BlendedNet: A Blended Wing Body Aircraft Dataset and Surrogate Model for Aerodynamic Predictions

BlendedNet is a publicly available aerodynamic dataset developed by the MIT DeCoDE Lab and MIT Lincoln Laboratory. It contains 999 unique blended wing body (BWB) aircraft geometries, each simulated under approximately nine distinct aerodynamic cases, resulting in 8,830 successfully converged high-fidelity CFD simulations. The geometries were systematically generated by sampling across geometric design parameters and flight conditions, and analyzed using Reynolds-Averaged Navier–Stokes (RANS) simulations with the Spalart–Allmaras turbulence model, employing 9–14 million volume cells per case. BlendedNet addresses the scarcity of high-fidelity aerodynamic data for BWB aircraft, enabling future research in data-driven aerodynamic analysis, performance prediction, and design optimization.