Airfoil

Airfoil#

_static/img/problems/airfoil.png

Version

0

Design space

Dict('angle_of_attack': Box(0.0, 10.0, (1,), float32), 'coords': Box(0.0, 1.0, (2, 192), float32))

Objectives

cd: ↓

Conditions

mach: 0.8

reynolds: 1000000.0

area_initial: None

area_ratio_min: 0.7

cl_target: 0.5

Dataset

IDEALLab/airfoil_v0

Container

mdolab/public:u22-gcc-ompi-stable

Import

from engibench.problems.airfoil.v0 import Airfoil

Airfoil 2D shape optimization problem.

Note

This problem requires gcc and gfortran to be installed. See the simulator section for more details.

Problem Description#

This problem simulates the performance of an airfoil in a 2D environment. An airfoil is represented by a set of 192 points that define its shape. The performance is evaluated by the MACH-Aero simulator that computes the lift and drag coefficients of the airfoil.

Design space#

The design space is represented by a dictionary where one element (coords) is a numpy array (vector of 192 x,y coordinates in [0., 1.) per design) that define the airfoil shape, and the other element (angle_of_attack) is a scalar.

Objectives#

The objectives are defined and indexed as follows:

  1. cd: Drag coefficient to minimize.

Conditions#

The conditions are defined by the following parameters:

  • mach: Mach number.

  • reynolds: Reynolds number.

  • area_ratio_min: Minimum area ratio (ratio relative to initial area) constraint.

  • area_initial: Initial area.

  • cl_target: Target lift coefficient to satisfy equality constraint.

Simulator#

The simulator is a docker container with the MACH-Aero software that computes the lift and drag coefficients of the airfoil. You can install gcc and gfortran on your system with your package manager.

  • On Ubuntu: sudo apt-get install gcc gfortran

  • On MacOS: brew install gcc gfortran

  • On Windows (WSL): sudo apt install build-essential

Dataset#

The dataset linked to this problem is hosted on the Hugging Face Datasets Hub.

v0#

Fields#

The dataset contains optimal design, conditions, objectives and these additional fields:

  • initial_design: Design before the adjoint optimization.

  • cl_con_violation: # Constraint violation for coefficient of lift.

  • area_ratio: # Area ratio for given design.

Creation Method#

Refer to paper in references for details on how the dataset was created.

References#

If you use this problem in your research, please cite the following paper: C. Diniz and M. Fuge, “Optimizing Diffusion to Diffuse Optimal Designs,” in AIAA SCITECH 2024 Forum, 2024. doi: 10.2514/6.2024-2013.

Lead#

Cashen Diniz @cashend

Contents