Three wind tunnels in one browser tab

From RC trainers to hypersonic noses.

Drop in your CAD, put it in the tunnel that matches how it flies, sweep its whole flight envelope — then let AeroSculpt reshape it inside the limits you set: protected parts, payload volumes, propellers and material temperatures.

Start free How it works inside
RC2–60 m/s · spinning props Generalto Mach 1.2 · 3D CFD SupersonicMach 1.2–12 · heat & materials
≈ 3 s
per 3D CFD run on a laptop GPU — a whole sweep in seconds, 12× the CPU
2 m/s → M12
park flyers, UAVs and rockets up to hypersonic
≤ 8%
propeller thrust and power vs UIUC wind-tunnel data
≤ 5%
hypersonic sphere drag vs free-flight data, Mach 4–10
Pick your tunnel

The right physics for every speed

No single method is trustworthy from a park flyer to Mach 8, so AeroSculpt switches physics with the speed range — and labels every number with the method that produced it.

RC tunnel2–60 m/s

Propellers turning

Put each propeller where it will really be — then drag, nudge or snap it onto the centre line. Choose the blade family, size, pitch, rpm and 2–6 blades; each spins in the CFD as an actuator disk with its own thrust and swirl.

  • Two measured blade families: slim thin-electric and wide constant-chord
  • Live thrust, power and slipstream for each option before you run
  • Airframe drag measured inside each propeller's slipstream
  • Thrust against drag, and the estimated level top speed
3D CFD + blade-element propellersOpen RC tunnel
General tunnelto Mach 1.2

3D CFD to transonic

The lattice-Boltzmann large-eddy wind tunnel: rockets, UAV fuselages, pods and fairings across a range of airspeeds and angles of attack, compressibility-corrected up to the critical Mach number and estimated through the transonic drag rise.

  • Drag polar, drag and power across the whole envelope
  • Surface pressure, streamlines and flow slices at every angle
  • Protect fins and mounts; keep payload bays and motor tubes inside
  • Bayesian shape optimization of nose, tail, section and length
Lattice-Boltzmann LES + Kármán–TsienOpen General tunnel
Supersonic tunnelMach 1.2–12

Shock, heat and materials

Wave, base and friction drag from Mach 1.2 into the hypersonic range, with aerodynamic heating and the wall temperature your skin reaches — checked against each material's limit.

  • Stagnation and surface heating, radiative-equilibrium wall temperature
  • Material library from fibreglass to carbon-carbon, or define your own
  • Material regions for nose caps and leading edges
  • Optimize for drag or peak temperature within material limits
Shock-expansion surface methodsOpen Supersonic tunnel
Side by side

What each tunnel gives you

Accuracy figures are from the public validation suite, run on every release.

RCGeneralSupersonic
Speed range2–60 m/sup to Mach 1.2 (transonic estimate above the critical Mach)Mach 1.2–12, up to 80 km
Physics3D CFD with actuator-disk propellers (blade-element momentum theory, two measured blade families) 3D lattice-Boltzmann large-eddy simulation, Kármán–Tsien compressibility corrections Tangent-cone/wedge, Newtonian, Prandtl–Meyer, Eckert, Fay–Riddell
Key outputsThrust, shaft power, prop efficiency, drag in the slipstream, top speedDrag polar, drag and power vs speed, pressure and flow fields Wave, base and friction drag, wall temperature, heat flux, centre of pressure
OptimizesAirframe drag with the props turningDrag across the envelopeDrag or peak temperature, within material limits
Typical sweepMinutes (one run per angle and speed)Minutes (one run per angle)Seconds
Checked againstUIUC propeller data, 2–4 blades: thrust and power within 8%; slipstream vs momentum theory within 1% Sphere drag within 5%, cube within 11%; streamlined bodies conservative Sphere drag within 5%; thin-plate lift within 2%; stagnation heating within 7% of Sutton–Graves
Not yet trustworthyWing lift (see General)Lift on thin wings, 50–66% low; the transonic band is an estimateLow-supersonic body normal force; real-gas chemistry above Mach 6
How it works

Four steps, whichever tunnel

Import

Upload STL, STEP, IGES, OBJ or 3MF. Set the units and which way the nose points; the mesh is checked and repaired automatically.

Constrain

Protect screw holes, mounts and fins; keep the payload or motor inside; place propellers; give hot spots their own material.

Sweep

A range of airspeeds or Mach numbers and a range of angles of attack. Every result covers every combination.

Optimize

Choose what may change, how far, and how many designs to try. Download the winner in your own units and orientation.

One project, any tunnel

Switch a project between tunnels at any time: test a sounding rocket subsonically, then take the same model through Mach 4 and check its nose cone against the fibreglass limit.

RC
T(V), P(V)
thrust and power at every speed
General
Cd(α, V)
drag polar across the envelope
Supersonic
Twall(M)
against each material's limit
All tunnels
−Δ drag
optimized shape, exported
Everything in the box

A test lab and a design optimizer, in the browser

No meshing, no solver setup, no cluster to manage.

