#1st place (MMGP variant) : Optimal Morphing Strategies for Efficient Computations
Abbas Kabalan
2024
in
Competition: NeurIPS 2024 ML4CFD Competition: Harnessing Machine Learning for Computational Fluid Dynamics in Airfoil Design
in
Competition: NeurIPS 2024 ML4CFD Competition: Harnessing Machine Learning for Computational Fluid Dynamics in Airfoil Design
Abstract
In this presentation, we introduce an extension to the MMGP method, focusing on enhancing the efficiency of morphing computations. Specifically, we outline several strategies for more effective morphing calculations, performed within the reference domain. This approach facilitates the construction of a reduced-order basis for morphings, which can be leveraged during the inference stage. Additionally, we propose an alternative architecture for the GP model, highlighting the method's flexibility. Finally, we provide insights into computing optimal morphings that minimize the number of modes after applying PCA to the fields.
Video
Chat is not available.
Successful Page Load