電機資訊學院: 電子工程學研究所指導教授: 黃俊郎蔡嘉羚Tasi, Chia-LingChia-LingTasi2017-03-062018-07-102017-03-062018-07-102015http://ntur.lib.ntu.edu.tw//handle/246246/276683好的地震定位系統需要精確且快速的定位出地震的發震位置，有準確的測站觸發時間跟能模擬實際地震波傳遞情形的速度模型能夠提高定位的精準度。本研究著重在如何取得能模擬實際地震波傳遞情形的速度模型來提高定位的準確度。 要模擬實際地震波傳遞情形需要知道實際的地底地質結構，而簡單的分層速度模型方便執行及取得但無法精準的模擬實際的地震波傳遞情形。因此本研究提出使用簡單的速度模型及過去地震歷史資料的學習來建地震波傳遞時間模型。此模型可模擬實際地震波傳遞時間跟簡單的速度模型傳遞時間的差值，間接地可模擬實際地質結構跟分層速度模型結構的不同。 另外提出了一個地震定位的演算法，使用地震波傳遞時間模型預測的傳遞時間差值來校正測站觸發時間並定位。實驗結果指出，使用本研究提出的地震定位演算法可以定位出更準確的震央。The objective of an earthquake locating system is to quickly and accurately locate the hypocenter, which requires correct triggered times of earthquake sensors and a proper velocity model. This work is related to the velocity model with which the traveling time from one point to the other can be derived. The real velocity model can hardly be obtained due to insufficient knowledge on the actual geological structure. This research proposes to construct through learning a model for the difference of traveling times obtained by (1) a simple one-dimensional velocity model, and (2) the observed traveling time from historical events. With the traveling time residual model, the locating system can obtain a more accurate traveling time by adding the residual returned by the model to the observed traveling time. An iterative earthquake locating flow that employs the traveling time residual models to calibrate the P wave traveling times is also proposed. Experiment results indicate that the proposed earthquake locating algorithm can improve the locating accuracy.1804371 bytesapplication/pdf論文公開時間: 2016/8/12論文使用權限: 同意有償授權(權利金給回饋學校)地震定位機器學習偏移量校正Earthquake locatingMachine learningCalibrationthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/276683/1/ntu-104-R02943099-1.pdf