Loading...

Comparison of the Different Classical Structural Optimization Methods and Algorithms

Moghani, Mazdak | 2009

487 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 39519 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Joghataei, Abdolreza
  7. Abstract:
  8. In this thesis, the classical methods of structural optimization were studied and compared. After a comprehensive study on different methods of structural optimization, a reduced number of methods were selected in order to be used in this research. Exterior penalty function method, linear and quadratic extended penalty function methods were the selected constrained optimization methods. For unconstrained optimization, univariate method, Powell method, steepest descent method, Fletcher-Reeves method, and quasi-Newton methods of DFP and BFGS were used. And from different one-dimensional search methods studied, we used the golden section method. The structures which were optimally designed by above methods were different kinds of 2D and 3D trusses. Analyzing of these trusses was done by stiffness method. Also sensitivity analysis was used for calculating the derivatives of behavior variables (stress in truss members and displacement of nodes) subject to primary variables (cross section areas of truss members). Computer programs were developed on the basis of selected methods, using Visual C++ language. From results gathered by several runs of these programs, using different set of methods on different trusses, we concluded that, among constrained optimization methods, the exterior penalty function method was the quickest and the most reliable in converging to the desired optimum design. Moreover, the quasi-Newton methods of DFP and BFGS were the quickest of unconstrained methods, although zero order methods like Powell method showed more reliability in converging to the minimum.
  9. Keywords:
  10. Optimization ; Structures ; Trusses ; Algorithms ;

 Digital Object List

  • محتواي پايان نامه
  •   view

 Bookmark

No TOC