AErOmAt — Surrogate-Assisted Aerodynamic Optimisation
2016–2019 Surrogate Models Quality Diversity CFD BMBF Funded
AErOmAt — Aerodynamic Energy Optimisation through Metamodel-Assisted Adaptation of Structures
Computational aerodynamic optimisation is limited by the high cost of CFD simulation. AErOmAt developed new methods combining surrogate models with quality diversity algorithms (MAP-Elites/illumination) to dramatically reduce simulation costs while exploring diverse, high-performing designs.
Key contributions
- Surrogate-assisted phenotypic niching, (SPHEN) — combining Gaussian process surrogates with illumination algorithms for data-efficient exploration of aerodynamic design spaces and their morphological and flow features
- Hierarchical surrogate modelling — multi-level surrogate models for further reducing computational cost
Partners
Prof. Dirk Reith (lead, H-BRS / TREE), Prof. Alexander Asteroth, Fraunhofer SCAI, DLR, University of Siegen. Funded by BMBF.