Do Flat Floors Slow You Down?

[Griffith500]

...

You simply can not admit where you are wrong and try to defend yourself by being patronizing. That is 'why the attitude'. Again you write paragraphs without actually addressing what you were relying to. Even the link said nothing about aerodynamics being the art of increasing drag, it just went into detail about different types of fluid behaviors. Cant you just simply admit that saying things like dimples in golf balls increase drag was plain wrong?

I can see what you were trying to say, you meant to speak about different ways of decreasing drag, but you didn't and ended up saying the opposite. It would kill you to admit you were wrong it seems.

I am perfectly capable of admitting I am wrong - for instance, I was mistaken about the flow separation occurring sooner (on the object) in the turbulent regime, it actually occurs later, which reduces the width of the wake etc. That's because I remembered the effect and falsely re-inferred the cause, so my reasoning wasn't wrong, at least.

I wasn't wrong by saying that adding dimples increases surface drag, which typically reduces aerodynamic performance - I qualified that by saying it was paradoxical that it should improve the overall performance of a golf ball. However, the special case of the golf ball was being used as a general rule of thumb, which is wrong - that's what my "paragraphs" were trying to explain to you. Your inability to comprehend that, especially in context (which you may have missed in your haste to stick your oar in, probably because, like CSLACR, you feel you must avenge yourself for some perceived slight), isn't my problem, as much as you want it to be.

I love the irony in your accusing someone else of not being able to admit they're wrong, though. Do regale us again of how conservation of energy doesn't apply to bicycles (by denying Newton's laws of motion as applied to "free" bodies), or how rolling resistance is more important to a road car's top speed than aero drag (which is the only way mass would matter).

Anyway it seems this whole argument is about nothing.

That hasn't stopped people spouting nonsense about golf balls, "air molecules" and cosmic speed limits, though, has it? You contributed to that nothingness by missing the point for so long.



Do let us know how the test with the R34s goes, though. That's what we ought to have been doing pages ago. 👍

 

[ryzno]

But we're talking about what limits the top speed of a car. Light speed physics doesn't matter, because no car will ever come close to it.

Mass certainly doesn't matter for top speed in GT6 and here's an experiment to prove it:

Take a stock Red Bull X2014 Standard to Route X. Adjust the gearbox to allow a top speed of 400 km/h or more. Do a lap and make a note of the top speed (the downhill part doesn't count).

Then add 200 kg of ballast to the car. This equals a mass increase of 35%. If mass limits the top speed then surely an increase of that size would be noticable. Do another lap and make a note of the top speed.

Here's my results:

Run 1: 395 km/h
Run 2: 395 km/h

Conclusion: Mass is irrelevant.

Try a less biased car. That doesnt require the adjustment of gears.
Ford Ka:
0 ballest, 110 mph.
200 ballest, 109 mph.

I did a complete lap for each test. Useing the back strech as the judge. No hill just flat out.

I know you will probly say oh poo, 1mph. Ok your right.
But if the engine cant over come the aero drag, plus added resistance of weight, it will never achive your said results. Right?

IRL why do Real Racing teams worry about weight? Because it makes a difference be it only 1mph... And acceleration but thats another story.

PS no physics brain here, just what I observed.

 

[Griffith500]

Try a less biased car. That doesnt require the adjustment of gears.
Ford Ka:
0 ballest, 110 mph.
200 ballest, 109 mph.

I did a complete lap for each test. Useing the back strech as the judge. No hill just flat out.

I know you will probly say oh poo, 1mph. Ok your right.
But if the engine cant over come the aero drag, plus added resistance of weight, it will never achive your said results. Right?

IRL why do Real Racing teams worry about weight? Because it makes a difference be it only 1mph... And acceleration but thats another story.

PS no physics brain here, just what I observed.

Or, put another way, an increase in mass of over 20% resulted in less than 1% decrease in top speed. Well, the limits are between 0.45% and 1.4%, because of rounding of the speed readout in the game.

Compare aero: an increase in Cd (or cross-sectional area or air density) of 20% will result in an almost 9% decrease in top speed (i.e. 100*(1-sqrt(1/1.2))). That's an order of magnitude more important already, and there's no way to know that PD got the contribution of rolling resistance right.

Another member independently found the ratio between rolling resistance and aero drag (at speed) was 1:20 using basic physics. (EDIT: GT6's numbers supplied suggest a ratio of only 2.5:1)

You're right, of course, it does have an effect; but the faster you go, the less it actually matters in real terms. Thanks for doing the test and supplying the numbers. 👍

 

[eran0004]

Try a less biased car. That doesnt require the adjustment of gears.
Ford Ka:
0 ballest, 110 mph.
200 ballest, 109 mph.

