SMART: Scalable Mesh-free Aerodynamic Simulations from Raw Geometries using a Transformer-based Surrogate Model
Hagnberger and Niepert
What it does
Predicts physical quantities at any query point from a point cloud of the geometry and the simulation parameters, with no simulation mesh as input; a physics decoder attends to the encoder's intermediate latents, so geometry features and the field are updated together.
Why it matters
Meshing a new design is costly, and mesh-free surrogates usually lose accuracy; SMART is competitive with, and often better than, models that take the mesh. Luminary Cloud has built its own implementation as the base of its large physics models.