Research On Dynamic Deconvolution Method Based On High-precision Time-frequency Analysis

Deconvolution theory is a successful application of invers mathematics in seismic prospecting and is a common method in seismic data processing.In seismic prospecting,the resolution of seismic data is the key factor that restricts the prospecting accuracy.With the development of prospecting technology,the demand of high-resolution seismic data are increasing.However,it is always a difficult problem of obtaining high-resolution seismic data in seismic prospecting.In order to satisfy the required of high-resolution of seismic data,in this paper,we propose a novel high-resolution method which is improved on classical deconvolution.This method improves the resolution of seismic data and supports for the geologists to interpret the abundant geological information in seismic data.The details of the researches and results are as follows:(1)Introducing the dynamic convolution model.This paper has systematically analyzed some common deconvolution methods in high-resolution processing,such as the least squares deconvolution,impulse deconvolution,predictive deconvolution and f-x deconvolution.The deconvolution is all based on stationary convolution model which assumes that the wavelet shape of the seismic wave propagating in the underground medium remains unchanged.However,this assumption is contrary to the actual propagation of seismic waves in underground.Therefore,based on the traditional stationary convolution model,this paper introduces the dynamic convolution model,which makes the deconvolution method proposed in this paper more suitable for the real propagation characteristics of seismic waves.(2)Introducing the four-parameter Generalized S-transform(GST).The seismic signal processed by deconvolution is a typical nonstationary signal.Besides,Time-frequency analysis is a powerful tool for analyzing nonstationary signals.It can clearly shows the characteristics of frequency with time varying.Therefore,it is necessary that selecting a high-resolution time-frequency analysis method to deconvolution seismic signals.In this paper,we compare the Short-Time Fourier Transform(STFT),Continuous Wavelet Transform(CWT),S-Transform(ST)and the GST and find that the four-parameter GST can better consider the resolution of both time and frequency,and thus have good multi-resolution characteristics.Besides,the four-parameter GST is closely related to the Fourier transform(FT).Using the inverse FT,we can directly realize the inverse transformation of GST which is completely lossless.Those characteristics are not only improving the flexibility in actual seismic processing,but also clearly improving the accuracy of seismic data analysis.(3)We propose a dynamic deconvolution method based on four-parameter GST.In this paper,combining the four-parameter Generalized S-transform with high precision and multi-resolution characteristics and dynamic convolution model which can present the time-varying characteristics.Besides,based on the theory of inner convolution,we propose a method named dynamic deconvolution based on four-parameter Generalized S-transform.The proposed method replaces the GST spectra of the non-stationary seismic signal with the product of the dynamic wavelet spectra and the generalized S spectra of the reflection coefficients(similar to the convolution model in frequency domain).Then,applying deconvolution to seismic signal in time-frequency domain.Finally,do lossless S inverse on the results obtained by deconvolution and convolution with the wave.So that,we can obtain a high-resolution seismic data.(4)The results of synthetic model and field data obtained by dynamic deconvolution method based on four-parameter GST verify that the proposed method can effectively improve the resolution of seismic data while the features of seismic data are kept,and compensate energy attenuation,which caused by underground medium during the propagation process.Moreover,the results confirm that the method has high feasibility and practicability in seismic data processing.