I did a complete lap for each test. Useing the back strech as the judge. No hill just flat out.

I know you will probly say oh poo, 1mph. Ok your right.
But if the engine cant over come the aero drag, plus added resistance of weight, it will never achive your said results. Right?

IRL why do Real Racing teams worry about weight? Because it makes a difference be it only 1mph... And acceleration but thats another story.

PS no physics brain here, just what I observed.

There's an added mass that needs to be accelerated which means that the engine needs to work for a longer time before it achieves the same speed. If all other things remain the same, added mass does not give a lower top speed.

In this case, some things do change. Rolling resistance would increase, for instance, because the tyre pressure is adjusted for a car that's 200 kg lighter. The X2014 probably have a much higher tyre pressure than the Ford Ka (because the X2014's tyre pressure would be adjusted to deal with the additional "weight" from the massive downforce generated by its aero parts), which would explain why the difference is greater for the Ka than for the X2014.

Real racing teams worry about weight because of acceleration, braking and cornering.

 

[ryzno]

Ok, but we cant adjust tire pressure. So dont you think that can be a possible factor?
Im not arguing the theory of engine torque/power vs. top speed, in perfect conditions. But another possible flaw in GTs physics.
edit:By flaw I mean non adjustable psi or knowlage of actual psi.

 

[CSLACR]

If that's all that matters to you, then yes, oh how you foiled me with your wit! But you're right, I don't typically trouble my brain with things my ass can figure out. 👍

Your ass didn't figure it out, so once again, this doesn't make sense.

It's not me "foiling you with my wit", to be clear, it's "you saying dumb stuff".

Stick to the aero stuff, you were doing great with that. (really)

 

[Griffith500]

Your ass didn't figure it out, so once again, this doesn't make sense.

It's not me "foiling you with my wit", to be clear, it's "you saying dumb stuff".

Stick to the aero stuff, you were doing great with that. (really)

I was aware of the switch in negation, which was deliberate. I was implying I didn't bother to engage my brain in response to the drivel you wrote, which had no bearing on the discussion.

I.e., if it truly were trivial (whether the flat-floor behaviour in the game is "accurate"), I'm sure you wouldn't have been compelled to post the "dumb stuff" you came out with that prompted me to respond in the first place. So it's quite amusing to see you say it's "trivial" when you got it wrong. There's no shame in getting it wrong, of course, it's more the subsequent scramble for the moral high ground that drained my patience.

Indeed, if it were so trivial at all, we wouldn't be 180 posts deep on the subject. Well, maybe half that if we ignore the speed of light nonsense, etc.

 

[Exorcet]

Skimming over the thread, someone said you can't know the real answer (the question being do flat floors add drag) without CFD/wind tunnel. Technically true. However the general trend for a flat floor should be lower drag and lift. Since GT6 probably isn't using CFD or real data specific to each individual car, I'd expect them to use the general trend.

Even if we're talking about a flat floor without a diffuser, the the floor should provide an overall benefit. Without the flat floor you can get stagnation points under the car which result in drag and lift. The air is just going to hit whatever is sticking out from under the car and increase pressure. Since it's under the car, that pressure is going to be pushing the car up. Adding the flat floor removes these stagnation points, so the drag is lowered. The flat floor also allows the flow to maintain total pressure, so Bernoulli's equation applies and pressure is lowered by an increase in air velocity. Air velocity will tend to increase under the car because the stagnation point of the nose is higher than the ground clearance so the dividing streamline that hits that stagnation point is forced down under the nose of the car. This leads to a decrease in cross sectional area for the flow, which means the flow must go faster.

Throw in the diffuser and you get additional downforce from streamline curvature at the throat. Past the throat cross sectional area increases, flow slows down, and pressure rises. However the floor is no longer parallel with the ground so some of that increase in air pressure goes toward pushing the car forward. Drag is lowered as a result.

 

[Griffith500]

I was wondering when you'd show up, Exorcet! I like the "bulk" descriptions you've given, makes a change from the "it's all black magic" view, or even the microscopic view you get with the theoretical basis of fluid mechanics (i.e. my own perspective).

It seems, though, that the "flat floors" in GT6 just add "downforce" in the usual way; i.e. using the same calculation as for the wings etc., yielding downforce and drag in the same relative proportions as those other items.


As a slight aside, what's your take on the potential for a coarse / crude CFD code to be run in-game (in the menus only, not in-race, e.g. with a "windtunnel" backdrop - which we've seen...) for when you make changes to the car's shape etc.? Does that sound a) feasible and b) reliable enough for a game?

