P-wave Onset Picking Based On Parameters Of Energy And Period

Earthquake disaster prevention and mitigation is the long-term goal of my country’s earthquake disaster mitigation planning.As an effective means of disaster reduction,earthquake early warning has been applied in many countries,such as Japan,the United States and the "National Earthquake Intensity Rapid Report and Early Warning Project" under construction in our country.P wave is the fundamental parameter to realize rapid early warning in earthquake early warning system,and plays an important role in earthquake location,source parameter estimation and impact field estimation in earthquake early warning.For example,the earthquake early warning system uses the travel time information of the P wave to locate the earthquake in real time,uses P-wave period,amplitude and energy information to estimate the magnitude,or directly use the P-wave amplitude information to estimate the ground motion peak value.It can be seen that the rapid and accurate identification of seismic P-waves is the basis for the realization of "quick and accurate" earthquake early warning systems.At present,the automatic firstarrival P-wave pickup method commonly used in earthquake early warning systems is a single-parameter identification method based on the characteristics of signal amplitude mutation.This method is difficult to eliminate the problem of false triggering caused by large-scale background noise and the slow growth of P-wave amplitude.Caused by the problem of lag in pickup results.Therefore,this paper conducts research on the automatic pickup method of P-wave first arrivals in earthquake early warning.The main research contents and results are as follows:(1)In order to select the P-wave and background noise data of ground stations with different noise levels,different earthquake sizes and different distances,this paper selects13455 of 549 earthquakes recorded by the Japanese K-NET network and the China Strong Earthquake Observation Network.The vertical acceleration records are used as a database for method research and evaluation.Among them,2,882 records of 352 earthquakes were from the China Strong Earthquake Observation Network,and 10,573 records of the remaining 197 earthquakes were from the K-NET network.The recorded magnitudes range from 2 to 9,and the epicentral distances range from 0 to 500 km.(2)Several rough first-arrival picking methods(long-and short-term average method,high-order moment method and periodic method)and two P-wave first-arrival precise picking methods(based on AIC criterion and periodic method)commonly used in earthquake early warning systems are analyzed.The problems and the root causes of the problems in the P-wave first-arrival accurate picking method based on the BIC criterion.(3)The differences in energy and period between the P-wave signal and the background noise signal were analyzed,and the differences between the three parameters of the energy gradient,the preeminent period and the preeminent period gradient in the seismic signal and the background noise were analyzed,and a threshold based on three parameters was established.The method of automatically picking up the P wave of the discriminant when it is rough.At the same time,the period parameter is introduced into the AIC accurate picking criterion to solve the problem of the lag of weak signal pickup results.The pick-up accuracy of the P-wave first-arrival pick-up method based on a single parameter and the P-wave first-arrival automatic pick-up method based on the threeparameter threshold criterion are compared.The results show that 88.6% of the pick-up results are within ±0.5s of the P-wave first-arrival pick-up method based on the threeparameter threshold discrimination criterion,while only 76% of the P-wave first-arrival pick-up method based on a single parameter.The P-wave pickup method based on the three-parameter threshold criterion has higher pickup accuracy.(4)An automatic pick-up method based on support vector machine for rough arrival of P wave is proposed.The method takes energy growth gradient and excellent period gradient as input parameters,classifies seismic signals and background noise,and combines the improved AIC method to achieve accurate pickup of P-wave arrivals.The pick-up accuracy of the P-wave first-arrival automatic pick-up method based on support vector machine and the P-wave first-arrival automatic pick-up method based on the threeparameter threshold criterion are compared.The results show that 88.6% of the pick-up results are within ±0.5s of the P-wave first-arrival automatic pick-up method based on the three-parameter threshold criterion,while the support vector machine-based method reaches 92.6%.SVM-based methods have higher pickup accuracy.