類神經網路於隔震結構動力分析之應用

Application of the A Neural Network Approach to Dynamic Analysis of Base Isolated Structure

謝景棠(Ching-Tang Hsieh)

指導教授 ： 陳振華 ； 鄭金國

中原大學/工學院/土木工程研究所/碩士(2004年)

https://doi.org/10.6840/cycu200400511

摘要

中文摘要

台灣位居於地震頻繁區域，對於地震的威脅，時常令人擔憂，早期的結構耐震設計是以建築物本身的強度及韌性來抵抗地震。近一、二十年來，由於科技進步，發展出的隔震消能技術，以隔震系統減少結構物所受之地震力，或以提供的消能裝置降低結構物之反應。然而，對於裝設隔震消能系統之結構承受地震力作用時，常使得隔震消能系統進入非線性，而結構其他部分則尚在線性範圍。傳統分析方法中，須先依據力學觀念建立隔震消能元件之非線性力學行為模式。事實上，準確地建立隔震消能元件之非線性模式，並非易事。因此，本文旨在利用類神經網路建立此非線性行為模式，然後結合傳統線性結構動力分析程序，進行結構動力分析。

由於類神經網路是一種非傳統的運算方法，其具有容錯性及可訓練性，對於非線性問題之處理具有相當效率與精度，因此本文擬首先利用類神經網路學習隔震裝置(例如LRB)的非線性動力行為，然後結合傳統結構動力數值分析方法進行含隔震裝置之建築物的動力分析，並與傳統非線性分析方法所得者比較，以探討類神經網路對於此一非線性問題處理之可行性，並建立相關分析方法，作為日後分析含隔震裝置之建築物行為模擬之依據。

關鍵字

類神經網路 ； 非線性分析

並列摘要

Abstract

Taiwan is a seismically active area, so that the seismic resistance capacity is a main concern in structural design. During the recent two decades, base isolation devices and energy dissipation devices have provided a new and efficient alternative for decreasing the seismic hazard for structures. By implementing these devices in a structure, the structure itself will often be its linear elastic range during an earthquake, while the devices will perform their inelastic behaviors. In a conventional analysis, the model of the nonlinear behavior for the devices has to be established based on some mechanics principles. However, it is usually not an easy to establish the mechanic model to describe the complex nonlinear behavior of a device. Consequently, the main purpose of this present is to apply the neural network technique to learn the complex nonlinear behavior of a device; then, to combine the network with traditional structural dynamic analysis procedure to predict the responses of a structure with the devices subjected to earthquake input.
The neural network technique has caught researchers’ attention in the recent two decades, due to its powerful and adaptive abilities to treat various complex problems. However, this is the first time to incorporate neural network technique into traditional structural dynamic analysis. The neural network will be used to learn the nonlinear behavior of base isolation devices like LRB. Then, the established neural network will be implemented traditional procedure of structural dynamic analysis for predicting the responses of a five-story shear building with the base isolation device subjected to earthquake input. The results will be compared with those from a traditional approach to show the applicability of the proposed procedure in this present.

並列關鍵字

neural network ； nonlinear analysis

參考文獻

張哲維，類神經網路於有限元素模式修正之應用，碩士論文，私立中原大學 土木工程學研究所，民國90年7月。

蘇信華，類神經網路於模態識別之應用，碩士論文，私立中原大學 土木工程學研究所，民國90年 7月。

Bodri, B.,“A neural-network model for earthquake occurrence”,Journal of Geodynamics, 32, 289-310 (2001).

Chassiakos, A.G. and Masri, S.F.,“Modelling unknown structural systems through the use of neural networks”, Earthquake Engineering & Structural Dynamics, 25(2), 117-128 (1996).

Dai, H. and MacBeth, C.,“Application of back propagation neural networks to identification of seismic arrival types”, Physics of the earth and planetary interiors, 101, 177-188 (1997).