 

[UltimateLurker]

Even if we're talking about a flat floor without a diffuser, the the floor should provide an overall benefit.

I agree with everything you said except this. There really is no such thing as no diffuser. Air exits the rear of the car in a controlled, or non-controlled manner, and this is essentially a diffuser, be it a bumper cover lip, or purposeful air vanes. If you add a flat floor to the bottom of the car without controlling the air at exit, you run the risk of creating lift.

Back in the early 80's, Reeves Callaway tried to do this with the Callaway twin-turbo Corvette, and the result was that the rear tires would come off the ground at about 160 mph, but would stay firmly planted without the flat floor.

Now, I don't pretend to know everything about this or how it all works, but I do have first-hand knowledge of the event. As a result, no 1987 Callaway Corvette's were made with a flat floor.

 

[Exorcet]

It seems, though, that the "flat floors" in GT6 just add "downforce" in the usual way; i.e. using the same calculation as for the wings etc., yielding downforce and drag in the same relative proportions as those other items.

Yes, it seems like a poor modeling choice on PD's part. Wings create drag when generating lift and there are specific relationships at work for wings that quantify how much you get. Lift tends to produce drag with other shapes, but the simple equations that outline how it works with wings don't apply to arbitrary shapes (like cars). One huge factor is that drag is not a singular thing and the balance of drags (between pressure, induced, and skin friction) are wildly different for a wing when compared to a car. Even if a diffuser by itself creates drag, it can lower drag on the car overall.

As a slight aside, what's your take on the potential for a coarse / crude CFD code to be run in-game (in the menus only, not in-race, e.g. with a "windtunnel" backdrop - which we've seen...) for when you make changes to the car's shape etc.? Does that sound a) feasible and b) reliable enough for a game?

True CFD isn't really an option in game, good results would be too resource hungry and the level of detail you would need to get real results from an underbody simulation is something that not even premium cars capture. You could run a really coarse simulation, but by the time is becomes computationally feasible you would probably be better off extrapolating from simplified equations or real world data. It has been done in a few cases, like X-Plane:



which allows users to design a custom plane, the aero of which is calculated real time. Techniques like that are good at the overall level (747 vs P-51) but are less impressive when it comes to fine detail. The other problem is that wings are really simple compared to cars.

If PD really wanted to move away from generic aero performance from aero parts, they would probably need to work with manufacturers and use their wind tunnel/CFD data to come up with specific aero behavior for each car. I think in the sim's current state, a good compromise is using actual equations on wing shape and position with generic modeling for the flow field around the car based on body type (coupe, sedan, hatchback, etc).

I agree with everything you said except this. There really is no such thing as no diffuser. Air exits the rear of the car in a controlled, or non-controlled manner, and this is essentially a diffuser, be it a bumper cover lip, or purposeful air vanes. If you add a flat floor to the bottom of the car without controlling the air at exit, you run the risk of creating lift.

Back in the early 80's, Reeves Callaway tried to do this with the Callaway twin-turbo Corvette, and the result was that the rear tires would come off the ground at about 160 mph, but would stay firmly planted without the flat floor.

Now, I don't pretend to know everything about this or how it all works, but I do have first-hand knowledge of the event. As a result, no 1987 Callaway Corvette's were made with a flat floor.

I can't say what the problem might be with the Corvette without looking at it, but in general you do want to control the airflow. One of the downsides of flat floors is more complex underbody flow interaction with tires because the airflow tends to be faster and at higher total pressure. Some road cars with flat floors tend to have tire deflectors as well to control this, but this starts to get into a higher level of detail than what we're looking at in GT.

The lack of a diffuser at the rear of a car shouldn't be an insurmountable problem. If you simply cut the floor at the back of the car, you get a clean separation edge which shouldn't cause any huge flow issues or lift. It's basically like the top of a SUV with the addition of ground effect.

 

[Griffith500]

Yes, it seems like a poor modeling choice on PD's part. Wings create drag when generating lift and there are specific relationships at work for wings that quantify how much you get. Lift tends to produce drag with other shapes, but the simple equations that outline how it works with wings don't apply to arbitrary shapes (like cars). One huge factor is that drag is not a singular thing and the balance of drags (between pressure, induced, and skin friction) are wildly different for a wing when compared to a car. Even if a diffuser by itself creates drag, it can lower drag on the car overall.

Yes, it seems those interactions can cause lots of confusion when relying on anecdotal evidence!

