Surface Wave Suppression Of Multi-component Seismic Data Based On Multivariate Empirical Mode Decomposition Method

Surface wave is characterized by low frequency,low velocity and high energy.For seismic reflection exploration,Surface wave is one of interference wave and masks reflected waves from middle-deep layer.Attenuating surface wave is an important step in seismic data processing.In this thesis,based on noise-assistant Multivariate Empirical Mode Decomposition(NA-MEMD),surface wave attenuation method is proposed for multi-component seismic data.In this thesis,Empirical Mode Decomposition(EMD),Multivariate Empirical Mode Decomposition(MEMD)and noise-assistant Multivariate Empirical Mode Decomposition(NA-MEMD)methods are introduced.A multi-component simulation signal,a stack of three different harmonic waves,is decomposed respectively by EMD,MEMD and NA-MEMD methods,which exhibit their advantages and limitation.A serious drawback of EMD method is mode mixing problem for aliasing in the intrinsic mode functions(IMFs),so three different waves cannot be separated completely.MEMD method uses multi-component joint analysis and their similar features to mitigate the mode mixing problem.A better decomposition result is obtained.Independent Gaussian white noises are added to all components in NA-MEMD method,which contribute to completely separate stack signals of different frequency scale.Therefore,The NA-MEMD method is suitable to extract surface wave for multi-component seismic data.In this thesis,I employ NA-MEMD method to decompose seismic data and attenuate surface wave in time-space domain and frequency-space domain respectively.In time-space domain,the decomposition is implemented for time sampling signals trace by trace.Multi-component seismic data are decomposed into the different Intrinsic Mode Functions(IMFs).According to different characteristic frequency scale of seismic reflection wave,surface wave and noise,IMFs including surface wave are removed from all IMFs,and the residue IMFs are reconstructed to attenuate surface wave.In frequency-space domain,the decomposition is implemented for space sampling signals on every frequency data.Firstly,seismic data is transformed from time domain to frequency domain by 1D FFT.Secondly,the signals of real part and imaginary part are decomposed respectively into different IMFs.The IMFs from Low frequency and high wavenumber of surface waves are removed.And the residue IMFs are reconstructed.Finally,a 1D inverse FFT is used to transform free surface wave data to time domain.Multi-component seismic synthetic data and real data are tested by proposed method to attenuate surface wave,the results show that surface waves in all-component data(X,Y,Z component)are well attenuated and high S/N ratio seismic reflected data are obtained.Compared with EMD,F-K filter and band-pass filter methods,the proposed method is better for surface wave attenuation of multi-component seismic data.