<ul><li>A new surrogate modeling framework for aerodynamic applications called MARIO (Modulated Aerodynamic Resolution Invariant Operator) has been introduced based on Neural Fields.</li><li>MARIO addresses non-parametric geometric variability through efficient shape encoding and utilizes the discretization-invariant nature of Neural Fields, enabling training on significantly downsampled meshes while maintaining accuracy during full-resolution inference.</li><li>The framework has been validated on two datasets, including the AirfRANS dataset for a two-dimensional airfoil benchmark and the NASA Common Research Model for three-dimensional pressure distributions on a full aircraft surface mesh, confirming its accuracy and scalability.</li><li>Benchmarking shows that Neural Field surrogates like MARIO can offer rapid and precise aerodynamic predictions while reducing computational cost and memory requirements compared to traditional CFD solvers and existing surrogate methods.</li></ul>

Towards scalable surrogate models based on Neural Fields for large scale aerodynamic simulations

Discover more