True CFD isn't really an option in game, good results would be too resource hungry and the level of detail you would need to get real results from an underbody simulation is something that not even premium cars capture. You could run a really coarse simulation, but by the time is becomes computationally feasible you would probably be better off extrapolating from simplified equations or real world data. It has been done in a few cases, like X-Plane:

www.youtube.com/watch?v=O_NDeSPCMks

which allows users to design a custom plane, the aero of which is calculated real time. Techniques like that are good at the overall level (747 vs P-51) but are less impressive when it comes to fine detail. The other problem is that wings are really simple compared to cars.

Oh, I didn't mean anything properly real-time, I was thinking a coarse "offline" solver with reasonable delays whilst it runs the simulation in the background - like the delay involved in taking a photomode picture, say. Again, I don't know how well that might work, although I expect leveraging the parallel computation potential on the PS3 or, better, PS4 might help if precision isn't an issue (you must have seen GPGPU creep into workflows near you ). There are a lot of "tool programmers" involved with GT over recent years, too.

I assume X-Plane is using the "basic" 2D airfoil model, taken as slices along its length (with corrections for non-parallel flow?) - it's pretty cool, I remember something similar in Rigs Of Rods. I can see why that wouldn't work well for a car, given a wing is wider than it is long, generally, and the effects of interest on a car occur typically along its length. As such, I think even look-up tables could be better: so as an example you'd pre-compute lift and drag as 3D vectors for relative angles of incidence of the free stream, perhaps in 2 or 3 dimensions also. That's potentially a lot of computation, but it might be possible to fill in the gaps / blend to decent enough "effect" between fewer actual computations - but I haven't looked at it at all, so I could easily be wrong.

It might be better for PD to do this before shipping the game, and then applying some kind of real-time correction as you make modifications - e.g. an airfoil model for the rear wing, "corrected" by things like the turbulence coming off the body etc., or an increase in lift for high, er, "angles of attack" for flat-floors...

If PD really wanted to move away from generic aero performance from aero parts, they would probably need to work with manufacturers and use their wind tunnel/CFD data to come up with specific aero behavior for each car. I think in the sim's current state, a good compromise is using actual equations on wing shape and position with generic modeling for the flow field around the car based on body type (coupe, sedan, hatchback, etc).

...

Or they could do this and parameterise it some way, just the same, assuming it all comes in the same format from each manufacturer. That depends on how "reliable" the data provided is, of course, but then it's no worse than PD's in-house code "favouring" certain cars because of particular, I don't know, turbulence models used, or whatever. At least PD can blame the manufacturers for "inaccuracies"!

 

[UltimateLurker]

I can't say what the problem might be with the Corvette without looking at it, but in general you do want to control the airflow. One of the downsides of flat floors is more complex underbody flow interaction with tires because the airflow tends to be faster and at higher total pressure. Some road cars with flat floors tend to have tire deflectors as well to control this, but this starts to get into a higher level of detail than what we're looking at in GT.

The lack of a diffuser at the rear of a car shouldn't be an insurmountable problem. If you simply cut the floor at the back of the car, you get a clean separation edge which shouldn't cause any huge flow issues or lift. It's basically like the top of a SUV with the addition of ground effect.

Agreed.

 

[Exorcet]

Oh, I didn't mean anything properly real-time, I was thinking a coarse "offline" solver with reasonable delays whilst it runs the simulation in the background - like the delay involved in taking a photomode picture, say. Again, I don't know how well that might work, although I expect leveraging the parallel computation potential on the PS3 or, better, PS4 might help if precision isn't an issue (you must have seen GPGPU creep into workflows near you ). There are a lot of "tool programmers" involved with GT over recent years, too.

I can't really give a precise breakdown on how much computing power or time it would take to run a background aero solver and I'm not really the most knowledgeable when it comes to computers. From what I know of the PS4, it's roughly on par with one of my better laptops which takes a few hours to run a decent simulation on a FSAE car using SolidWorks Flow Simulation. The license for that is about $10,000. There are cheaper programs that are probably a bit less accurate, but might still be useful. They would also have simulation times in hours on a single computer. For comparison, auto manufacturers use multiple thousand core computing centers with each simulation taking about 100 or more cores. These simulations run anywhere from a half day to a day or two (typical) to a week.

A built in GT aero solver might use the panel method (breaks the shape up into panels and control points of flow tangency) on a simplified aero model, which could run under an hour, but I'm not really sure where it would realistically fall in terms of accuracy/time.

I assume X-Plane is using the "basic" 2D airfoil model, taken as slices along its length (with corrections for non-parallel flow?) - it's pretty cool, I remember something similar in Rigs Of Rods. I can see why that wouldn't work well for a car, given a wing is wider than it is long, generally, and the effects of interest on a car occur typically along its length. As such, I think even look-up tables could be better: so as an example you'd pre-compute lift and drag as 3D vectors for relative angles of incidence of the free stream, perhaps in 2 or 3 dimensions also. That's potentially a lot of computation, but it might be possible to fill in the gaps / blend to decent enough "effect" between fewer actual computations - but I haven't looked at it at all, so I could easily be wrong.

