利用類神經網路探討建物磚牆提供之勁度

Abstract

在房屋進行結構分析時，磚牆常是被忽略的。但是磚牆的存在使得其四週樑、柱之受 力情形改變，當大地震來襲時，極可能因而發生脆性破壞。磚牆在房屋結構中之分佈並不 規則，若要一一考慮，不僅複雜且極為不便。影響結構物反應之因素有構件撓曲剛度EI、 樓層質量、接頭剛域、結構物之阻尼比等，本文針對構件撓曲剛度EI、樓層質量及磚牆勁 度三個因素來研究。 本研究以類神經網路為方法，利用類神經網路最佳化的功能，求得近似實際結構量測 所得振動訊號時之磚牆等值斜撐桿件有效寬度、構件撓曲剛度及樓層質量。 當神經網路完成學習之後，以其求得等值斜撐有效寬度、構材撓曲剛度、樓層質量 。其結果顯示，等值斜撐有效寬度應為0.2429倍之等值斜撐長度；構材之撓曲剛度EI需提 高至1.207倍；結構物之樓層質量應提高為1.2倍。 Brick wall is usually ignored in structure analysis of a building. But the existence of brick wall will make the change of beams or columns around it, and the brittle failure will happen when it undergoes earthquakes. The distribution of brick wall in a structure of building is too irregular to concern individually. The factors that influence the response of structure a reflexure rigidity EI, mass of floor, rigid zone of beam-column, damping ratio and so on. This thesis only considers EI, mass of floor and stiffness of bric kwall. This thesis aims to find the equivalent bracing width of brick wall, EI and mass of floor which match the measured value by using optimization o fArtificial Neural Network（ANN）. When ANN completes the learning, it will output the three values above.F rom the results, it is noted that the equivalent bracing width should equal to 0.2429 times the equivalent bracing length; EI should be risen to 1.207 times , and the mass of floor should be risen to 1.2 times.