Virtual wind tunnel · STEM Racing & rockets

CFD-Class

CFD-Class-Racing →
ModelNone loaded
Grid–
Cell size–
SolverStarting…
Version3.1.1

Your wind tunnel is empty.
Upload a STEP file of a STEM Racing car or a model rocket, or open a saved project.

01Behind the simulation

How does it work?

01From CAD to triangles

A STEP file stores the exact curved surfaces of your design. The OpenCascade CAD engine (the same one inside FreeCAD) turns each surface into thousands of small flat triangles. Mesh detail controls how closely the triangles follow the curves.

02Building the wind tunnel

Your model is placed in a virtual box of air, split into tiny cubes called voxels (like 3D pixels). Voxels inside the model become solid. There is space in front, lots of space behind for the wake, and room at the sides. The sides and roof are open: air pushed aside by your model can flow out, just like in open air, so the tunnel walls don't squeeze the flow.

03Lattice-Boltzmann method

Each voxel holds 19 "packets" of air moving in different directions. Every time-step, packets hop to the next voxel (streaming), bounce off your model's surface, then mix together (collision). The bounce uses the exact position of the surface inside each voxel, so curves aren't treated like Lego blocks. Repeat this thousands of times and realistic airflow appears. Your graphics card does millions of these updates every second using WebGPU.

04Pressure, speed and drag

Where air slows down against the nose, pressure rises (red). Where it speeds up around curves, pressure drops (blue). Every packet that bounces off your model gives it a tiny push. Adding up all those pushes gives the drag force (this is called momentum exchange). Dividing by the wind's "push" and the frontal area gives the drag coefficient Cd. A lower Cd means less drag. The split into pressure drag and friction drag comes from reading the pressure and the air speed just next to every triangle.

05Why is Cd only for comparisons?

Real airflow depends on the Reynolds number (speed × size ÷ air's viscosity). A real STEM Racing car is around 250,000. In Standard mode the grid can only simulate a few hundred to a few thousand. That's like testing your car in syrup: the airflow pattern is similar, but friction is far more important than in real life, and the flow stays smooth (laminar) instead of turbulent. So the numbers are useful for ranking your own designs against each other but not as real-world values.

Precision mode gets closer: the air is about 5 times thinner, the cells are smaller, and a turbulence model handles the swirls that are too small for the grid. It is still not a professional CFD result (the thin layer of air right against the surface is smaller than a cell), so compare precision tests with precision tests.

06Precision mode: smaller cells, turbulence, two stages

Smaller cells capture your shape better (thin wings, the gap under a car) and let the air be thinner. Each time the number of cells doubles, a run takes about 2½ times as long.

The turbulence model (Smagorinsky large-eddy simulation) adds extra "eddy viscosity" only where the air is being churned up, standing in for swirls smaller than a cell.

With the zoom box (experimental), stage 1 simulates the whole tunnel. Stage 2 then re-runs only a smaller box hugging your model, with smaller cells, using the air flow stage 1 found at the edges of that box. It is like a camera zooming in.

Long runs keep the screen awake and save a checkpoint every 5 minutes, so a run can carry on if something interrupts it.

07Tips for good results
  • Export a single solid body (or a few touching solids) from Fusion, Onshape, Inventor or SolidWorks.
  • Check the blue wind arrow and model orientation before pressing Run.
  • Only compare designs run with the same accuracy setting, wind direction and ground setting — the test list does this for you.
  • Treat differences of less than about 5% as "no clear difference" — tiny changes in where the grid falls can move Cd by a few percent, especially under a car.
  • For STEM Racing cars, use CFD-Class-Racing (spinning wheels, the CO2 cartridge and the race-time predictor). For rockets, point the wind along the rocket's length and switch the ground off.
  • Changing the shape (nose cone, wings, pods)? Compare the pressure drag part too. In Standard mode friction is exaggerated and can hide shape differences.