類神經網路於基礎裸露橋梁耐震能力

Abstract

台灣因為地理環境因素，地形陡峭，河流短而急且降雨集中，河床逐年下降，造成跨河橋梁時常發生基礎裸露的問題，當地震來臨時，受沖刷橋梁的受力情況已異於原設計，為此，針對橋梁的現況做一快速之耐震能力評估便相當重要。本研究係利用公路橋梁耐震設計規範中之靜力分析方法所設計的橋梁為模型，以側推分析 ( Pushover Analysis ) 和容量震譜法 ( Capacity Spectrum Method ) 進行橋梁的耐震能力評估。本研究所分析的橋梁為一跨簡支型式之鋼筋混凝土橋梁，使用有限元素軟體SAP2000建構完成後，輸入適當的柱塑性鉸，配合規範設計地震力下之水平地震力進行非線性靜力側推分析，得到結構物之容量曲線，接著使用容量震譜法評估橋梁的耐震能力；關於耐震能力之計算，ATC-40有建議一套方法，惟此法需繁複迭代且不易收斂的缺點，本文則是參考以結構性能為基準的評估法，於容量震譜上選定任意一結構物之性能點，反推其對應之地表加速度。最後，利用類神經網路 ( ANN ) 模擬整個橋梁耐震能力評估法，判斷橋梁遭沖刷後可承受之降伏及崩塌地表加速度值。由研究結果證實，ANN不僅有效預測出鋼筋混凝土橋梁之耐震能力，也可反映出不同沖刷深度下，橋梁耐震性能之非線性變化的關係。

In Taiwan, the rivers originate in the central part mountains of the island, they have short courses and rapid streams due to the fact of that the terrain is precipitous and the climate is generally marine. The annual typhoons always bring intensive rain, therefore, the problem of abutment scouring of bridges is one of major damages to bridges in Taiwan. Another natural hazard is from earthquakes, when they occur to the bridges that foundation have been scoured, current situation of bridges could not withstand this violent force. This work presents the results of an analytical study on the seismic response of simply supported bridges. The main goal of the work is to predict the seismic capacity for four variables, scouring depth, upper load, diameter of pillar and steel area, in order to investigate the nonlinear behavior of bridges under predefined earthquake force by artificial neural network (ANN) model, especially, the abutment scoured bridge. Furthermore, the modeling approach and appropriateness of analyses, like pushover analysis and capacity spectrum method are also concerned in this work. Finally, the simulated results show that a well-trained artificial neural network can simulate the relations between the seismic responses of bridges and the scoured depth effectively.