Flow simulation¶

Push air or water past a part, or spin the part in it, and see what the flow does. Use it to compare designs: which shroud has less drag, which way a fan pumps, where the pressure sits on a blade.

Wind tunnel¶

A part held still in a moving stream.

1. Open a part and start the Flow operation from the toolbar.
2. Leave the mode on Tunnel.
3. Pick the fluid, air or water. The speed follows it: 10 m/s in air, 1 m/s in water. Water is a thousand times denser, so air's speeds pose a problem this method has no business answering.
4. Set the direction the stream blows from and its speed.
5. Optionally pick a torque axis. The result then also says how hard the stream twists the part about it, which is the question behind "would this vane turn, and which way". Pick it like a rotation axis (below); leave it alone and the result omits that line. Changing it later re-reads the run you already paid for instead of starting another.
6. Check the flow domain, the box drawn around the part. It is sized automatically; drag a face or type into the fields to change it, Auto puts it back.
7. Pick a resolution. These are time budgets rather than mesh sizes: draft, standard and fine aim at roughly one, five and ten minutes. The line beneath shows the cell size and the estimated runtime.
8. Solve flow. A ring in the corner notification fills as the run steps through its budget.

Rotation¶

A part spinning about an axis: a fan, an impeller, a stirrer, a turbine.

1. Switch the mode to Rotation.
2. Set the speed in rpm (negative reverses it) and how many revolutions to run.
3. Pick the rotation axis: click the field, then an origin axis, a body edge, or a circular edge such as a hub bore, which contributes the axis it turns about. A dashed line shows what you picked. A part whose blades are a circular pattern comes with the axis already filled in.
4. Domain, resolution and solving work as above. The domain for a rotating body is a loose duct around it rather than a tunnel.

Reading the results¶

The result appears in the right-hand card with the fields drawn in the viewport.

• Drag force / Thrust and Moment: the force the fluid puts on the body and the moment of that loading. A tunnel calls it drag, a rotation calls it thrust, which is what the fan pushes back against its mounting with. Hover the moment for its components and the point it is taken about: the bounding-box centre for a tunnel, the rotation axis for a rotation.
• Torque @ axis (tunnel, when you picked a torque axis): how hard the stream twists the part about that axis, so whether a vane, flap or unpowered rotor would want to turn and which way. Positive turns right-handed about the axis. This is as trustworthy as the drag force; the caveat below is about a body you drive, not one the flow pushes on.
• Flow rate (rotation): how much fluid the part moves, in litres per minute.
• Torque (rotation): how hard the motor has to turn the part, marked approx. Multiply by angular speed for shaft power. See below.
• Lattice: how many cells and steps the run used.
• converged or step cap (tunnel): whether the flow settled, or the run used its whole budget without settling. A step-cap result is still a result, averaged over the run's last stretch, but less trustworthy. A rotation reads fixed run instead: it turns exactly the revolutions you asked for and has no settling test to pass.

Three layers, each toggleable: surface pressure on the body, streamlines coloured by speed (they draw themselves in from inlet to outlet when a result appears, so the flow direction is visible before you touch anything), and a cut plane through the domain showing speed or pressure, off until you ask for it.

Anything the solver noticed appears in amber under the numbers, in plain words: a body poking into the damped edge of the domain, features too thin for the lattice to resolve, a run that did not settle.

Accuracy¶

The solver puts the part on a grid of cubic cells, so its surfaces come out as tiny steps rather than smooth curves. That is what makes it fast and it is also what bounds the answer.

• Refine when a feature is thin. At a coarse preset a 3 mm fan blade may be one or two cells thick, not enough to resolve what it does to the flow. The solver warns when the whole body is that thin; a finer resolution is the cure.
• Trust directions and comparisons more than absolute numbers. Which of two shrouds has less drag is answered well. What the drag is, to two figures, is not.
• A rotating body's torque is an order of magnitude, not a figure. On the fan this was developed against, three resolutions spanning a factor of 1.75 in cell size agree on the sign and stay within a factor of 1.3 on the size: enough to say "about six watts to turn this", not enough to choose a motor on. Flow rate and direction from the same run are in the same class. The Torque @ axis line of a tunnel study is a different and better number, measured on a body standing still, and is as good as the drag force beside it.
• Check the convergence badge. A "step cap" result has not settled. Raising the flow-throughs count or coarsening the domain gives the run more time. Tunnel studies only; a rotation's "fixed run" badge is not a warning, and flow_throughs is not one of its settings.