Research On Seismic Phase Recognition Based On Variational Mode Decomposition

With the rapid development of computer technology and test instruments and systems,earthquake warning technology has become an important means of effectively reducing the losses caused by earthquakes in recent years.Therefore,by picking up the arrival time of the first arrival wave of P-wave and S-wave,the earthquake early warning can be realized,thus reducing the casualties and property losses.In this paper,the variational mode decomposition(VMD)method is introduced into the automatic seismic pick-up and improved.VMD combines with NLMS adaptive filtering algorithm to improve the accuracy and stability of STA / LTAAIC based on the new feature function for seismic phase identification of different signal-tonoise ratios.The main content of the paper includes:Firstly,the traditional time-frequency analysis method is difficult to describe the local characteristics,and the non-stationary seismic signal is difficult to meet the needs of highresolution frequency time amplitude.Therefore,the time-frequency analysis method based on modal decomposition is introduced into the research content of automatic seismic phase recognition.The simulation results show that compared with EMD and its improved algorithm,VMD solves the problem of modal aliasing and has strict theoretical support,so VMD is more suitable for the time-frequency analysis of non-stationary seismic signals.In the time-frequency analysis of the Hilbert transform of VMD,the effect of suppressing the tuning effect is obviously better than that of EMD and its improved algorithm.Secondly,for the end effect of VMD,the waveform mirroring extension method is introduced and the SVR-waveform mirroring extension method is proposed.The simulation results show that the SVR-waveform mirroring extension method is better than the waveform mirroring extension method in suppressing the end effect.Aiming at the problem of VMD empirical presetting,a GA-PSO optimization algorithm based on average amplitude spectral entropy is proposed to search for the optimal parameter combination corresponding to the minimum average amplitude spectral entropy.The simulation results show that the mean value of the optimal average amplitude spectral entropy and the average amplitude spectral entropy converges quickly,the algorithm is stable,and the modal component is almost the same as the original component.Finally,using the improved VMD Hilbert method,a step-by-step seismic phase identification method based on the synthesis of STA / LTA and AIC criteria is proposed.In view of the problem of misrecognition under low signal-to-noise ratio,a STA/LTA-AIC step-by-step recognition method based on the new feature function after VMD-NLMS adaptive filter reconstruction is proposed.The results show that the STA / lta-aic step-by-step recognition algorithm based on the new feature function can effectively identify the seismic phase under the condition of low signal-to-noise ratio.