發(fā)布時間：2018-04-30 11:57

  本文選題：地震預警 + 烈度預測�。� 參考：《中國地震局工程力學研究所》2017年碩士論文

【摘要】：地震預警是近二十年來新發(fā)展起來的一種減小地震災害的有效手段,其利用從地震發(fā)生到破壞性地震波到來之前的這段時間快速獲取并向公眾發(fā)布地震信息,從而達到減小人員傷亡和財產(chǎn)損失的目的。越來越多的研究者通過利用地震實時波形快速計算的地震儀器烈度來快速評估地震破壞程度。為了能夠在地震預警的過程中利用P波初始信息快速獲得地震儀器烈度的估計值,本文提出兩種實時持續(xù)預測儀器烈度的辦法,并利用研究成果對四次不同臺網(wǎng)記錄到的地震進行測試和分析。本文所做主要工作如下:(1)搜集并選取了1999-2016年日本KiK-net強震臺網(wǎng)記錄到的629次地震,以及基于各自臺網(wǎng)記錄到的2008年汶川地震、1999年集集地震、2014年魯?shù)榈卣鸬膹娬鹩涗?對這些強震記錄進行了預處理和篩選,并通過STA/LTA法和AIC法對震相進行了自動撿拾。對所有選擇臺站的震源參數(shù)信息以及觸發(fā)后1-20秒的7個地震動參數(shù)進行了提取,并計算每個臺站的最終儀器烈度。(2)提出一種持續(xù)預測地震儀器烈度的辦法,利用P波觸發(fā)后實時的PGA和PGV對最終儀器烈度進行持續(xù)預測。地震發(fā)生后,儀器烈度的變化呈現(xiàn)一定的規(guī)律性,該方法通過函數(shù)來描述這種烈度的增長形式,并找出函數(shù)中的參數(shù)與震源距的關系。對上述方法的預測效果進行檢驗,并對結果進行改進,將每秒的烈度值分開利用,對臺站觸發(fā)20秒內(nèi)的烈度值和最終烈度值的關系進行了統(tǒng)計,得到了三種模型。最后,本文對模型進行了檢驗,并得到了相對穩(wěn)定可靠的結果。(3)提出了一種基于人工神經(jīng)元網(wǎng)絡的地震儀器烈度辦法,同樣利用P波觸發(fā)后的信息,根據(jù)地震預警中信息獲取的實際情況建立了三種模型,分別對P波觸發(fā)后1-20秒的數(shù)據(jù)進行統(tǒng)計分析,并選擇合適的樣本對網(wǎng)格進行訓練。三種模型有著不同的適用范圍,預測結果也略有不同,隨著P波觸發(fā)后時間的推進,模型的預測結果也在逐步改善,到達20秒時,可得到相對較好的預測結果。(4)使用上述兩種儀器烈度預測方法對4次不同臺網(wǎng)記錄到的地震事件進行了烈度預測,并對具有代表性的11個臺站的預測結果進行了分析和對比。在此基礎上,確定了模型的適用范圍,并總結了一套儀器地震烈度連續(xù)預測的方法。
[Abstract]:Earthquake warning is an effective means to reduce earthquake disaster, which is developed in the past two decades. It uses the period from earthquake occurrence to the arrival of destructive seismic wave to quickly obtain and release earthquake information to the public. In order to reduce casualties and property losses. More and more researchers estimate the degree of earthquake damage by using seismic real-time waveform to calculate the intensity of seismic instruments quickly. In order to obtain the estimating value of seismic instrument intensity quickly by using the initial information of P wave in the process of earthquake early warning, this paper puts forward two methods to predict the intensity of the instrument in real time and continuously. Four earthquakes recorded by different network are tested and analyzed by using the research results. The main work of this paper is as follows: (1) collected and selected 629 earthquakes recorded by Japan's KiK-net strong earthquake network from 1999 to 2016, as well as strong earthquake records of 2008 Wenchuan earthquake, 1999 Jiji earthquake and 2014 Ludian earthquake recorded by the Japanese strong earthquake network. These strong seismic records were pretreated and screened, and the seismic phases were automatically picked up by STA/LTA and AIC methods. The source parameter information of all selected stations and 7 ground motion parameters of 1-20 seconds after trigger are extracted, and the final instrument intensity of each station is calculated. Real time PGA and PGV after P wave trigger are used to predict the final instrument intensity continuously. After the earthquake occurs, the variation of the instrument intensity presents a certain regularity. This method describes the increasing form of the intensity through the function, and finds out the relationship between the parameters of the function and the focal distance. The prediction results of the above methods are tested, and the results are improved. The intensity values per second are used separately, and the relationship between the intensity values and the final intensity values within 20 seconds triggered by the station is statistically analyzed, and three models are obtained. Finally, the model is tested, and a relatively stable and reliable result is obtained. A method of seismic instrument intensity based on artificial neural network is proposed, which also uses the information of P wave trigger. According to the actual situation of information acquisition in earthquake early warning, three kinds of models are established, the data of 1-20 seconds after P wave trigger are statistically analyzed, and the appropriate samples are selected to train the grid. The three models have different range of application, and the prediction results are slightly different. With the advance of the time after P-wave triggering, the prediction results of the three models are gradually improved, reaching 20 seconds. A relatively good prediction result can be obtained. (4) using the above two kinds of instrument intensity prediction methods, the intensity prediction of 4 seismic events recorded by different network is carried out, and the prediction results of 11 representative stations are analyzed and compared. On this basis, the applicable range of the model is determined, and a set of methods for continuous prediction of the seismic intensity of the instrument are summarized.
【學位授予單位】：中國地震局工程力學研究所
【學位級別】：碩士
【學位授予年份】：2017
【分類號】：P315.7

