Research On Seismic Signal Noise Suppression Method Based On Empirical Mode Decomposition (EMD)

In order to solve people's demand for oil,natural gas and other energy sources,seismic exploration has developed from shallow surface to deep exploration.During the exploration process,the monitoring instrument is in a complex natural environment,so that the collected signal contains a large amount of environmental noise,which reduces the quality of the signal and adversely affects the interpretation of the later geological data.Therefore,an effective method is needed to help us obtain high quality seismic signals.In this paper,based on the non-stationary characteristics of seismic signals,the empirical mode decomposition method(EMD)is chosen to suppress the noise in seismic signals.Since the seismic signal may contain an impact amount,the mode component of the decomposition has a mode mixing problem.Combined with auxiliary noise to improve the method,a ensemble empirical mode decomposition method and a complete ensemble empirical mode decomposition method are proposed.Although these two methods solve the mode mixing problem,they have low computational efficiency and introduce false components defect.So,a complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)method is proposed,which has the completeness of the empirical mode decomposition method and has the characteristics of fast calculation speed and high calculation precision.In this paper,based on CEEMDAN combined with wavelet threshold function,the noise suppression of seismic signals is realized,which avoids the problem of removing high-frequency components to achieve the loss of high-frequency seismic signals caused by noise suppression.However,since the basis function of the wavelet transform is fixed and cannot match all the mode components well,the filtered signal still has a lot of high frequency noise.The method of discarding the residual component of the most high-frequency mode component will suppress the noise to a certain extent,which will cause the loss of the effective signal problem.In this paper,the improved wavelet threshold function is applied to each mode component,and the filtered component and residual component are reconstructed to achieve signal noise suppression.However,the spectrum of the noise-reduced signal still contains significant high-frequency noise,and the noise-reduced signal still has the problem of losing effective signal energy.An improved filter-Savitzky-Golay(SG)filter,which is based on least squares smoothing method,is introduced in this paper.The filter can keep the width and height of the peak value of the signal well,and is suitable for the analysis of non-stationary signals.In this paper,the noise suppression of seismic signal is realized based on CEEMDAN and SG filter.The mode component which contains noise is smoothed by Savitzky-Golay filter,and the smoothed amount is superimposed with the unprocessed component to reconstruct the noise suppression signal.A correlation coefficient between the observed signal and each component is calculated by applying the Fréchet distance to determine the mode component containing the main noise.Through the comparative analysis of experimental data,this method can effectively suppress high frequency noise and improve the signal to noise ratio.