利用震動訊號特徵偵測邊坡深層滑動行為

Abstract

台灣位於板塊交界且亞熱帶氣候，造山運動盛行且降雨豐沛，高山面積大且地質破碎，又因地震與豪雨發生頻繁，經常引發邊坡災害，造成生命及財產之威脅，為了減少損失發生，如何監測邊坡破壞的課題逐漸被重視，然而無論是何種邊坡破壞，皆會造成地表震動發生，因此若能從邊坡上的加速度計對震動訊號進行分析，並當偵測到滑動訊號時，迅速發起警戒，即可有助於監測與災害應變。 本研究以蘭台大規模崩塌潛勢區為研究區域，並以11筆地震事件、3筆降雨事件和現地監測資料事件做為研究案例，觀測頻率則為0.1~25 Hz。在11筆地震事件中，先以牛鬥測站震度大於三級進行篩選，再來將微機電系統MEMS1的加速度資料進行短時傅立葉轉換得到時頻圖，初步判定邊坡滑動訊號為17 Hz，並將微機電系統中MEMS1與牛鬥測站與MEMS3的頻譜圖進行比較，並依照震源機制去分類各個卓越頻率段和檢驗邊坡地形效應，本研究亦進行愛氏強度分析與儀器傾斜角分析，結果發現其主要能量累積之方向有指向平行崩塌滑動方向之趨勢而儀器傾斜角則較看不出崩塌滑動方向之趨勢。 在降雨案例中，分析了山竹颱風、利奇馬颱風與米塔颱風三筆降雨資料，並在米塔颱風過後發現邊坡滑動之訊號17 Hz，並將該時間的加速度歷時進行愛氏強度分析與儀器傾斜角分析，所得出之結果與地震相近。 在現地監測資料的統整，利用定置型傾斜儀、孔內伸縮計和地電阻比對微機電系統，並在現地監測資料有反應的時候，微機電系統也能偵測到邊坡滑動訊號，且在愛氏強度分析與儀器傾斜角分析中，也與地震、降雨事件分析一致。此外，在豪雨過後，地電阻剖面顯示出地下水位面變厚，整體視電阻率上升，以此交互比對邊坡滑動之發生。

Taiwan is located at the boundaries of tectonic plates and subtropical climate, which lead to orogenic movement and heavy rainfall. Thus, the mountain area is large and the geology is weak. Moreover, earthquakes and heavy rainfall occur frequently, which often cause serious landslide events and threaten casualties and economy loss. To decrease the loss, monitoring and mitigation measure becomes more important. However, no matter what kind of slope movements occur, ground vibration also generate. Hence, analyzing seismic signals from the station located on the slope, identifying landslide signals, and announcing the alter will be helpful for mitigation. This research locates in Lantai area and focuses on eleven earthquake cases, three rainfall cases and field monitoring. The frequency ranges from 0.1 to 25 Hz. In the earthquake cases, first, use seismic intensity exceeding three on Nioudou station to choose the seismic events. Then, moving windowed fast fourier transform gets MEMS1 spectrum and identifying landslide signal happens on 17 Hz. Next step is comparing the spectrum to Nioudou and MEMS3 of spectrum. Use a focal mechanism to classify each dominant frequency and verify the topography effect. In addition, Arias intensity and tilt angle analysis are also conducted. The results show that the main energy cumulating direction is possible to parallel to the sliding direction and that tilt angle is hard to identify the sliding direction. In the heavy rainfall cases, use moving windowed time-frequency spectrum to analyze three typhoon cases including the Mangkhut, Lekima and Mitag. Then the landslide signal 17 Hz is found after the Mitag typhoon. Also, put this section of acceleration to do Arias intensity and tilt angle intensity. The result is similar to the result from the earthquake cases. In the field investigation, use the in-place inclinometer, borehole extensometer, and ground resistance analysis to compare the MEMS signals. When these monitoring systems have responded, landslide signal from MEMS can be detected. Also, the result of Arias intensity and tilt angle analysis has high consistency to the earthquake and heavy rainfall cases. In addition, resistivity image profiles indicate that the groundwater table will rise and overall apparent resistivity increases after heavy rainfall. Hence, the result can be used to cross-validate whether the slope happens to landslide or not.