透過您的圖書館登入
IP:18.188.154.252
  • 學位論文

辨別地震與非地震訊號之現地即時強震預警系統研發

Research on Distinguishing between Earthquake Signal and Non-Earthquake Signal for On-Site Earthquake Early Warning System

指導教授 : 張國鎮

摘要


鑒於台灣地狹人稠,為增加強震波來臨前的預警時間以及解決區域型地震預警系統警報盲區的問題,現地型強震即時預警系統的建置勢在必行。現地預警系統利用地震初達波前幾秒的特性來預測此地震事件可能達到之最大地表加速度(Peak Ground Acceleration, PGA),根據預測震度大小視情況發布警報。然而,由於現地型系統只使用單一地震測站所記錄的振動歷時資料,便容易受到其他非地震事件所造成的影響,例如車輛、行人、施工等等未知振動源,以及電子訊號等非振動訊號的干擾,而發生預警系統誤報的情形。故本研究擬使用支持向量分類法(Support Vector Classification, SVC)與奇異譜分析(Singular Spectrum Analysis, SSA)等人工智慧(Artificial Intelligence)與訊號處理方法嘗試解決這個問題,期能將容易被高估PGA的非地震事件在預測PGA前先行去除,以減少系統誤報的狀況,並確保實際震度大於四級的地震事件能夠分類正確。 支持向量分類法是一種機器學習(Machine Learning)的技術,其常被應用於資料識別等相關領域,而奇異譜分析為一種時間域訊號處理的分析方法,被廣泛應用在古典時間歷時分析、動力反應系統、多變量統計、多變性幾何、以及各類相關訊號處理的領域。本研究以這兩種理論為基礎,提出了兩種辨別流程法來區分地震事件與非地震事件,並使用來自於中央氣象局與國家地震中心現地測站的地震資料庫,以及收集自現地測站的非地震資料庫,將兩資料庫的加速度歷時資料進行兩階段的測試,以驗證本研究所提方法之可行性。

並列摘要


The regional earthquake early warning system (EEWS) is not suitable for Taiwan due to the fact that most of the destructive seismic hazard comes from in-land earthquakes, which makes the lead-time before destructive earthquake wave arrives provided by the regional warning system can be null. On the other hand, on-site warning system can provide more lead-time at the region close to an epicenter since only the seismic information on the target site is required. Instead of leveraging the information of several stations, the on-site system extracts some P-wave features from the first few seconds of vertical ground acceleration of a single station and performs the prediction of the coming earthquake intensity at the same station according to these features. However, the system may be triggered due to some vibration signals that are not caused by an earthquake event or interference from electronic signals, which may result in false alarm at the station. Therefore, this research performs a study on the classification between true earthquake and non-earthquake events by means of Support Vector Classification (SVC) and Singular Spectrum Analysis (SSA). Support Vector Classification is a machine learning technique that has been widely used in automatic data identification in the past decades. Singular Spectrum Analysis is a signal processing algorithm which is popular for classical time series analysis, multivariate statistics, and dynamical systems. This research proposes two methods to distinguish the vibration signals caused by non-earthquake events from the one caused by earthquake events based on the above two algorithms. The feasibility of the proposed method will be verified by using data collected from Taiwan Strong Motion Instrumentation Program (TSMIP) and earthquake early warning stations of National Center for Research on Earthquake Engineering (NCREE).

參考文獻


[2] Satriano C., Wu Y.M., Zollo A. and Kanamori H., “Earthquake early warning: Concepts, methods and physical grounds,” Soil Dynamics and Earthquake Engineering, 31, 106–118 (2011)
[4] Hsu T.Y., Huang S.K., Chang Y.W.,Kuo C.H., Lin C.M., Chang T.M., Wen K.L., and Loh C.H., “Rapid on-site peak ground acceleration estimation based on support vector regression and P-wave features in Taiwan.”, Soil Dynamics and Earthquake Engineering, Vol.49, pp. 210–217, 2013
[6] Bernd Heisele, Purdy Ho, and Tomaso Poggio, “Face Recognition with Support Vector Machines: Global versus Component-based Approach”, Proceedings of the IEEE International Conference on Computer Vision, vol.2, pp. 688-694, 2001
[7] Harris Drucker, Donghui Wu, and Vladimir N. Vapnik, ”Support Vector Machines for Spam Categorization”, IEEE Transactions on Neural Networks, vol. 10, n0.5, pp. 1048-1054, 1999
[8] Kihoon Choi, Satnam Singh, Anuradha Kodali, Krishna R. Pattipati, John W. Sheppard, Setu Madhavi Namburu, Shunsuke Chigusa, Danil V. Prokhorov, and Liu Qiao, " Novel classifier fusion approaches for fault diagnosis in automotive systems," IEEE Transactions on Instrumentation and Measurement, vol.58, Issue 3, pp. 602-611.

延伸閱讀