Skip to Main content Skip to Navigation
Theses

GBNM : un algorithme d'optimisation par recherche directe. Application à la conception de monopalmes de nage

Abstract : In this work a fixed cost essentially local optimization method, which becomes global if the number of evaluations increases is developed. " Globalization " is achieved by probabilistic restarts based on past searches. An improved Nelder-Mead method is the local optimizer. Improvements concern the variables which are bounded, the nonlinear inequality constraints which are taken into account by adaptive penalization, and the failures by simplex degeneration which are detected and handled through restart. The resulting method is called " Globalized and Bounded Nelder-Mead (GBNM) ". Numerical experiments on multimodal test functions and composite laminated design problems illustrate the performance of the method as compared to an evolutionary optimizer. More complex applications are dealt with GBNM: the flexural stiffness optimization and the identification of the ply drop positions of swimming monofins.
Document type :
Theses
Complete list of metadatas

Cited literature [88 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00850658
Contributor : Florent Breuil <>
Submitted on : Wednesday, August 7, 2013 - 4:41:37 PM
Last modification on : Wednesday, June 24, 2020 - 4:18:32 PM
Long-term archiving on: : Friday, November 8, 2013 - 4:24:10 AM

Identifiers

  • HAL Id : tel-00850658, version 1

Citation

Marco Antonio Luersen. GBNM : un algorithme d'optimisation par recherche directe. Application à la conception de monopalmes de nage. Modélisation et simulation. INSA de Rouen, 1993. Français. ⟨NNT : 2004ISAM0014⟩. ⟨tel-00850658⟩

Share

Metrics

Record views

616

Files downloads

2085