以類神經網路作鋼筋混凝土結構最佳化設計

淡江大學機構典藏 > 工學院 > 土木工程學系暨研究所 > 學位論文 > Item 987654321/94327

請使用永久網址來引用或連結此文件: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/94327

題名:	以類神經網路作鋼筋混凝土結構最佳化設計
其他題名:	Optimal design of reinforced concrete structures using artificial neural networks
作者:	陳志偉;Chen, Chih-wei
貢獻者:	淡江大學土木工程學系碩士班高金盛;Kao, Chin-Sheng
關鍵詞:	類神經網路;鋼筋混凝土;最佳化設計;artificial neural networks;Reinforced Concrete;Optimal Design
日期:	2013
上傳時間:	2014-01-23 14:21:29 (UTC+8)
摘要:	在傳統鋼筋混凝土結構設計中，結構尺寸與造型配置等變數一般都是利用以往經驗來設定初始值，例如梁柱長度與斷面型式、尺寸等等，事實上，這些變數之間都是相關的。工程師因設計時間過於倉促，最終設計成果往往未能有效率地使各個變數間發揮其最佳效能。而結構最佳化設計問題屬於具有限制條件、非線性以及離散變數的最佳化問題，傳統最佳化設計方法例如梯度法或線性規畫等方法並不適用於求解此類問題。類神經網路法有別於以往傳統方法，是一種平行分散式處理計算模式。其基本的運作原理乃以大量、簡單的處理單元，或稱神經細胞互相連接，藉由整體處理單元對外界輸入訊號的簡單運算來處理資訊，擁有類似於人腦的許多特性及優點。本研究主要目的在於將類神經網路法配合交叉驗證法與訓練測試法運用於鋼筋混凝土梁構件、柱構件、梁柱構架等結構成本最少化之最佳化設計，本文亦進一步綜合建立鋼筋混凝土建築結構成本最少化之最佳化設計模式。利用本文之研究成果，可使傳統僅滿足耐震需求的鋼筋混凝土建築結構，進一步達到成本最少化之最佳化設計。 In the traditional design of reinforced concrete structures, structural variables such as size and shape configurations generally use past experience to set the initial value such as the length, cross-section type and size of beam and column. In fact, these variables are correlated. Due to overly hasty design, the final results designed by engineer often failed to efficiently play the best performance for all structural variables. The optimization designs of structure belong to the constrained, nonlinear and discrete variable optimization problems. The traditional design methods such as optimal gradient method or linear programming are not applicable for solving this type of problem. Artificial Neural Networks method is different from the traditional one which is a kind of parallel distributed processing computing model. The basic principle of operation is based on a large but simple processing unit, or called Neuron connected to each other; by using the whole processing unit by the simple arithmetic of the external input signal to process information, which is similar to many features and benefits of human brains. Main purpose of this paper lies in the Artificial Neural Networks training with cross-validation method and Train-and-Test method used in reinforced concrete beam members, column members, beams and other structural cost minimization framework of optimal design. The paper has further consolidated to establish the cost minimization of reinforced concrete buildings of structural optimization design patterns. Using the results of this study can make traditional reinforced concrete structure which only meets seismic requirements further achieve cost minimization of the optimal design.
顯示於類別:	[土木工程學系暨研究所] 學位論文

文件中的檔案:

檔案	大小	格式	瀏覽次數
index.html	0Kb	HTML	294	檢視/開啟

在機構典藏中所有的資料項目都受到原著作權保護.

TAIR相關文章

資料載入中.....