Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/2239
標題: 使用全域靈敏度公式與近似分析方法的多重領域最佳化
Multidisciplinary Design Optimization Using Global Sensitivity Equation and Approximate Analyses
作者: 蔡起源
Tsai, Chi-Yuan
關鍵字: Multidisciplinary Design Optimization;多重領域最佳化;Global Sensitivity Equations;Response Surface;Moving Least Square;Artificial Neural Networks;全域靈敏度公式;回應表面法;移動最小平方法;類神經網路
出版社: 機械工程學系
摘要: 
工程最佳化有時包含著數個領域,而各領域的反應值也可能是其他領域的一個輸入因子。傳統的作法是在個別的領域中對其設計變數求取最佳解,對於各個領域的相互影響並未加以考慮,因此對於高度耦合的系統傳統的作法未必能產生真正的最佳解。
本文主要以全域靈敏度公式(GSE)來求取個別領域的設計變數對於其他領域反應的靈敏度,亦即是求取系統反應對設計變數的全微分,然後結合各領域形成單一的最佳化問題。為形成全域靈敏度公式以,本文近似法如移動最小平方法(MLS)或類神經網路取代有限元素分析或實驗,建立近似的目標函數和系統反應函數,再利用這些近似函數配合全域靈敏度公式求取一耦合系統的全域靈敏度然後以SLP法求解。
本文以數個例子分別比較使用傳統方式(忽略耦合關係)求取最佳解與使用全域靈敏度公式求取最佳解,結果發現全域靈敏度公式需要近似函數有極高的準確性,並且對於耦合系統有精確解才能得到可靠的最佳解。

The multidisciplinary design is usually found in many engineering applications. The input of one discipline may be from an output of another discipline in the system. Traditionally the optimum solutions are found without considering the mutual influence of disciplines. Therefore the solutions obtained may not be reliable.
In this paper Global Sensitivity Equations (GSE) are used to get sensitivities of responses in one discipline with response to design variables in other disciplines. In other words the total differentials of respect system response are found by GSE. After obtaining the global sensitivities, a single optimization problem can be created by combining all disciplines together and further solved by the sequential linear programming method. In order to form GSE some approximation approaches are used. The finite element analysis and the mechanism analysis are replaced by approximate methods such as moving least square, response surface and artificial neural networks.
Optimum solutions obtained by uncoupled method are compared with optimum solutions obtain by GSE in this paper. The result shows that credible solutions obtained by GSE for coupled system heavily depend on the accuracy of those approximate functions.
URI: http://hdl.handle.net/11455/2239
Appears in Collections:機械工程學系所

Show full item record
 
TAIR Related Article

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.