one-way wave equation,time-consuming and chaotic threads,are solved.The hybrid ABC based on central processing units(CPU)is compared with GPU-based one.The numerical modeling and computation time for the two ABCs show the GPU-based hybrid ABC could suppress boundary reflections with high efficiency.Fourthly,elastic VSP RTM is studied.The dissertation compares three kinds of decomposition methods for P-wave and S-wave as well as their imaging results: the Helmholtz decomposition,space domain vector decomposition and wavenumber domain vector decomposition.The VSP RTM images on synthetic and field data indicate that,for the trade-off between imaging accuracy and efficiency,the Helmholtz decomposition is suitable for imaging in simple structure,whereas the space domian vector decomposition is suitable to the complex structure.The consistency of the imaging result with true structure on synthetic data indicates effectiveness of the above methods.Fifthly,the least square RTM(LSRTM)on surface and VSP data are implemented to achieve the images with true amplitude.The results on synthetic and field data show that the LSRTM imaging results have higher resolution and illumination than RTM imaging.Sixthly,the high efficiency approach for modeling and RTM,combined the efficient boundary storage strategy,checkpointing strategy,GPU parallel algorithm with the POSIX thread of CPU parallel algorithm,are developed.The numerical modeling based on CPU is compared with the one based on GPU.Modeling and efficiency analysis indicate that the modeling on GPU could greatly improve the efficiency.Moverover,for considering memory and storage in 3D VSP RTM,the checkpointing is applied.The POSIX thread is introduced at each checkpoints.The dual-parallel strategy for GPU parallel and CPU parallel on 3D RTM imaging could improve the efficiency further by saving the storing and reading time at checkpoints.Seventhly,the 2D acoustic,3D acoustic and 2D elastic VSP RTM imaging is implemented on synthetic and field data.The time-space LS-based RGFD,GPU-based hybrid ABC are applied on 2D and 3D synthetic and field data acoustic VSP RTM and the space LS-based SGFD,perfectly matched layer(PML)ABC are applied on 2D synthetic and field data elastic VSP RTM.The consistency of the VSP imaging result with real structure on synthetic data,and the consistency of VSP imaging result with surface imaging result on the field data indicate the effectiveness of the above methods.