Research On The Closed-loop SRME Based On The 3D Sparse Inversion

In traditional marine seismic exploration,multiple is abundant in the subsurface structure since the air-sea surface has high reflectivity.Besides,most of the traditional seismic imaging methods are based on the primary assumption.That’s why multiple is always assumed as noise,which need to be eliminated in the seismic data processing.The following seismic imaging result will be accurate only if the input data is multiple free.If there is still some residual multiples in the input data,it will low the accuracy and validity of the following data processing.So,multiple elimination is always one of the most challenging research topics in the geophysical areas.Currently some new geophysical explorations,which different from the traditional seismic exploration have come to our eyes.Passive seismic exploration chooses to use the source in the subsurface,that is,using the natural earthquake and formation fracture as non-artificial source to generate the seismic waves.Additionally,blending seismic exploration increase the production efficiency greatly because of the short time delay between each shot.Although the data obtained by those three explorations are different,multiple is still regarded as a common problem which they are facing.Therefore,we proposed different multiple elimination method based on traditional Closed-loop SRME(surface-related multiple elimination)for different exploration after we analyzing the characteristic of the data acquired via different ways.After some modifications,primary estimation can be obtained for different acquisitions.First of all,we proposed Closed-loop SRME based on the 3D sparse inversion for traditional seismic exploration.Steepest descent method is replaced by L1-norm bi-convex optimization,which can stabilize the inversion process.Furthermore,to improve the accuracy of the estimated result,3D sparse transform is added into the inversion process as a powerful constraint.Secondly,for noise source passive seismic exploration,we proposed noise source data primary estimation via Closed-loop SRME based on the 3D sparse inversion.Focal transform is introduced into the Closed-loop SRME based on the 3D sparse inversion framework so that noise can be eliminated and travel time errors can be corrected.Finally,for blended acquisition,primary-multiple model for blended data is concluded firstly and then introduced into the Closed-loop SRME based on the 3D sparse inversion framework.We proposed unblended primary estimation from blended data via Closed-loop SRME based on the 3D sparse inversion.For multiple elimination in traditional acquisition,Closed-loop SRME based on the 3D sparse inversion is proposed.Compared with the traditional Closed-loop SRME,we use L1-norm bi-convex optimization to replace the steepest descent framework to stabilize the inversion process.Besides,linear operator multiplication is used to express the primary-multiple model for traditional seismic data.The above modified primary estimation framework is solved by L1-norm spectrum project gradient method(SPGL1).3D sparse transform(2D Curvelet transform and 1D Wavelet transform)is introduced into the inversion process,which can make the estimated result more accurate.The proposed method shows remarkable results on synthetic data and field data test.For multiple elimination in passive seismic acquisition,we proposed noise source passive data primary estimation via Closed-loop SRME based on the 3D sparse inversion.Focal transform is introduced into the Closed-loop SRME based on the 3D sparse inversion after we analyzed passive seismic acquisition,seismic interferometry and the characteristics of noise source passive data.We proposed to use focal transform to eliminate the noise and correct the travel time errors in virtual shot gathers obtained by noise source passive data.After we combined the focal transform and Closed-loop SRME based on the 3D sparse inversion framework,the proposed method can provide noise free primary estimation with correct travel time in the far-offset areas.Synthetic data tests have demonstrated the validity of our proposed method.For multiple elimination in blended seismic acquisition,we proposed unblended primary estimation from blended data via Closed-loop SRME based on the 3D sparse inversion.Blended wave propagation equation of the unblended primary,blending operator and raw data can be obtained after we analyze the blended acquisition and wave propagation.Blended wave propagation equation is introduced into the Closed-loop SRME based on the 3D sparse inversion framework.After such modification,unblended primaries can be estimated directly from the blended data.Synthetic data test and field data test show remarkable and reliable result via our proposed method.After those modification to Closed-loop SRME based on the 3D sparse inversion,primary can be estimated accurately from traditional,passive,blended seismic acquisition.We test our proposed methods on synthetic data and field data,which show accurate and reliable industrial primary estimation.Besides,the proposed methods provide references for multiple elimination via various seismic exploration.