BlendedNet++: A Large-Scale Blended Wing Body Aerodynamics Dataset and Benchmark

TL;DR

BlendedNet++ addresses the scarcity of field-resolved, large-scale BWB aerodynamic data by releasing 12,490 geometries with dense surface fields and standardized geometry-disjoint splits. It provides a six-model forward-surrogate benchmark and a diffusion-based inverse-design framework, enabling fair, reproducible comparisons across architectures and optimization strategies. Empirical results show FiLMNet as the leading forward surrogate, while conditional diffusion offers fast, diverse inverse designs with strong CFD-consistency validated by high-correlation metrics. The work lays a scalable, community-friendly foundation for advancing field-level aerodynamics and inverse design in large design spaces.

Abstract

Despite progress in machine learning-based aerodynamic surrogates, the scarcity of large, field-resolved datasets limits progress on accurate pointwise prediction and reproducible inverse design for aircraft. We introduce BlendedNet++, a large-scale aerodynamic dataset and benchmark focused on blended wing body (BWB) aircraft. The dataset contains over 12,000 unique geometries, each simulated at a single flight condition, yielding 12,490 aerodynamic results for steady RANS CFD. For every case, we provide (i) integrated force/moment coefficients CL, CD, CM and (ii) dense surface fields of pressure and skin friction coefficients Cp and (Cfx, Cfy, Cfz). Using this dataset, we standardize a forward-surrogate benchmark to predict pointwise fields across six model families: GraphSAGE, GraphUNet, PointNet, a coordinate Transformer (Transolver-style), a FiLMNet (coordinate MLP with feature-wise modulation), and a Graph Neural Operator Transformer (GNOT). Finally, we present an inverse design task of achieving a specified lift-to-drag ratio under fixed flight conditions, implemented via a conditional diffusion model. To assess performance, we benchmark this approach against gradient-based optimization on the same surrogate and a diffusion-optimization hybrid that first samples with the conditional diffusion model and then further optimizes the designs. BlendedNet++ provides a unified forward and inverse protocol with multi-model baselines, enabling fair, reproducible comparison across architectures and optimization paradigms. We expect BlendedNet++ to catalyze reproducible research in field-level aerodynamics and inverse design; resources (dataset, splits, baselines, and scripts) will be released upon acceptance.

Figures (12)

• Figure 1: BlendedNet++ overview: (left) geometry parameterization and meshing, (middle) surrogate benchmarks for field-level prediction (FiLMNet, Transolver, PointNet, GraphSAGE, GraphUNet, GNOT), and (right) inverse-design pipeline (conditional diffusion, gradient-based optimization) over the 9-D planform box. The dataset couples integrated coefficients $(C_L,C_D,C_M)$ with dense surface fields $C_p$ and $C_f$ for 12.49 unique BWB geometries.
• Figure 2: Parameterization of the BWB planform.
• Figure 3: Representative renderings from BlendedNet++, illustrating the shape diversity across the dataset. The variations in fuselage geometry, wing span, and overall configuration highlight the richness of aerodynamic design space captured in the dataset.
• Figure 4: Example from BlendedNet++, showing the 3D shape alongside surface flow quantities. From left to right: baseline geometry, surface pressure coefficient ($C_p$), and skin-friction coefficients in the streamwise ($C_{fx}$), and vertical ($C_{fz}$) directions.
• Figure 5: Example visualization of an automatically generated computational fluid dynamics mesh.