Spinning propellers

Blade-element momentum theory sized from the numbers on the propeller (10×7 at 9,000 rpm), injected into the CFD as thrust and swirl.

Heating and materials

Stagnation and surface heating, radiative-equilibrium skin temperature and a material library with service limits — or your own.

Constraints you can see

Click the model to place protected regions, keep-inside volumes, propeller hubs and material regions. The optimizer respects all of them.

Whole-envelope sweeps

Ranges of airspeed or Mach number and angle of attack instead of one point; designs are scored across all of it.

CAD-style features, automatic redesign

Add nose and tail cones, fairings over exposed parts, pods, wings and fins, cut slants or stretch a wing on your own upload. The optimizer reshapes any dimension you leave unlocked and learns which ones matter.

Export and report

Download the optimized STL in your own units and orientation, plus a printable engineering report.

Under the hood

Solvers and optimizer built from first principles

No commercial CFD underneath: every method below is implemented in this codebase, tested, and checked against published data.

GPU lattice-Boltzmann CFD

D3Q19 regularized collision with a Smagorinsky large-eddy model and curved-wall (Bouzidi) boundaries. The CUDA and CPU kernels are generated from one source and agree to 0.2 %; on a laptop GTX 1650 Ti the tunnel runs at about 400 million cell updates a second.

9 s 3-angle rocket sweep vs 56 s on 6 CPU cores

An optimizer that knows its own error

Before searching, it re-runs the baseline shifted by a fraction of a cell to measure the solver's noise, ranks every named parameter in a screening pass, then searches the strongest ones with a Gaussian-process model. Gains smaller than twice the measured error are not reported.

−7.7 % supersonic rocket drag 28 designs in 30 s

Wings at fixed lift

Vortex-lattice lifting surfaces with NACA 4- and 5-digit sections, viscous section polars from a neural surrogate (NeuralFoil), stall and stability margins, and a fuselage drag model — so a wing is never "improved" by simply lifting less.

−15.8 % wing drag at fixed lift NACA 2412 → 2408

Supersonic to Mach 12

Second-order shock-expansion theory along each meridian, modified Newtonian flow, Fay–Riddell stagnation heating and radiative-equilibrium skin temperature checked against each material's service limit.

±7 % ogive drag, Mach 1.5–4 vs published data

Propellers in the loop

Blade-element momentum theory for each propeller, injected into the flow as an actuator disk with thrust and swirl, so the airframe is tested in its own slipstream.

≤ 8 % thrust and power vs UIUC wind-tunnel data

Validation on every change

54 reference cases — sphere and cube drag, a streamlined body, finite wings, NACA 0012, cones, ogives, stagnation heating and propellers — re-run on the GPU in about four minutes. Failing cases stay on the public page.

4 min full validation suite was ~60 min on the CPU
Honest physics

Validated against published data, open about its limits

Every tunnel is checked against reference data on every release: sphere and cube drag, propeller wind-tunnel data, linear supersonic theory and stagnation-heating correlations. Where a result is not yet trustworthy we say so, next to the number.

View validation results

Questions

Which tunnel should I use?

Anything with a propeller that matters for its drag: the RC tunnel. Rockets, UAVs, pods and fairings up to Mach 1.2: the General tunnel. Rockets, missiles, re-entry bodies and high-speed aircraft above Mach 1.2: the Supersonic tunnel. A project can switch tunnels at any time.

How is the transonic range handled?

The General tunnel's CFD is computed at low Mach number and corrected for compressibility (Kármán–Tsien) face by face, which is sound up to the body's critical Mach number — where the flow first goes sonic, typically Mach 0.7–0.85 — and that number is reported for every body. Above it, shocks form; the drag rise is estimated by bridging the corrected CFD to the supersonic surface methods at Mach 1.2. That band is an estimate and is labelled as one in every result: a true transonic answer needs a compressible flow solver.

How are propellers simulated?

Each propeller's thrust and torque come from blade-element momentum theory using its diameter, pitch, blade count and rpm, checked against UIUC wind-tunnel data for APC propellers. That loading is applied to the flow as an actuator disk, so the slipstream and its swirl wash over the airframe. Blades are not resolved one by one.

How hot can it get?

The Supersonic tunnel runs to Mach 12 and 80 km. Above Mach 6 real-gas chemistry starts to matter and above Mach 8 it is significant; the methods use a perfect gas, so temperatures there are flagged as approximate.

What is a compute credit?

One draft CFD run costs 1 credit (standard 3, high 8). In the General tunnel a run is one angle of attack; with propellers turning it is one angle and one airspeed. A whole supersonic sweep costs 1 credit, and supersonic optimizations 1 credit per ten designs.

Is the Lite plan really free?

Yes: all three tunnels, 40 compute credits a month, one propeller, short optimizations, no card required.

Who owns my designs?

You do. Models and results are private to your account and deleted when you delete the project or your account.

Put your airframe in the right tunnel

Free to start. Upgrade for more fidelity, more propellers, wider sweeps and export.

Create a free account