Research On The Method Of Rapid Identification And Early Warning Of Seismic Events

Earthquake early warning,as an effective means of earthquake prevention and mitigation,has received significant attention from various countries and regions.Real-time monitoring of seismic motion and rapid estimation of earthquake magnitude play important roles in earthquake early warning.Among them,real-time monitoring of seismic motion involves extracting the arrival time information of seismic phases from real-time seismic waveforms and using it to discriminate triggering events.This process directly affects the accuracy and reliability of subsequent magnitude estimation and location.Rapid estimation of earthquake magnitude is a method to determine the size of an earthquake based on the empirical relationship between extracted feature parameters from the first 3 seconds of seismic data before the P-wave arrival time and the magnitude.The accuracy of this process directly influences the accuracy of subsequent earthquake early warning information.However,current seismic monitoring methods still cannot rapidly and accurately identify earthquake events and explosion events.Traditional single-parameter methods exhibit significant discreteness in rapid magnitude estimation,particularly the problems of overestimation for small earthquakes and underestimation for large earthquakes,which urgently need to be addressed.To address these issues,the main research work conducted in this paper is as follows:（1）Selection and Preprocessing of Earthquake Data.Simulated data from664 earthquake records from the Kik-net seismic network in Japan were chosen for this study.The obtained seismic signals were preprocessed using methods such as a0.075Hz Butterworth high-pass filter,denoising,mean removal,and baseline correction.The STA/LTA（Short-Term Average/Long-Term Average）method was applied with different time window lengths to pick the seismic P-wave arrivals.Experimental results demonstrate that selecting an appropriate time window length can improve the accuracy of picking seismic wave arrival times.（2）Proposed the v_rms-P_d（Root Mean Square Velocity-Displacement Amplitude）compatibility test criterion for discriminating earthquake events and explosion events.The v_rms-P_dcompatibility test criterion divides the compatibility interval into fully compatible,partially compatible,and incompatible regions based on the variance of the v_rms-P_dcompatibility test relationship equation.The seismic event discrimination is determined based on the region to which the calculated results belong.Experimental results show that this method can select potentially destructive earthquake events,thereby improving the discrimination rate of triggering events.It can effectively identify earthquake events and explosion events.（3）Optimization of Support Vector Machine Model Parameters.The seismic motion feature parameters(characteristic periodτ_c,displacement amplitude P_d,and root mean square velocity v_rms)were normalized and assigned different weights to form multiple parameter combinations.These combinations were used as input feature parameters for the SVM-M（Support Vector Machine with Multiple Parameters）model to estimate earthquake magnitudes.Experimental results show that the magnitude estimation method of the SVM-M model improves the discreteness of traditional single-parameter estimation methods and addresses the issue of overestimation for small earthquakes.The accuracy of earthquake early warning level classification is significantly higher than that of single-parameter methods,providing more accurate judgment criteria for earthquake early warning.