Aerodynamic Quantities for 8,996 Airfoil Shapes at Two Angles of Attack

Aerodynamic flow field and summary coefficients for 8,996 airfoil shapes, computed using the HAM2D CFD model. The dataset provides simulations at two angles of attack (4 and 12 degrees) for a benchmark in AI/ML tool testing. It was generated in February 2023 using the Bridges-2 supercomputing system.

Use Cases

• Train surrogate models to predict coefficients of lift and drag from airfoil shape parameters for rapid aerodynamic design.
• Analyze the relationship between flow field values like momentum, energy, and vorticity and the resulting aerodynamic coefficients.
• Benchmark generative AI models for creating new airfoil shapes by learning from the four-dimensional parameter space of higher-order perturbations.
• Investigate the aerodynamic impact of non-linear shape deformations while holding attributes like thickness approximately constant.

Strengths

• Contains 8,996 distinct airfoil shapes, providing a substantial sample for data-driven aerodynamic analysis.
• Each shape is simulated at two specific angles of attack (4 and 12 degrees) under controlled conditions (Mach 0.1, Reynolds 9M).
• Derived from a parameterization learned from the G2Aero database of over 19,000 airfoils, capturing a wide array of shape characteristics.
• Simulations use a detailed CFD setup including a 400-point surface grid and specific turbulence/transition models, ensuring technical rigor.

Limitations

• Limited to only two angles of attack, restricting analysis of performance across a full operational envelope.
• Simulations are based on a specific CFD model (RANS with Spalart-Allmaras turbulence model), which may not capture all flow physics accurately.
• The dataset focuses on isolated, non-linear shape perturbations, so it does not represent complete, arbitrary airfoil designs.

Provenance

Source

Department of Energy (INTEGRATE and Foundational AI for Wind Energy projects).

Collection Method

Computational Fluid Dynamics (CFD) simulations using the HAM2D code on the Bridges-2 supercomputing system.

Time Range

Simulations executed in February 2023.

Freshness

Simulations performed in February 2023; data reformatted and published in July 2023.

Geography

null