Research On Seismic Multiple Attenuation By Sparse Inversion

With more and more complicated objective regions of exploration and gradually popular marine exploration, the serious interference of multiple can’t be suppressed well by the conventional filter method and the wave field extrapolation method. Because of that, a new multiple suppression algorithm based on wave theory is put forward by scholars. As it is completely data-driven and does not require any prior information of underground medium,the Surface-related Multiple Elimination(SRME) algorithm is widely used in industry and has good suppression effect. However, nothing is perfect. There are some problems to this algorithm, such as wavelet effect when replacing the prediction operator with the full wave field and primary damage caused by adaptive subtraction. To solve the problem, the Estimation of Primary by Sparse Inversion method(EPSI) is studied deeply in this thesis. The algorithm is similar to SRME, both of them are based on data matrix and feedback model, but EPSI utilizes iterative inversion to estimate the primary impulse response and source wavelet realizing the direct reconstruction of primary. It can break the two-step strategy of SRME algorithm and avoid the remaining wavelet effect in multiple prediction process and the damage to primary in adaptive subtraction. However, the algorithm requires that the input data must be full wave field regular grid data, and it can hardly be satisfied by general acquisition, so Anti leakage Fourier Transform(ALFT) for seismic data regularization is deeply studied in the thesis. ALFT algorithm eliminates the frequency spectrum leakage phenomenon of irregular seismic data by recursive subtraction, and realizes the accurate estimation of Fourier coefficient, and then the reconstruction and regulation of irregular grid data can be achieved. To settle the computational efficiency bottleneck problem caused by massive seismic data, the GPU/CPU cooperative parallel acceleration technology based on CUDA architecture is introduced in this paper. In the algorithm, a high degree of parallelism is improved by using GPU, and then efficient implementation of the algorithm can be realized.The model data and actual data test show that both EPSI algorithm and ALFT algorithm have achieved good results, meanwhile the speedup ratio of GPU algorithm can be about 70.The algorithm and the corresponding improvement strategy provides some solution for Marine seismic exploration of multiple wave suppression problems in this paper.