基因演算法和類神經網路在斜張橋最佳化設計及健康診斷之應用

Application of Genetic Algorithm and Neural Network to Optimal Design together with Health Diagnosis of Cable-Stayed Bridge

林祐川(Yu-Chuan Lin)

指導教授 ： 宋裕祺

國立臺北科技大學/工程學院/土木與防災研究所/碩士(2005年)

摘要

人工智慧(Artificial Intelligent )發展迄今已半世紀，由於電腦運算速度的日益提升使其在各領域的應用在最近幾年更加廣泛，本文採用人工智慧中的基因演算法( Genetic Algorithm)和類神經網路( Artificial Neural Network )作為研究工具。其中基因演算法在處理最佳化問題上有很好的能力，該方法會依照問題特性利用啟發性的搜尋方式來尋求最佳解，所需計算時間也較其他最佳化理論所需者快速，加上基因演算法本身具有跳脫局部極值的能力，這些優點都是基因演算法逐漸受到重視的原因。而類神經網路憑藉其優良的自學習和自適應能力，在範例資料健全的情況下能從中整理出正確的對應規則，使其處理複雜的對應問題上有極高的評價。
本文利用基因演算法的優點和特性來協助我們求解結構最佳化設計的問題。以往斜張橋最佳化設計多是利用數學規劃法來求鋼索預拉力的組合，本文則分別採用基因演算法和混合基因演算法( Hybrid Genetic Algorithm)進行求解。至於結構健康診斷則分別利用類神經網路和基因演算法進行靜態識別，其方法是藉由斜張橋橋拱頂部的轉角來反推纜索的內力組合，並求出各構件對應的安全係數，作為結構健康診斷的依據，所得之結果冀能提供為類似工程參考之用。

關鍵字

斜張橋 ； 最佳化設計 ； 結構診斷 ； 基因演算法 ； 類神經網路

並列摘要

Genetic algorithm (GA) and Neural Networks (NN) are two important approaches of artificial intelligence (AI) to deal with highly nonlinear problems effectively. With the strong ability in searching global minimum or maximum, GA is recognized as a powerful procedure for optimization. With the significant ability in learning, NN is able to reflect the nonlinear mapping relationships between input and output. As a result, GA and NN have been successfully applied to various engineering field. However, few of their applications to cable-stayed bridge were found. This thesis thus focuses on the applications of GA and NN on the structural optimal design and structural health diagnosis of cable-stayed bridge.
Based on the structural optimization approach, this thesis uses the theory of minimum strain energy of the bridge in deriving the objective function as the quadratic form of the post-tensioning cable forces. In addition, the equality constraints for the restriction on the displacements of the pylon and the un-equality constraints for the limitation on the envelopes of the cable forces are both implemented in the optimization model. GA is then conducted to find the post-tensioning cable forces of the bridge for the structural optimal design. Besides, GA and NN are used, respectively, to find the corresponding post-tensioning cable forces of the bridge subject to the measured rotations of pylon for the structural health diagnosis.
The Mau-Lo Hsi Cable-stayed Bridge is adopted as a case study. The results obtained revealed that the presented method indeed fulfill the structural optimal design and the structural health diagnosis and might be a useful reference for similar bridge engineering.

並列關鍵字

Cable-Stayed Bridge ； Optimal Design ； Health Diagnosis ； Genetic Algorithm ； Artificial Neural Network

參考文獻

40. 宋裕祺、張荻薇（1998），斜張橋之最佳化設計，中國土木水利工程學刊，第十卷，第二期。

1. Ali N, Behdinan K, Fawaz Z. (2003). “Applicability and Viability of a GA Base Finite Element Analysis Architecture for Structural Design Optimization,” Computers and Structures, Vol. 81, pp.2259~2271.

2. Camp C, Pezeshk S, Cao G. (1998). “Optimized Design of Two-Dimensional Structures Using A Genetic Algorithm,” Journal of Structural Engineering, Vol. 124(5), pp.551~559.

3. Coello CAC. (2000). “Use if a Self-adaptive Penalty Approach for Engineering Optimation Problems,” Comput Industry, Vol. 41, pp.113~127.

4. Goldberg D E. (1989). “Genetic Algorithms in Search, Optimization and Machine Learn,” Addison Wesley, Reading, MA..