臺灣博碩士論文加值系統

本文主旨是探討以多層最佳化設計方法配合靈敏度分析, 求解幾何非線性
結構的最佳化設計問題, 並將最佳化設計過程中, 佔程式執行時間最長的
兩大部份: 非線性有線元素分析程式與最佳化分析程式, 撰寫成適合向量
平行運算之電腦程式, 並討論此程式於向量平行電腦CONVEX C3840 上執
行的情形。在非線性結構分析方面, 以切線勁度法配合Jacobian共軛梯度
法,取代傳統高斯消去法, 以利向量平行處理。在程式執行結果方面, 試
著比較線性與幾何非線性結構最佳化設計結果的差異。且設定一個較好的
切線勁度法及Jacobian共軛梯度法收斂誤差值, 以利最佳化設計時運用;
再根據CONVEX C3840在向量平行處理運算上的特色, 分析電腦程式執行的
結果, 並提出其在運用上的可行性 , 及未來研究方向。

The purpose of this study is to investigate the optimum design
of geometrically nonlinear structure by means of multilevel
optimization method with sensitivity analysis. In this study
the tangent stiffness method is used to solve the nonlinear
problem. The Jacobian conjugate gradient, an iterative solution
algorithm, is employed to promote the performance of finite
element analysis for structure. The proposed solution
procedures are programmed in FORTRAN for implementation on
vector-parallel supercomputer CONVEX C3840. The difference of
weight optimum design between linear elasticity and considering
the geometrical nonlinearity behavior from large displacements
are discussed. The suitable tolerances for tangent stiffness
method and Jacobian conjugate method are proposed. And the
results of examples are presented. Further discussion and
suggestions for the results are also stated.