http://swrc.ontoware.org/ontology#InProceedings
Acceleration of GA in aerodynamics by search space reduction and artificial neural
en
genetic algorithm
artificial neural network
aerodynamics
aerodynamic optimization
aircraft design
design optimization
CPU
search space reduction
Rousseau Yannick
中村 佳朗
Rousseau Yannick
Nakamura Yoshiaki
宇宙航空研究開発機構
Japan Aerospace Exploration Agency (JAXA)
JAXA Special Publication: Proceedings of Aerospace Numerical Simulation Symposium 2004
JAXA-SP-04-012
155-159
2005-03-25
Genetic algorithms (GAs) have been successfully applied to numerical optimization problems, unfortunately GAs still remain computationally expensive, and the high computational cost make the use thereof impractical in most Aerodynamic optimization problems. Computational cost reduction is undoubtedly thought to be a common problem of most Aerodynamic shape optimization. In this approach, Artificial Neural Network (ANN) is used for function approximation; a number of mathematical computations are performed on the approximated function to obtain a reduced model. This resulting mathematical model is used to locate the variables that affect most the cost function. This approach is first tested on analytical functions, then numerical experimentations are conducted to solve shape optimization problem for the design of a wing profile. For each evaluation required by the optimizer, the Navier-Stokes equations with the Baldwin-Lomax turbulence model are solved.
1349-113X
AA11984031
資料番号: AA0048469026
レポート番号: JAXA-SP-04-012