CFD in racing simulators

As sim racers, I think I'm safe in assuming that most of us are, to some degree, realism-freaks. To one degree or another, most, if not all of us has at some point been concerned with the features of simulators to recreate real racing, wether it was when choosing what sim to buy or to complain about something not properly done.

Anyway, I'm not really up to par with the progress of CFD today, as far as I know, even in F1 it can't properly recreate all the vortices and flows around their really complex bodywork, I've read that NASA at the moment may be the only ones with such resources. But even if that's the case, we're walking towards a time when wind tunnels are gonna be useless (I think).

What I'm wondering is: When will we be able to go on Steam and buy a software that's capable of simulating a grid with 50 cars, all their wings, wheels, vortices, dirty and clean air, using CFD instead of simplified values to determine drag and downforce? Also, when will we be able to go on a store and buy a computer capable of handling that? Maybe 15 years from now, maybe 50, maybe not even on 100?

What do you guys (especially the more informed on it) think about this? Not that I'm that much of a realism nerd, it's just something I've been wondering.

 

First the devs need to choose the right model for the CFD
https://turbmodels.larc.nasa.gov/

 

If you assume Moore's law and that it takes a couple hours to reach a consistent solution right now that implies that in about 50 years hardware will be fast enough to do at realtime. I'm not sure how that works in the real world though, computers are approaching some actual atomic limits on component sizes where it won't really be possible to keep putting more transistors into chips.


I don't know if it's worth chasing that goal though, you have to look at the magnitude of errors caused by any given simplification vs. how much of your computing time it'll need. For example if a winglet's providing 0.1% of your overall downforce, then improving the simulation of that winglet by 5% changes your overall downforce by 0.005%. If anything else in your simulation is wrong by more than that amount, like for example you measured the springs as 68000N/m and they're actually 68003N/m cause of variable spring rate, then you're not focused on the right thing.

More likely it'd go the direction of flight simulators where they pick a reasonable number of elements (hundreds rather than millions) and use a non-realtime simulation to determine coefficients to feed into the simple one.

 

This is much more likely. Run your game 3d model into CFD to build a complete relevant aeromap and then use that data in a realtime simple model.

Full realtime CFD in a game is not happening imo.

There is somewhat of a misunderstanding even with today's CFD. Most people not in the know think it's as simple as putting a 3d model into the software, click play and wait 20 hours for the results. It's incredibly more complex to get a result that correlates to the real thing than we have been led to believe. Most of the recent CFD "work" in racing has been only that...trying to get the same numbers the windtunnel is giving.

 

...as demonstrated by Red Bull's recent issue where they seem to have three different results from CFD, Wind Tunnel testing and the track. But it would be interesting to imagine having computers capable of running a full grid of cars in CFD in real time. Let's not forget that the computers used to put the man on the moon for the first time were slower than our current phones (If it's on the internet it must be true).

 

Beware...The law of diminishing returns!
Having pre-determined values doesn't mean it's a bad thing. Is it bad that we've pretty much all agreed on the gravitational constant? Or should we try to recalculate it on the fly each time? I'll got with the former. I know what you're asking @Poguinhas and it's a great question for a philosophical discussion, however, I'm not intelligent enough to provide you with good answers. Will we ever get there? Maybe...Will it be any more accurate 100 years from now compared to 90 years from now? Personally, I don't know. I'm 42 years old now....Let's say I can hopefully live another 20 years -- I'm sure we will have way cooler sims than we have now, especially with how advanced VR should be in 20 years...but, I still don't think we'll be even close to doing all calculations in real time. I don't think it would do anything besides waste CPU cycles.

 

If in terms of real world simulation then you would really need to map and calculate every air particle on the track, allow for how it reflects off all objects, wind directions and other cars. Simply getting real time single car cfd is not likely and using it to get data then inputting this into simpler formulas that is close enough to the limited amount of variables chosen to be included is really as good as it will get in most of our lifetimes for simulation games.

You would also need to calculate all the moving parts of the car, how the air deals with this and effects air flow and in the end your mostly wasting cpu cycles for a game when it can likely all be done much simpler where resources can always be used elsewhere even when we have hardware to burn. In the future there will need to be huge steps in processing power just to up the resolution to match the human eyes limit in terms of resolution for a vr type display that will replicate the fov and clarity of what we see. Calculating fluid dynamics in real time will be the same as using real time ray tracing for graphics and both seem unlikely for a very long time.

 

I am pretty sure it is going to be used as a marketing stunt sooner than later tho

 

Computational fluid dynamics (CFD) is a branch of fluid mechanics that uses numerical analysis and data structures to solve and analyze problems that involve fluid flows. (Google is a wonderful thing)

 

[ASSETTO_CORSA]
SLEEP_TIME=-1
WARNINGS_AS_ERRORS=0
SIMULATION_VALUE=0 ; increase this for more simulation value
COMPUTATIONAL_FLUID_DYNAMICS=0 ; increase this for even more simulation value

 

it produces numbers mostly, but also visual representations like this that help you design a car and understand what those numbers mean:


like in this case from numbers alone you probably would not see the huge vortex created at the back, which influences the whole underbody aero

 

i dont know why but i read this as "it will be simulated in pcars2".

 

Speaking of aero, I'd appreciate having winds in a sim.

 

You never know what they are working on :O
https://www.assettocorsa.net/forum/i...me-track-yt-channel.44177/page-12#post-863692

 

i'm just wondering if the aero data of cars (mods and official contents) is calculated from data. or is it possible to do that.

 

Data calculated from data? Not sure what you mean...

 

so let's say you put in a 3d model of the car into one of the cfd simulators. can you use the output data from CFD and put it back to AC.