Multi-fidelity efficient global optimization : Methodology and application to airfoil shape design

Abstract : Predictions and design engineering decisions can be made using a variety of informa- tion sources that range from experimental data to computer models. These information sources could consist of different mathematical formulations, different grid resolutions, dif- ferent physics, or different modeling assumptions that simplify the problem. This leads to information sources with varying degrees of fidelity, each with an associated accuracy and querying cost. In this paper, we propose a novel and flexible way to use multi-fidelity informa- tion sources optimally in the context of airfoil shape optimization using both Xfoil and ADflow. The new developments are based on Bayesian optimization and kriging metamodeling and allow the aerodynamic optimization to be sped up. In a constrained optimization example with 15-design variables problem, the proposed approach reduces the total cost by a factor of two compared to a single Bayesian based fidelity optimization and by a factor of 1.5 compared to sequential quadratic programming.
Complete list of metadatas

Cited literature [27 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-02233293
Contributor : Open Archive Toulouse Archive Ouverte (oatao) <>
Submitted on : Thursday, August 1, 2019 - 10:32:41 AM
Last modification on : Wednesday, August 7, 2019 - 2:34:15 PM

File

Meliani_24161.pdf
Files produced by the author(s)

Identifiers

Citation

Mostafa Meliani, Nathalie Bartoli, Thierry Lefebvre, Mohamed-Amine Bouhlel, Joaquim Martins, et al.. Multi-fidelity efficient global optimization : Methodology and application to airfoil shape design. AIAA Aviation 2019 Forum, Jun 2019, Dallas, United States. pp.1-18, ⟨10.2514/6.2019-3236⟩. ⟨hal-02233293⟩

Share

Metrics

Record views

48

Files downloads

44