ETHZ-EDAC/aeroelastic-wings

Aeroelastic Wings

531,464 simulated wing designs, 26 columns, 4.65 GB, for prediction and inverse design. The accompanying setup specifies a 15 m Aluminum 7075 wing, 1 m root chord, and 50 m/s freestream. The inspected solver couples vortex-lattice aerodynamics with bending and torsion.

Quick start

pip install huggingface_hub pandas
hf download ETHZ-EDAC/aeroelastic-wings --repo-type dataset --local-dir data/processed/wings

Run hf auth login first if authentication is required.

import pandas as pd

for batch in pd.read_csv(
    "data/processed/wings/combined_wings_data_clipped.csv",
    usecols=["CMa", "max_von_mises_stress", "mass", "CL", "CD"],
    chunksize=10_000,
):
    batch["CL/CD"] = batch["CL"] / batch["CD"]
    print(batch.head())
    break

Columns

Column	Unit / representation	Meaning
chord_ctrl_pos	Dimensionless	Chord Bezier control-point position along normalized half-span.
chord_weight_ctrl	Dimensionless	Chord control-point weight.
chord_root	m	Root chord; always 1.0.
chord_tip	m	Tip chord.
twist_root	degrees	Root geometric twist.
twist_tip	degrees	Tip geometric twist.
twist_mean_pos	Dimensionless	Normalized half-span position where twist reaches its root-tip mean.
spar_sweep	degrees	Spar sweep angle.
max_jacobian_eigenvalue	Solver coordinates	Largest eigenvalue real part of reduced tangent stiffness.
min_jacobian_eigenvalue	Solver coordinates	Smallest eigenvalue real part of reduced tangent stiffness.
is_statically_stable	Boolean	Structural stability: minimum eigenvalue exceeds 1e-6.
max_bending_moment	N·m	Signed peak bending moment.
max_torsion	N·m	Signed peak torque.
max_von_mises_stress	Pa	Maximum estimated von Mises stress.
L	N	Lift.
D	N	Model-reported drag.
CL	Dimensionless	Lift coefficient.
CD	Dimensionless	Drag coefficient.
CM	Dimensionless	Pitching-moment coefficient.
CMa	1/rad	Pitching-moment derivative with respect to angle of attack.
CLa	1/rad; legacy label	Exported rolling-moment derivative, not lift-curve slope.
max_heave_displacement	m	Signed peak heave displacement.
max_twist_displacement	degrees	Signed peak elastic twist.
mass	kg	Full-wing structural mass.
aeroelastic_wing	JSON string	ID, geometry, airfoil coefficients, displacements, performance, mechanical state.
airfoil_familys	Five-integer JSON array	Spanwise family selections; categorical IDs 0–13.

Signed peaks retain the sign at greatest absolute magnitude. Compute CL/CD as CL / CD; parse serialized fields with json.loads. Eigenvalues are not frequencies; structural stability is distinct from longitudinal stability. The exporter assigns aero["Cla"] to CLa; AeroSandbox distinguishes this rolling derivative from lift derivative CLa.

Files and integrity

Files include the original CSV, generation config.yaml, statistics in metadata/dataset_profile.json, and the figure in metadata/property_distributions.png. SHA256SUMS verifies the original CSV and configuration:

cd data/processed/wings
sha256sum -c SHA256SUMS

Coverage and limitations

All numeric values are finite, geometry JSON parses, and geometry IDs are unique. All stability flags are true, but stresses reach 543.837 MPa, exceeding the configured 430 MPa yield strength. These checks do not independently validate physics or exclude similar geometries.

No official evaluation splits are supplied. Record split IDs, dataset revision, and checksum; fit preprocessing on training data only.

License and citation

MIT, as declared in repository metadata. Cite ETHZ-EDAC/aeroelastic-wings with the exact revision and CSV checksum.