基於支撐向量回歸之客製化即時結構受震反應系統

FEASIBILITY STUDY ON ESTABLISHING CUSTOMIZED REAL-TIME SEISMIC STRUCTURAL RESPONSE SYSTEM BASED ON SVR

林子剛(Tzu-Kang Lin) ； 林光奕(Kuang-Yi Lin) ； 林又(Yo Lin)

《中國土木水利工程學刊》 32卷6期 (2020/10) Pp. 483-491

https://doi.org/10.6652/JoCICHE.202010_32(6).0002

摘要

樓層加速度對於支承於結構物之設備及非結構構材的耐震設計上扮演重要角色，過大的樓層加速度足以影響結構物在震後是否仍可維持正常機能運作，尤其對於高科技廠房中之精密儀器。倘能在地震來臨前預測最大樓層加速度（Peak Floor Acceleration, PFA）並提供警報，以適度關閉可能損壞之機具，即可降低損失。本研究擷取單一測站在垂直向地表加速度前3秒之P波參數及結構物之固有特性，共包括平均週期、最大加速度、最大速度、最大位移、累計絕對速度、速度平方積分、結構物樓層高度及固有週期等8個參數，利用支撐向量回歸（Support Vector Regression, SVR），以台灣中央氣象局之結構物強震監測系統自1992年至2016年代表性的地震紀錄進行訓練及測試，建立回歸模型以預測各樓層最大加速度。另利用結構物縮尺模型取代實尺結構物，透過振動台試驗獲得結構物之地震歷時資料庫，建立客製化PFA預測模型，以克服結構物強震監測系統無法全面普及監測之問題。本研究結果有助於設計者對結構物內之設備於地震下的振動特性有進一步的瞭解，提供高樓層結構物震度發布的參考，並在最短時間內針對該結構進行初步健康診斷。

關鍵字

支撐向量回歸 ； P波參數 ； 最大樓層加速度 ； 地震預警系統 ； 地震災害防制 ； 縮尺模型

並列摘要

Floor accelerations play an important role in seismic design of non-structural components and equipment supported by structures. Large floor accelerations may cause damage to service and may also result in structural damage or even collapse of the buildings. For precision instruments in high-tech plants, even for small floor accelerations can cause great damage. Properly shutting down the machine before the severe damage caused by strong earthquake is one of the main purposes of this research. The six P-wave parameters, including peak measurement of acceleration, peak measurement of velocity, peak measurement of displacement, effective predominant period, integral of squared velocity, and cumulative absolute velocity are estimated from the first three seconds of a vertical ground acceleration time history. Then, a new predictive algorithm is proposed which utilizes the previous parameters with the floor height and the fundamental period of the structure as the new inputs of the support vector regression model. Representative earthquakes, which were recorded by Structure Strong Earthquake Monitoring System (SSEMS) of Central Weather Bureau (CWB) from 1992 to 2017, are collected to construct the support vector regression model to predict peak floor acceleration (PFA) of each floor. The result shows that the accuracy of predicting PFA located within one-level difference of the seismic intensity scale of Taiwan is up to 96.96%. The proposed system can be integrated into the existing earthquake early warning system (EEWS) to provide a full protection to life and economy.

並列關鍵字

support vector regression (SVR) ； p-wave features ； peak floor acceleration (PFA) ； earthquake early warning ； seismic hazard mitigation ； reduced-scale model