The Study Of High Precision Spectral Analysis Methods On Seismic Data

Spectral analysis plays an important role in the processing and interpretation of non-stationary seismic signals.By converting one-dimensional signals in time domain to two-dimensional signals in time-frequency domain,the local abnormal information related to the research target could be obtained,and then the delineation of layers and the reservoir prediction as well as hydrocarbon detection could be achieved.However,with the increasing complexity of prospecting objects,the traditional methods have become increasingly incapable of satisfying the demand for fine prospecting.The ability of precise detection and delineation on targeted objects is in need of being enhanced and the available techniques in subject to the features of seismic data are required to be contrived.Aiming at the high-precision spectral analysis of seismic data,this paper focuses on the study of the correlation between the basis function of timefrequency analysis method and signal characteristics,as well as the time-frequency characterization of attenuated seismic channels.Based on the comparison of traditional time-frequency analysis methods,two high-frequency focusing spectral analysis methods are developed to improve the accuracy of time-frequency characterization of seismic data.Traditional time-frequency analysis methods can be divided into linear,bilinear,nonlinear and parametric methods.Based on the principle of the method and the analysis of the basis function,the constant frequency step signal model,the frequency linear change signal model,the frequency nonlinear change signal model and the seismic channel model are designed.The calculation results of these model signals and different time-frequency analysis methods of noise signals are calculated respectively,and the time-frequency analysis methods are compared and analyzed from the perspective of time-frequency spectrum delineation effect and time-frequency concentration.The results show that the linear method and the parametric method are stable and efficient,whereas the time-frequency energy aggregation is poor;Bilinear method has high time-frequency energy aggregation,but it could be interfered easily by cross term energy;the nonlinear method has higher time-frequency energy aggregation,nevertheless,it is likely to be affected by noise,resulting in the instability of the calculation results.The commonly-used time-frequency analysis method in seismic prospecting usually requires that the frequency characteristics of non-stationary signals do not change or change linearly in the local time domain,then,based on the orthogonal timefrequency grid or rotating time-frequency grid,a line segment parallel to the time axis or frequency axis or with a certain slope is used to approximate the time-frequency characteristics of the signal in the local time-frequency plane.However,the local time-frequency characteristics of non-stationary signals are often nonlinear in practice,so the time-frequency analysis method based on local polynomial Fourier transform is proposed.In the local time-frequency plane,the timefrequency grid is allowed to bend,and the high-order polynomial is used to fit the nonlinear change of the instantaneous frequency to improve the energy concentration of the time-frequency representation.The application effects of practical seismic data show that the local polynomial Fourier transform method can effectively describe the instantaneous frequency and amplitude variation characteristics of local seismic signal,and the accuracy of target detection with frequency and amplitude anomalies can be enhanced in the meanwhile.In addition,due to its advantages in signal representation,the matching pursuit method has been widely used in seismic data processing whose effectiveness depends on the similarity between the time-frequency atom and the local characteristics of the analyzed signal.If the resemblance between them is high,fewer time-frequency atoms are required to realize signal sparse decomposition,while if that is low,not only the representation of the local characteristics of the signal by decomposed time-frequency atoms is difficult to complete,but more atoms are also needed to approach the signal.Therefore,the definition and selection of time-frequency atom library is critically important.The traditional time-frequency atom usually includes four parameters,namely scale factor,center time,center frequency and phase factor.For the conventional seismic data,the time-frequency relationship of seismic channel is relatively gentle,and the four parameter time-frequency atom can generally meet the requirements of seismic channel decomposition.However,when the seismic data encounter strong attenuation,the time-frequency relationship of the seismic channel becomes more intense.Under such condition,the four parameter time-frequency atom is hardly to effectively characterize the time-frequency characteristics of the signal.Therefore,the matching pursuit method based on the rotating time-frequency atom library is proposed.Based on the conventional four parameters,the frequency rotation operator is introduced to build a five parameter time-frequency atom library,which can realize the sparse decomposition and time-frequency characterization of the attenuation channels.The application of practical data shows that the rotating time-frequency atomic library can effectively describe the attenuation characteristics of seismic traces,and the prediction of reservoirs with attenuation characteristics can be accomplished precisely by the combination with spectral decomposition and other technologies.