Inverse Design — 02

Airfoil Optimizer

NeuralFoil-powered shape optimization. Specify your operating points (Re, CL, weight) and constraints — get a CST airfoil that minimizes drag across the envelope. Solved with CasADi/IPOPT via AeroSandbox; analytic gradients keep typical runs under 5 seconds.

⚠ Read this before trusting an optimized airfoil

Airfoil optimization is easy to get wrong. Single-point drag minimization can produce pathological shapes — knife-edge leading edges, ultra-cusped trailing edges — that beat a baseline at exactly one CL and fall off a cliff everywhere else.

This tool defaults to multi-point optimization across your CL range for this reason. Always review the shape, polar curve, and constraint slack before using the output. Validate in a higher-fidelity solver (XFOIL, RANS) for anything that flies or carries load.

See Peter Sharpe's AeroSandbox notes and Mark Drela's Pros & Cons of Airfoil Optimization for context.

Starting Airfoil

Operating Points

weighted multi-pointEach row is an aerodynamic condition the optimized airfoil must perform well at. The optimizer minimizes the weighted-mean drag across all rows, with each (Re, CL) treated as a hard constraint via per-point alpha decision variables.
#SpeedReCLWeight
1
Σweights = 1.0
Re ∈ [5.0e+5, 5.0e+5]
CL ∈ [0.60, 0.60]
Advanced (Ncrit, Mach, iterations)

Constraints

% chord
% chord
deg
(recommended)
IPOPT · NeuralFoil largeIPOPT (Interior Point OPTimizer) is the constrained nonlinear solver driving the shape updates. Gradients flow via CasADi automatic differentiation through KulfanAirfoil + NeuralFoil — no finite differences. NeuralFoil is the surrogate model evaluating drag for each candidate shape; bigger models are more accurate but slower per call.

Ready when you are

Pick a starting airfoil, add your operating points (Re, CL, weight), then run. Typical optimization completes in 2–6 seconds with the large NeuralFoil model.

Airfoil Optimizer — FoilTools | Foil.tools