Simultaneous Sources Separation Method Via SVD

In the field of seismic exploration,good economic benefit and collection quality have always been the important criterion for the exploration technology.In recent years,the development of simultaneous-source seismic acquisition technology has balanced for both of these indicators.Compared with the traditional acquisition technology,this technique has shorter acquisition time and higher exploration quality.However,due to its technical characteristics,the collected gun sets can interfere with each other,and form strong interference.Therefore,it is one of the application keys of simultaneous-source acquisition technology to realize the separation of the guns.In general,simultaneous-source separation method based on SVD in frequency domain adopts the fixed rank to estimate effective signals.The higher rank can introduce noise,but the rank is too low to damage the effective signals.So the rank estimation is a key point in SVD method.In view of the above problems,this paper proposes a Rank-increasing in frequency(RIF)strategy to realize simultaneous-source data separation.Firstly,the singular spectrum analysis(SSA)method are used to obtain the singular value distribution,and the main energy of effective signal would be reconstructed by the largest singular value.According to the relationship between the effective signal and the crosstalk noise,the noise will be estimated.Then,the noise is subtracted from pseudo data.This is a iterative process.In the subsequent iteration processes,the rank dimension in the SSA method is gradually increased,and the estimation precision of the effective signal will be improved.The processing of synthetic data and actual data shows that the above iterative method can deal with complex data,improve the denoising ability and retain effective signal.In order to prove the feasibility and superiority of the rank-increasing iteration strategy in frequency domain,this paper uses the synthetic complex model data based on the wave equation and actual towing marine acqusition data to test the RIF algorithm,the iterative rank-reduction(IRR)algorithm and the rank-increasing iteration(RI)algorithm in time domain.The results of complex model data and actualocean data processing show that this method can effectively separate simultaneous-source data and significantly improve signal-to-noise ratio.