Discrimination Between Earthquakes And Explosions Using Wavelet Packet Transform

Usually, in the process of actual earthquake monitoring and recording waveform analysis, we apply visual analysis and whole judgment to the waveform characteristics based on experience. Then, we can qualitatively identify natural earthquakes from artificial explosions. However, there are some explosions, whose waveform records are similar to that of earthquakes. It is difficult to exclude or identify them through macro features. In the paper, we will apply the method of local spectral density analysis based on wavelet packet to analysis and identification of earthquakes and explosions in the northern part of Ningxia and adjacent areas. By comparing the maximum value of instantaneous spectrum in the same frequency bands of different recording signals, we make a study of quantitative index of identification of earthquakes from explosions, with a view to enhancing effectiveness of identification.On the basis of having been studying the basic knowledge of wavelet and basic skills of Matlab programming, and discussing some key questions of seismic wavelet, noise reduction method of wavelet packet , selection of wavelet basis function and method of time-frequency spectral analysis, we use wavelet packet transform to realize time-frequency spectral analysis of seismic signals and blasting signals. First of all, we carry out wavelet packet decomposition with scale j=5 to seismic signals and blasting signals, then, time-frequency analysis of decomposition signals, and normalized treatment, finally, mapping out normalized time-frequency spectrum. At the same time, we can calculate maximum values of instantaneous spectrum in each decomposition frequency band of P-wave of seismic signals and blasting signals, and make a research on differences among the values in the same band. Similarly, we can make a research on differences among the values in the same band of S-wave of seismic signals and blasting signals. Then, we can obtain a number of quantitative indices and thresholds of identification of earthquakes from explosions. Finally, to enhance effectiveness of the indices, we unify all the individual indices with each other, in accordance with the principle that discrimination results from more than half of the indices are the types of events. The results are as follows. Firstly, there are significant differences between frequency at the maximum value of time-frequency spectrum of P-wave (or S-wave) of seismic signals and that of blasting signals from Yinchuan seismic station. The recognition rate is 84.85% (or 87.88%). Secondly, there are differences between the maximum values of instantaneous spectrum in the two decomposition bands (0.78125 ~ 1.5625 Hz and 1.5625 ~ 2.34375 Hz) of P-wave band (0 ~ 6.25 Hz) of seismic signals and that of blasting signals from Yinchuan seismic station. Thirdly, there are differences between the maximum values of instantaneous spectrum in the six decomposition bands (0 ~ 0.78125Hz, 0.78125 ~ 1.5625Hz, 1.5625 ~ 2.34375Hz, 3.125 ~ 3.90625Hz, 3.90625 ~ 4.6875Hz and 4.6875 ~ 5.46875 Hz) of S-wave band (0 ~ 6.25 Hz) of seismic signals and that of blasting signals from Yinchuan seismic station. Fourthly, with the ten indices unified above, we re-discriminate the earthquakes from explosions in the paper. And discrimination types are all consistent with that of the events. Especially, seven explosions implemented are all distinguished from the events. And the third event, which was considered as earthquake from rapid report of earthquake and proved to be explosion after implementation, is identified as explosion with the synthesis identification criterion.