Research On Seismic Multiple Denoising Based On Boundary Element Method And Deep Learning

In seismic data processing,multiple wave suppression is an important task,especially in marine seismic exploration,multiple wave problems are more serious.As a kind of noise,multiple waves will interfere with people's identification of effective waves,thus forming false and unreliable seismic imaging,which will affect the accuracy of seismic data interpretation.The quality of multiple wave suppression and the degree of effective wave information retention affect the final accuracy of seismic exploration.Therefore,the multiple suppression method has important research value.The numerical simulation of seismic wave propagation is an important research method for seismic exploration and seismology.It is based on the theory of seismic wave propagation through numerical calculation to simulate the propagation of seismic waves in the medium.Boundary element method and finite difference method are two commonly used numerical simulation methods.Boundary element method is a semianalytical-semi-numeric boundary type method based on integral equations,which has the advantages of dimensionality reduction,analysis,high accuracy,and automatically meeting the far-field radiation conditions;the finite difference method is a typical differential-based region This method has the advantages of high efficiency,practicality and easy numerical realization.Firstly,the related research on boundary element method is done in this paper,including the numerical calculation method of Helmholtz boundary integral equation and the processing of singular integral.Based on the self-interaction of the total boundary integral equation matrix and the different interpretations of the extrapolation operators,a hybrid boundary integral equation cracking algorithm for multi-layered media is proposed,which can extract multiple waves from the full wave field information of seismic waves.Information,which can achieve the purpose of removing multiple waves.At the same time,the algorithm can also solve the problem of forward modeling of the thin-layer boundary integral equation,greatly reducing the calculation amount,and reflecting multiple scattering information.Then,the staggered-grid finite difference method was researched,and the seismic records containing multiples were obtained through finite difference forward modeling,and the seismic records without multiples were synthesized by adding absorption boundaries.Finally,using the seismic data obtained by finite difference forward modeling,deep learning method is used to study the suppression of multiples.The results show that the effect of deep learning methods on multiple wave suppression is at the same level as traditional methods,but it does not require accurate modeling of signals and noise.It can realize adaptive denoising and has potential application value in multiple suppression.