Curated Airfoil Shapes with 19,164 Landmarked Profiles

19,164 curated airfoil shapes, each defined by 1,001 shape landmarks. The dataset was constructed by perturbing CST coefficients of 13 baseline wind turbine airfoils and augmenting with filtered shapes from the BigFoil database. It includes a derived parameter space from principal geodesic analysis of separable shape tensors for machine learning applications.

Use Cases

• Train generative models in the provided PGA space to synthesize novel airfoil shapes for aerodynamic optimization.
• Apply regression models to predict aerodynamic performance coefficients using the 1,001 landmark coordinates as input features.
• Use the 18 CST coefficients from the baseline airfoil perturbation process as a compact parameterization for shape optimization studies.
• Cluster the 19,164 airfoils based on their landmark coordinates to identify dominant shape families and design trends.

Strengths

• Contains 19,164 distinct airfoil shapes, providing a substantial collection for analysis.
• Each airfoil is consistently represented by 1,001 high-resolution shape landmarks for precise geometric modeling.
• Includes a derived data-driven parameter space (PGA space) built using principal geodesic analysis on separable shape tensors.
• Underwent thorough preprocessing, including filtering for data quality and regularization via 8-order CST parametrization.

Limitations

• The dataset is a static collection with no associated aerodynamic performance data (e.g., lift, drag coefficients).
• Contains a bias towards wind energy applications, as 13,000 shapes are derived from NREL and IEA reference wind turbine airfoils.
• The geometric representation is fixed to 1,001 landmarks per airfoil, which may not be the optimal resolution for all applications.

Provenance

Source

Department of Energy, constructed from baseline wind turbine airfoils and the consolidated BigFoil database.

Collection Method

Generated by perturbing CST coefficients of baseline airfoils and applying rigorous filtering and regularization to shapes from source databases.

Freshness

Last updated in October 2024.