Force history drag and lift coefficients over time
Reading the numbers
t* counts how many times the air has travelled one object length past the model. Averages settle after roughly 10 to 30.
Cd and Cl are averaged over the last stretch of the run once it settles. The lamp turns green when the average stops moving.
Simulated Re may sit below the real one. Fine-scale turbulence the grid can't resolve is handled by a Smagorinsky model, so treat results as engineering estimates.
Check it yourself. Validation, in the top bar, runs textbook cases against published wind-tunnel data. The sweep below plots drag and lift across a range of speeds or angles.
Sweep drag and lift across a range of airspeeds or angles
Force history drag and lift coefficients over time
The 3D tunnel
Same physics as 2D, one dimension up: 19 velocity directions per cell (D3Q19), a Smagorinsky turbulence model and curved-wall bounce-back, all on the GPU through WebGPU.
Surface colours show pressure or speed just off the skin. The slice cuts through the flow; drag its slider to sweep it.
Vortex cores mark where the air spins faster than it stretches (Q-criterion): trailing vortices off a car's rear pillars, rings behind a sphere.
Your own models. Import an STL, OBJ, glTF/GLB or 3MF file (or drop it anywhere on the page). It is turned into a distance field once, so it runs at any angle and resolution, and the Body tab works out its printed mass, balance and stability.
How far to trust it. A browser grid carries Re ≈ 10⁴ at most, so a wall model sets skin friction at the real Reynolds number. Sharp-edged shapes land close to wind-tunnel data (a square plate within about 5 % on Medium). Rounded bodies read high: a sphere about 20 % on Medium, and the Ahmed car-like reference body roughly twice its measured drag on Medium and 1.6 to 1.8 times on High and Ultra. Most of its excess shows up on the nose, where suction around the rounded edges should nearly cancel the pressure on the face. Run Validation (top bar) to see how your grid does. Comparisons between shapes and settings are more trustworthy than single numbers.
Speed along the flight by distance downrange
How the flight is worked out
Every millisecond or so the solver adds up gravity, drag and spin lift on the projectile and steps its position forward (4th-order Runge–Kutta). Air density, speed of sound and gravity change with height along the way.
Drag comes from the drag data you pick. Built-in curves are measured 3D data for each shape family and rise sharply near the speed of sound.
Stability compares where the air pushes (centre of pressure) with where the weight sits (centre of gravity). If the push is ahead of the weight, the projectile flips and tumbles.