A Research On Image Simulation Of Seismic Wave Propagation Process In Epicenter And Adjacent Area

The natural earthquake events are shear sources,and the focal depth is usually several kilometers or more.The physical process of the source is longer and the attenuation of the incident waveform is slower.However,the man-made explosion is an expansion source,the depth of the hypocenter is usually less than 1 kilometers,and the duration of the physical process of the source is shorter.The wave rays of this kind of event mainly spread in the soft soil layer of earth crust,so the attenuation of the event waveform is obviously faster.Therefore,the wave propagation processes in the epicentral and adjacent regions of these two events(earthquake and explosion)should be significantly different.There are many observation stations in regions nearby any seismic event,they are discrete observation points in the view of earth surface-the epicentral and adjacent areas.By some interpolation method,any point ever in the position of none observation station in the epicentral and adjacent areas could have a simulated value of seismic wave.In the whole process of event,the simulated seismic wave value in all points of the epicentral and adjacent areas could be calculated by the discrete observation points in the same region.In any moment of the whole process of the event,the simulated seismic wave value in all points could be regarded as a wave field distribution image.The image series of the whole event process is the characteristic of the wave propagation process of corresponding event.According to the attenuation characteristics of the seismic source,and the seismic wave characteristics of the data being collected by discrete monitoring stations,this thesis makes full use of all stations data by constructing difference color image from multi-stations observation of each event in each moment,then by forming the images series for the whole event,finally by extracting the variation characteristics of image series to perform the automatic identification of the seismic source type.This thesis proposes a visualization method:Multi-stations Interpolation Pattem(MIP)--At any simultaneous moment,the all seismic wave values in all the observation stations are visualized as a color image.And all the waveforms of all observation stations in the same area are transformed into a series of color images.Through the change of the color image series,the seismic event could be observed how to spread to the surrounding area;through the characteristics of the propagation and change,the rapid identification of the seismic source type is possible and effectiveThe experimental results show that the average brightness of the MIP spatiotemporal series image of natural earthquake is higher than that of man-made explosion,and the brightness fluctuation of the whole image is lower than that of man-made explosion.The recognition rate of the features being extracted from the MIP series images in the vertical direction(Up-Down,UD)is higher than that in the East-West(E-W)and North-South(N-S)directions.This thesis also studied the effect of sampling time on MIP series of images,found that in natural earthquake and man-made explosion MIP space-time image series of the entire event process in certain period of time has a high recognition rate,in line with the mechanism principle and decay properties of the corresponding source types.The seismic source type identification uses the machine learning logic regression algorithm,uses the least square to carry on the automatic recognition to 62 earthquake events,the average recognition rate achieves 80%,and the highest recognition effect has 90%.After constructing the color time-series images,the coefficients of variation(CoV)with time are extracted,and it is found that the 2 features of kurtosis and variance of the CoV can well identify and classify the seismic source types.