Adaptive Multi-fidelity Surrogate Modelling for High-quality Shape Optimization

Jeroen Wackers (1) , Hayriye Pehlivan Solak (1) , Michel Visonneau (1) , Riccardo Pellegrini (2) , Andrea Serani (2) , Matteo Diez (2)

1 LHEEA - Laboratoire de recherche en Hydrodynamique, Énergétique et Environnement Atmosphérique
2 CNR-INM - Institute of Marine Engineering - CNR

A multi-fidelity surrogate modelling approach for shape optimization, which relies on adaptive techniques to obtain good performance for a large range of problems, is presented and critically evaluated. Furthermore, an approach to adaptive selection of the fidelity levels to be used is presented. Adaptation is shown to be effective for solving complex problems. Finally, potential improvements in the noise canceling, the uncertainty estimation, and the adaptive sampling are identified.

Mots clés: Workshop; Multi-fidelity surrogate; Active learning; Computational fluid dynamics; Adaptive grid refinement

Domaines: Sciences de l'ingénieur [physics]

Jeroen Wackers, Hayriye Pehlivan Solak, Michel Visonneau, Riccardo Pellegrini, Andrea Serani, et al.. Adaptive Multi-fidelity Surrogate Modelling for High-quality Shape Optimization. Research workshop AVT-354 on Multifidelity methods for military vehicle design, Sep 2022, Varna, Bulgaria. ⟨hal-03815782⟩