COCOANet

A high-fidelity CFD dataset of parametric (CCA) Aircraft geometries and flight condition for aerodynamic surrogate modelling and Aerodynamic Shape Optimization, created as part of ShapeBench.

Dataset Summary

Geometries	401
CFD runs	3,570
Design parameters	16 geometric
Flight parameters	3 (angle of attack, velocity, altitude)
CAD Kernal	nTop
Solver	Flow360
Sampling	Latin Hypercube Sampling (LHS), seed 42

Files

File	Description
dataset_samples.csv	3,570 rows — 16 design params + flight conditions + corrected CL, CD, CFx, CFy, CFz, CMx, CMy, CMz
sampling_metadata.json	LHS seed, library, and sample counts
COCOANet_croissant.json	Croissant 1.0 metadata
cca_drone_Transolver.pt	Pretrained Transolver surrogate model checkpoint
norm_stats.pt	Normalisation statistics for the surrogate model
case_*/	Full CFD data per geometry: STL mesh (nTop), Flow360 solution fields, surface/volume data

Design Parameters

Dihedral angle, Max Wing Blend, Inlet Angle 1 & 2, Wing Position, Rear Point X, Inlet Location, NACA 4-digit code, Fore/Aft Top Angle, Top/Bottom Height Aft, Wing span, Rear tail offset, Root Chord, Tail Root chord.

Outputs

Corrected force and moment coefficients: CL, CD, CFx, CFy, CFz, CMx, CMy, CMz.

Notes

Some cases excluded due to CFD divergence or HPC node restarts. Full raw CFD data (3.65 TB).