Time-varying Wavelet Phase Extraction Method Based On Q Variation Segmentation And Wave Dispersion Prolongation

Due to the earth absorption which causes energy attenuation and phase velocity dispersion during the propagation of seismic wave,real seismic records have high-frequency attenuation and phase distortions.The wavelet phase cannot satisfy the assumptions of zero-phase,minimum-phase or constant phase.At present,the phase extraction is a key topic in the precise wavelet estimation.In order to extract the time-varying wavelet phase from nonstationary seismic records accurately,a phase extraction method based on Q variation segmentation was proposed and phase prolongation was realized by seismic wave dispersion analysis.This thesis finished the following work.Firstly,on account of the time-varying characteristic,the wavelet phase is difficult to be extracted straightly from the non-stationary seismic record.Furthermore,average segmentation and artificial segmentation are unable to describe time-varying characteristic properly.A segmentation method based on Q variation was proposed in this thesis.For the precise estimation of Q values,the time-frequency features of seismic records were analyzed by the modified S transform and the formation quality factor Q was estimated by combination of time-frequency spectral modeling and weighted centroid frequency shift method.In order to embody the time-varying feature of wavelet phase,the seismic data was divided into several segments according to the variation of Q for wavelet phase extraction.The simulation and real seismic data processing results showed that the segmentation results and the time-varying wavelet phase extracted by the proposed method agreed with the time-varying characteristic of real wavelet phase.Secondly,in view of that the length of segmented seismic data is short,the initial wavelet phase within each segment is difficult to be extracted accurately.A phase extraction method based on the bispectrum and the pure phase filter method was proposed by this thesis.The prior estimated phase was obtained by the linear matrix algorithm based on the relation between the double spectrum of seismic record and the wavelet phase spectrum.And then,the parameters of the pure phase filter were given according to the prior estimated phase results and the accurate parameters of wavelet phase were acquired by particle swarm algorithm eventually.The simulation experiments and real seismic data processing results showed that the proposed method can extract the initial phase of each segment in the condition of short data accurately.Finally,the current wavelet phase prolongation methods have not taken the actual circumstances of the wavelet propagation and the formation influence into account.In order to solve the above problem,a phase prolongation method was proposed based on the analysis of the seismic wave dispersion effect.This method analyzed the spread of the wave phase in the formation with constant Q value and the relation between the phase change and Q value was obtained by the solution of wave equation.The wavelet phase prolongation was realized by using this relation and the time-varying phase of wavelets were obtained eventually.The theoretical analysis and numerical experiments comparison were made between the proposed method and mainstream approaches to illustrate the advantages and effectiveness in phase extraction of time-varying wavelet.