本研究以高速鐵路於橋梁結構之淺基礎段與深基礎段之現地量測資料作為建立資料庫之基礎,其內容將包含近域與遠域以及不同頻率區間之振動資料。延續前期對於高速鐵路引致地盤振動之相關影響因子與頻率因素之研究結果,其中影響因子包含列車車速、結構型式、基礎尺寸、土層種類、剪力波速、量測距離、振動頻率、結構體積、前後背景值、衰減係數、衰減距離等。頻率因素則由前期之建議將振動資料分為整體頻率、低頻-中頻-高頻區間進行分析,探討振動值於各頻率區間下之差異。接著將使用類神經網路、隨機森林及支持向量機等三種機器學習演算法,對上述量測資料進行訓練與測試,並分析不同機器學習演算法之間的表現差異。最後,本研究使用整合性學習將三種機器學習演算法進行整合,分別為投票式整合、堆疊式整合及增強式整合,期望提升單一機器學習演算法之預測表現,並針對不同方式之整合效果進行比較。
This study evaluates various ensemble learning algorithms used to predict the vibrations induced by Taiwan high-speed trains on shallow and deep foundations. A wide variety of field-measured ground vibration data is utilized to analyze the prediction models. The main factors that affect the overall vibration level are established based on the measurement results. These factors are train speed, structure type, foundation size, geological condition, ground shear wave velocity, measurement distance, vibration frequency, background vibration, and attenuation coefficient. Various single machine learning methods, including support vector machine (SVM), artificial neural network (ANN), and random forest model (RFM), are initially used to predict vibration levels. Ensemble learning algorithms, such as voting-based algorithm, stack-based algorithm, and boosting-based algorithm are then adopted to improve the performance of prediction. Analytical results for the single learning methods and ensemble learning algorithms are discussed in detail.