A Multi-Fidelity Double-Delta Wing Dataset and Empirical Scaling Laws for GNN-based Aerodynamic Field Surrogates

Data-driven surrogate models are increasingly adopted to accelerate vehicle design, but open-source multi-fidelity datasets and empirical guidelines linking dataset size to model performance remain limited. This study investigates the relationship between training data size and prediction accuracy for a graph neural network (GNN)-based surrogate model for aerodynamic field prediction. We release an open-source, multi-fidelity aerodynamic dataset for double-delta wings spanning 272 geometries and 2448 flow snapshots evaluated with both vortex lattice method (VLM) and Reynolds-averaged Navier–Stokes (RANS) solvers. Using the dataset, we conduct a preliminary empirical scaling study of the MF-VortexNet surrogate and find that test error decreases with dataset size according to a power-law relation, providing guidance for future data collection and model design in multi-fidelity aerodynamic surrogate modeling.

Paper PDF: https://yiren6.github.io/files/Scitech_2026_VortexNetScaling.pdf

Recommended citation: Shen, Y.; Alonso, J. (2026). "A Multi-Fidelity Double-Delta Wing Dataset and Empirical Scaling Laws for GNN-based Aerodynamic Field Surrogates." AIAA SCITECH 2026.
Access from Publisher