【參考文獻】

相關期刊論文 前10條

1 金星;張紅才;李軍;康蘭池;韋永祥;馬強;;地震儀器烈度標準初步研究[J];地球物理學進展;2013年05期

2 金星;張紅才;李軍;韋永祥;馬強;;地震預警震級確定方法研究[J];地震學報;2012年05期

3 許沖;徐錫偉;;基于GIS與ANN模型的地震滑坡易發(fā)性區(qū)劃[J];地質科技情報;2012年03期

4 金星;張紅才;李軍;韋永祥;馬強;;地震預警連續(xù)定位方法研究[J];地球物理學報;2012年03期

5 楊帆;鄭寶柱;剡亮亮;;基于BP神經(jīng)網(wǎng)絡的地震傷亡人數(shù)評估體系研究[J];震災防御技術;2009年04期

6 康蘭池;金星;;四川地區(qū)地震動峰值衰減規(guī)律研究[J];地震學報;2009年04期

7 雷建成;高孟潭;俞言祥;;四川及鄰區(qū)地震動衰減關系[J];地震學報;2007年05期

8 俞言祥;汪素云;;中國東部和西部地區(qū)水平向基巖加速度反應譜衰減關系[J];震災防御技術;2006年03期

9 于山;王海霞;馬亞杰;;三層BP神經(jīng)網(wǎng)絡地震災害人員傷亡預測模型[J];地震工程與工程振動;2005年06期

10 金星,馬強,李山有;利用數(shù)字強震儀記錄實時仿真地動位移[J];地震學報;2005年01期

相關博士學位論文 前2條

1 張紅才;地震預警系統(tǒng)關鍵技術研究[D];中國地震局工程力學研究所;2013年

2 馬強;地震預警技術研究及應用[D];中國地震局工程力學研究所;2008年

相關碩士學位論文 前3條

1 李水龍;地震儀器烈度計算方法初步研究[D];中國地震局工程力學研究所;2014年

2 鄭宇;基于強震動記錄P波的后續(xù)地震動估計[D];中國地震局工程力學研究所;2013年

3 馬東;基于福建區(qū)域地震臺網(wǎng)的高速鐵路預警能力評估及烈度預測方法探討[D];中國地震局工程力學研究所;2013年

，

本文編號：1824469