台北101之動態響應：從模態分析到預測模型

Abstract

本研究針對 TAIPEI 101 進行長期結構健康監測（SHM）分析，使用近十年高解析度的環境振動資料，透過隨機遞減技術（RDT）擷取模態參數，確認溫度與風速為主導的環境因子。除此之外，我們發現颱風事件會造成短期頻率下降，其幅度與長期趨勢相當，而旋轉模態（Torsion mode）對環境與結構變化則展現出更高的敏感性。除長短期趨勢分析外，本研究亦評估支援向量迴歸（SVR）模型對模態頻率之預測能力，並提出固定式及漸進式兩種模型，儘管漸進式模型的更新能力提升了模型的適應性，固定與漸進模型皆在資料中斷後的動態與未建模之結構變化上表現不佳，顯示現有方法亟需具備物理背景與情境感知能力的預測架構。此外，調諧質量阻尼器（TMD）的異常也揭示了當前 SHM 模型的不足，往後研究時將 TMD 動態資料與其操作情境納入模型中，對於提升預測準確性與結構反應解釋能力至關重要。綜合以上，本研究顯示結合環境資料、長期監測與進階分析技術，我們能有效支持超高層建築之安全性與韌性管理，然尚有部分技術及資料分析有待改進。This study presents a long-term structural health monitoring (SHM) analysis of TAIPEI 101 using nearly a decade of high-resolution ambient vibration data. Through the Random Decrement Technique (RDT), we extracted modal parameters and identified temperature and wind velocity as dominant environmental drivers. Typhoons caused short-term frequency drops comparable to long-term trends, while torsional modes showed heightened sensitivity to environmental and structural changes.Support Vector Regression (SVR) models were evaluated for modal frequency prediction. While incremental learning improved adaptability, both fixed and incremental models struggled with post-gap dynamics and unmodeled structural changes, underscoring the need for physically informed, context-aware frameworks.Anomalies in Tuned Mass Damper (TMD) behavior further highlight gaps in current SHM models. Incorporating TMD dynamics and operational context is essential for improving predictive accuracy and structural interpretation. This study demonstrates the value of integrating environmental data, long-term monitoring, and advanced analytics to support the safety and resilience of supertall buildings.