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

TL;DR

This paper introduces BlendedNet, a public high-fidelity dataset of 999 blended-wing-body geometries with ~8830 converged RANS cases using the Spalart–Allmaras model, plus a two-stage surrogate that predicts surface aerodynamic coefficients from surface point clouds. A permutation-invariant PointNet regressor first infers nine geometric design parameters from sampled surface points, then a FiLM-conditioned network, guided by those parameters and flight conditions, outputs pointwise Cp, Cfx, and Cfz across the surface. The authors demonstrate strong predictive performance, with high R^2 scores for geometry parameters and low MSE/MAE for surface coefficients, and they achieve accurate integrated lift and drag when aggregating the predictions. BlendedNet aims to address data scarcity for unconventional configurations and to accelerate data-driven surrogate modeling for aerodynamic design, enabling rapid exploration of BWB geometries and operating conditions.

Abstract

BlendedNet is a publicly available aerodynamic dataset of 999 blended wing body (BWB) geometries. Each geometry is simulated across about nine flight conditions, yielding 8830 converged RANS cases with the Spalart-Allmaras model and 9 to 14 million cells per case. The dataset is generated by sampling geometric design parameters and flight conditions, and includes detailed pointwise surface quantities needed to study lift and drag. We also introduce an end-to-end surrogate framework for pointwise aerodynamic prediction. The pipeline first uses a permutation-invariant PointNet regressor to predict geometric parameters from sampled surface point clouds, then conditions a Feature-wise Linear Modulation (FiLM) network on the predicted parameters and flight conditions to predict pointwise coefficients Cp, Cfx, and Cfz. Experiments show low errors in surface predictions across diverse BWBs. BlendedNet addresses data scarcity for unconventional configurations and enables research on data-driven surrogate modeling for aerodynamic design.

Figures (10)

• Figure 1: Illustration of the blended wing body (BWB) planform parameterization, showing key geometric design parameters used in dataset generation.
• Figure 2: View of an example BWB surface mesh.
• Figure 3: Centerline slice of mesh with zoomed view of boundary layer cells.
• Figure 4: Scatter plots of aerodynamic coefficients: (Left) Lift coefficient ($C_L$) vs Drag coefficient ($C_D$), (Center) Pitching moment coefficient ($C_M$) vs Lift coefficient ($C_L$), and (Right) Pitching moment coefficient ($C_M$) vs angle of attack ($\alpha$). These relationships provide insights into aerodynamic performance and longitudinal stability characteristics.
• Figure 5: Visualization of aerodynamic coefficients for different cases. Each row represents a specific design and aerodynamic flight condition respectively: highest and lowest lift, highest drag, and highest and lowest lift-to-drag ratio. The columns correspond to the pressure coefficient ($C_p$) and the skin friction coefficients in the $x$, $y$, and $z$ directions ($C_{f_x}$, $C_{f_y}$, and $C_{f_z}$). It is important to emphasize that both the BWB geometry and the flight conditions must be taken into account, as their interaction jointly influences the overall aerodynamic performance.