How does Aerodynamic Simulation work?

[Whirligig Girl]

Obviously we all know how the stock Kerbal drag system is horrible and bad. We also know that FAR is a great replacement because it is able to work for a ship/plane of any shape.

But how does that work? How does it compare to X-Plane? X-Plane uses [Blade Element Theory] the shape of the vehicle to simulate the air; while other simulators like Microsoft Flight Simulator and Orbiter use predefined conditions for each plane, making them useless for design work. Does FAR just use predefined conditions for each part to simulate craft on any shape, or does it do something more advanced?

And how accurately does FAR model the real life aerodynamics?

Edited October 24, 2014 by GregroxMun

[Nuke]

lots of them use table lookup and interpolation between the two nearest samples (which is how math libraries handle things like trig and square roots). what i did (or rather started to do) was use simplified physics models for the vital parts of the aircraft, such as wings and fusalage segments. i then determine the four forces (thrust, drag, lift, gravity) operating on each one. linear forces just use the sum of all the forces on all physics models, i think angular forces just take the cross product between the position of the object and the force, and sum the results for all objects. it works for the most part, but its too simple to handle things like supersonic/hypersonic flight.

[rkman]

If you want to know how FAR works you'd be better off asking Ferram. He did some explaining in the FAR thread.

[paul23]

Normal simulations use FEA -finite element analysis-. Where they simply solve the differental equations numerically, there are many ways to solve differential equation numerically. The simplistic (and terrible, comparable to using bubble sort in programming) way is just calculate it step-by-step from an initial value (that's the forward euler method as described on the wikipedia article).

[K^2]

Normal simulations use FEA -finite element analysis-. Where they simply solve the differental equations numerically, there are many ways to solve differential equation numerically. The simplistic (and terrible, comparable to using bubble sort in programming) way is just calculate it step-by-step from an initial value (that's the forward euler method as described on the wikipedia article).

And if you try to apply this to a real wing, you'll watch the simulation fail miserably. The problem is that there are a crap ton of vortex states generated by any air foil, and they come in all shapes and sizes. So if you use a fixed grid FEA, you end up missing separation layer, and generating almost no lift. If you simply check for vortex condition and sub-tesselate your grid, you end up with too many elements to do anything with and your simulation stalls.

What people really end up doing is a horrible monstrosity stitched from FEA, particles, and a whole bunch of heuristics. And they still, generally, end up working just for specific kinds of geometries. For a game, it might be sufficient to do laminar flow (inviscid, incompressible) FEA and hacking in a fake separation layer that's going to be treated as a boundary by your FEA. It won't be perfect, and you would never, ever do something like this for a proper engineering simulation, but it might be close enough for a simulation game.

Fluid dynamics is just one of these problems, you know?

As for FAR, I haven't looked at it too closely, but as far as I know, it's a relatively simple version of blade element with some pre-defined polars for the blade elements. Anything sensible for a game like KSP will be some variation on that, to be honest.