CFD in racing simulators

Yes. See the model @AccAkut showed of his Audi, the numbers were extrapolated/interpreted into something the game understands.

 

nice catch!
i wonder how its going to work and if we will be able to choose wind direction?
also wonder what affect it will have on cars and if you will see cars struggle to take corners like in bahrain gp.

 

does kunos or any mods use this method to get data?

 

That Audi is a mod... http://www.racedepartment.com/downloads/krb-audi-s1-silhouette.14447/

And its unlikely Kunos do their own CFD, why would they when they presumably get the data from the manufacturers?

And its a lot of work to get a model suitable for CFD (standard poly models that all games use aren't suitable.) Modders have time to make the required changes, Kunos don't.

 

I think by the time a quantum computer can fit into my phone and tell me the best aero setup on the car, the ai piloted ones would've driven circles around me already. #roborace #obsolete

 

There's these expensive "treadmill" wind-tunnels for a reason. CFD causes error (obviously) and no matter how good it gets, what's the point if you have wind-tunnel data available to model aerodynamics?

Okay if the manufacturer is *****ing about giving the aero-data then you could make a CFD-simulation and model aero from the results. Yes, you could introduce more elements to the game with realistic AOA, yaw and speed-sensitivity qualities and that would produce way less error in the end result than running a realtime CFD with all of it's bugs.

 

CFD modelling is just an approximation of reality. Just like Assetto Corsa runs a model to simulate aerodynamics, CFD is another model, albeit more complex, which is run using finite elements analysis.
The current state of the art in CFD can only predict accurately simple flows on simple geometries. CFD is another tool in the engineers design tool box, wind-tunnel testing being another one. Today, they are complementary as each has its pros/cons and benefits.

Having said that, the computing power to run CFD models on current computers (even supercomputers) is extremely limited. Modellers normally have to simplify meshes tremendously, and even simplify geometry (2D over 3D, planar-symmetric, radially-symmetric, etc) vs a true 3D representation. A fully CFD analysis of a complete car does not exist today, even with a supercomputer. All that exist is a simplification of a 3D model to give approximations. Furthermore, simplifications will never cease to exist. For example, multi-physics modelling. Even if we have full 3D CFD model, we would need to add other physics to increase accuracy such as thermal modeling, mechanical deflections (flexing elements, etc), all running also with finite elements do play a role in how fluid forces affect a surface. Maybe in the future, who knows, but definitely not today.

As somebody else mentioned earlier on the thread, diminishing returns will kick-in way earlier than having a CFD model run in real time. We can have 99% of accurate results with perhaps <1% of the processing power. Even if we did have that amount of processing power to do real-time CFD, I think it would be wasted running CFD. To answer this very interesting question raised by Poguinhas, I would rather use such CPU for something more meaningful. It would probably be better used running more complex models of tyres, dynamics, brakes, AI, etc.

 

Well, but I did consider a scenario 100 years from now. 100 years ago computers didn't even exist, so I don't think it's that much to believe that 100 years from now we will have the processing power necessary for 40 cars running full 3d cfd in real time.
Of course a lot of things will change during that time, simulators, if they still exist, will probably have taken several changes in their direction.
But yes, if with the simplified data we use today we can have 99% accurate results, using real time cfd would be way too much effort for the reward, it's just that some times I find it interesting to imagine what direction innovation will follow, and many years from now cfd may be one of them. Especially considering that probably 100 years from now sims will be the only way to drive a petrol fueled car, for good or for worse.

 

Cd numbers probably. (race cars unlikely)

Cl numbers unlikely. (especially if its race cars, road cars maybe but even that is a stretch.)

As for AC aero simulation, it does the job and its seems very accurate in AC compared to real world.

 

Personally I think that Realtime Tyre Simulation should come before Realtime CFD aero.

Both of which are well beyond the capability of home computers, but discussions like this are a can or worms, do we want another rip in the fabric of time and space like after "zandvoort not laserscanned"-gate.

 

I don't think there is even that much non-Realtime simulation for tires actually, compared to what is possible in aero development. Maybe for slicks, but as soon as you enter some sort of tread pattern simulations won't get you far, and its down to real life testing on facilities like this. So many iterations of development tires went through the workshop when I worked there.