BlendedNet++: A dataset and benchmark for field-resolved aerodynamics and inverse design of blended wing body aircraft

In three linesBlendedNet++ is a dataset of 12,492 Blended Wing Body (BWB) aircraft geometries with RANS simulations for aerodynamic field prediction. Authors benchmark 5 deep learning architectures (Transolver best) and propose a generative inverse design pipeline using conditional diffusion models, validated by CFD with R² > 0.99.

Summary generated by Claude — human-verified