A Study And Application Of Seismic Data Processing Based On CEEMD

Seismic signal is a kind of complex non-stationary and nonlinear signal. Thetraditional signal analysis technologies which always based on Fourier theory aresuitable for processing stationary signal, but cannot obtain a satisfactory result inseismic signal analysis. Empirical mode decomposition (EMD) is a powerful signalanalysis method for non-stationary and nonlinear signal. EMD decomposes a signalinto a serious of sub-signals called intrinsic mode function (IMF), each IMF hasdifferent frequency components. The application of Hilbert transform on IMFs canobtain instantaneous attributes which have a meaningful physical significance and awide application prospect for seismic data processing and interpretation. This paperstudies the EMD method and its variations, and attempts to apply them for seismicdata processing.This paper first introduces the basic concepts and sifting iterations procedure ofthe EMD method, a simple synthetic signal is also applied to show the result ofdecomposition. EMD method experiences problems of end effects and modes mixing,which affect the accuracy of decomposition seriously, and destroy the physicalsignificance of IMFs. Then this paper introduces a noise-assisted analysis methodcalled ensemble empirical mode decomposition (EEMD). The addition of Gaussianwhite noise solves the mode mixing problem to some extent by using the dyadic filterbank behavior of the EMD. Next, this paper studies the newly variation calledcomplete ensemble empirical mode decomposition (CEEMD). CEEMD can provide abetter spectral separation of the modes and keep the completeness of EMD method. Abetter result is obtained through verifying the CEEMD method on a synthetic signal.The result shows that CEEMD is least affected by modes mixing problem, the methodcan also provide an accurate reconstruction of the original signal. The paper reviews and introduces the basic concepts of traditionaltime-frequency analysis methods include short-time Fourier transform (STFT),continuous wavelet transform (CWT), S transform, and then introduces the complexsignal analysis technology Hilbert transform and a combination method of EMD andHilbert transform concept called Hilbert-Huang transform (HHT). HHT has anexcellent performance in time-frequency analysis of non-stationary and nonlinearsignal, which can provide an obvious higher time-frequency resolution and a moreaccurate spectrum than common methods.Finally, this paper applies CEEMD method to seismic data processing. A bettertime-frequency resolution is obtained through verifying CEEMD method on bothsynthetic and real data, which has the ability to depict different frequency componentsdetailedly. CEEMD also shows a good performance in suppressing swell noise. Thede-noising result of real data analysis demonstrates that CEEMD-based de-nosingmethod in the time domain can eliminate the swell noise effectively while maintainthe effective signal. A satisfactory performance is showed when apply CEEMD tofiltering random noise in f-x domain. Compared with the result of f-x deconvolution,CEEMD-based method in f-x domain can suppress the noise more effectively, andthere is no obvious signal loss.