Research On Visual Identifying Distinct Seismic Sources

Waveform visualization is an important subfield of the current very popular research field: scientific visualization. It has been applied in a lot of areas:seismic exploration, man-made explosions monitoring, medical imaging, speech recognition, geophysical exploration, equipment fault detection, identification of power system faults, and so on. By transforming waveform data to some visual graph, the information and pattern being furtively hided in boring common digital number can be intuitively and legibly visualized in some form of graph. Visualization can reveal the entirely overall configuration and changing pattern of the physical quantities or physical processes being represented by the time-stamped time series-the waveform data. In visualization, researchers can easily examine the data in different viewports, alter algorithms which reveal the hided laws, even calculatedly simulate these physical quantities. In time domain, the seismic waveforms of earthquakes and explosions are very similar, and it is hard to distinguish between these two seismic events types. If an appropriate visualization is conducted, and temporal seismic waveform data is displayed in some form of graph, then the visual effects among different seismic sources may be significantly distinctive, which would resulting the possibility of rapidly distinguish distinct seismic sources..Being based on the mechanics of strong tremor events, this thesis investigates the scientific visualization schemes of seismic waveform data. The main research focus is aimed at the two strong tremor events:natural earthquake and man-made explosion. For natural earthquake, the mechanics of seismic source is rather complex, and the time duration of the source physical process is relatively elongated, resulting in slower attenuation of the corresponding waveform signal and wider spectrum of the signal. However, for a single man-made explosion, the ignition and sequel explosion are occurred instantaneously, and the time duration of the source physical process is rather brief, resulting in sharper attenuation of the signal; in addition, the hypocenter is rather shallow, then the high-frequency components of the signal often have been absorbed on the initial transmission way in the soft soil layer of earth’s surface, resulting in narrower spectrum. On these grounds, this thesis mainly investigates one visualization scheme:symmetric dot pattern (SDP). For every one seismic source event, there exist multiple observatory seismic waveform data which being acquired in different locations, each of these seismic waveform data can be displayed as a SDP graph, so that we can get a group of SDP graphs (abbreviated as: SDP matrix) for every one seismic event.The results of experiments in this thesis have showed that:the SDP matrix of natural earthquake is very diverse, whereas that of man-made explosion is relative dull and simple. For the sake of quantitative analysis, we calculate a quantity-the coefficient of variation (CoV) based on SDP, as a new recognition feature for distinguishing distinct seismic sources. The experiments show that the CoV feature is an effectual discriminatory feature. In this thesis, the configurations of SDP control parameters:sampling interval and difference calculating interval are further investigated for re-enforcing potential distinguishing visualization effects and increasing the correct recognition rate of seismic event source type. Finally, this thesis also derives that:if a preprocessing method combining empirical mode decomposition (EMD) and wavelet transform being employed, this preprocessing being equivalent to some kind of noise reduction, then distinguishing visualization effects can be further re-enforced whilst the correct recognition rate of seismic event source type also is improved.