Research On Surface-Related Multiple Attenuation By Sparse Inversion

With the increasing complexity of exploration target areas and the rising of marine exploration,the suppression of multiple waves has become a key issue to be solved urgently.When the underground geological structure is complicated,the filtering method based on the signal processing theory can not achieve a satisfactory multiple suppression result.Geophysicists proposed the Surface-related Multiple Elimination method(SRME),which is based on wave theory.This method is completely data driven and does not require any prior knowledge of the underground.So it is widely used in the multiple suppression of complex media.However,there still are many limitations,mainly due to the influence of the lack of near offset on the prediction of multiple waves,and the risk of harmful waves due to adaptive subtraction.In this paper,the surface multiple suppression method by sparse inversion(EPSI)is studied.This method is derived from the same feedback model as the SRME method.However,unlike the SRME method,EPSI does not need to introduce adaptive operators to adjust multiples,Instead,it takes the primary wave impulse response and source wavelet as unknown quantities,and iteratively inverts the values of both by the conjugate gradient method,and then reconstructs primary and free surface multiples,eliminating the adaptive subtractive step and avoiding the damage to the primary.In addition,the EPSI method can also be used to reconstruct the missing near-offset data.Compared with the SRME algorithm,Compared with the SRME method's over dependence on near-offset data,the method has obvious advantages.However,as an iterative inversion method,the high computational cost is the main problem that restricts its development.This paper proposes a new GPU acceleration strategy to improve the computational efficiency of the EPSI method.Compared with the traditional acceleration strategy,this strategy considers the data transmission factors and further improves the algorithm by reducing the data transmission between the host and the device segment.Computational efficiency.The model test shows that the EPSI method can effectively suppress multiples.Compared with the SRME method,the EPSI method can better protect the primaries.Applying the new GPU acceleration strategy to the EPSI algorithm can improve the computational efficiency under the premise of guaranteeing the calculation accuracy...