X-Plane using Blade element theory, which is basically as you described. The wing is broken up into multiple pieces, and the forces on those pieces are computed individually and then summed. This allows each part to see different conditions and to react differently to the flow. With the wing, there exist equations to are reasonable approximations to real behavior. For a car you would need to pre compute forces forces like you said. I wouldn't want to try and shortcut this process as it could easily lead to a lot of error since flow separation is so much more prevalent on a car than on a wing.

It might be better for PD to do this before shipping the game, and then applying some kind of real-time correction as you make modifications - e.g. an airfoil model for the rear wing, "corrected" by things like the turbulence coming off the body etc., or an increase in lift for high, er, "angles of attack" for flat-floors...

Yes, this is what I was saying before. You can model a wing reasonably well. To also model the effect of the wing on the car, you need the flow field around the car.

Or they could do this and parameterise it some way, just the same, assuming it all comes in the same format from each manufacturer. That depends on how "reliable" the data provided is, of course, but then it's no worse than PD's in-house code "favouring" certain cars because of particular, I don't know, turbulence models used, or whatever. At least PD can blame the manufacturers for "inaccuracies"!

In the auto industry I think codes are more standardized, while in aviation there are more proprietary codes. In either case, they will probably all agree very well on the level of detail that GT would use.

 

[Holdenhsvgtsr]

I'm just gonna leave this here

 

[Griffith500]

Haha, I saw "Crimsafe" and it immediately made me think "Aussie"!
And no, that's not a reference to the original colonists, just the endearingly straight-forward nature they engendered. The advert says it all: "if it isn't Crimsafe, it's not crim-safe!"

Nice video. 👍

I can't really give a precise breakdown on how much computing power or time it would take to run a background aero solver and I'm not really the most knowledgeable when it comes to computers. From what I know of the PS4, it's roughly on par with one of my better laptops which takes a few hours to run a decent simulation on a FSAE car using SolidWorks Flow Simulation. The license for that is about $10,000. There are cheaper programs that are probably a bit less accurate, but might still be useful. They would also have simulation times in hours on a single computer. For comparison, auto manufacturers use multiple thousand core computing centers with each simulation taking about 100 or more cores. These simulations run anywhere from a half day to a day or two (typical) to a week.

Sure, I think the reason it takes so long is because it's relying on the standard "general purpose" cores in your laptop, which, being general purpose, aren't really that suited to doing specialist stuff like flow simulation - the computation involved just doesn't make use of the full range of hardware "processes" available on such a chip, so most of it sits idle whilst the FP / vector unit gets hammered (well, something like that). I mentioned GPGPU as there is a big push now towards using the vast potential of what is effectively very cheap, massively-parallel, floating-point hardware to accelerate the process - graphics hardware has lots of capacity for the kinds of computations required in these problems. The PS3's Cell is also pretty good at FP calculations; its folding@home performance was around 30 times that of x86 CPUs in 2008.

Obviously, that requires modifications to the code, from the ground up, unless it's already highly parallelised as in the case of your car manufacturer example; but even then the "instruction set" is very different, so the actual computation steps will need to be re-optimised. I'm not sure what the speed-up is, but it's at least a factor of two, maybe five or more - I've read about some cases where a factor of 20+ was possible (based on cost), but I have no hands-on experience with any of it, and it clearly depends on the hardware gap.

Really, the only reason to use an existing software package is because of the complete workflow potential they provide (and the validation and support etc.), and in research it is very common to (re)write custom code (maybe first in something like MATLAB, then the code monkeys port it to something a little more sprightly like C++ or whatever). The techniques are well understood enough by the numerical boffins, and the improvements actually start in research (naturally) and filter down into the commercial products.

So I wonder if some of the more recent techniques can find application for "approximate" pseudo-simulation in a game - an example I've seen for wave simulation is to use pre-defined wave-equation solutions for specific geometry (e.g. spheres, rectangles or whatever) and find a way of applying that to general geometry. That had a speed-up of over 5 times before translation to GPUs (for equivalent resolution), and a tuneable reduction in accuracy, exchanged for some of that speed-up, "to taste". You just need someone with experience in implementing general numerical solvers, really.

I don't know that it's really hugely important how it's done (and I don't think the "brute force" methods used for serious industry applications need apply here), but I have to wonder how they managed when CFD was first starting being used - I guess there was a lot of careful reduction in degrees of freedom by some means, and I suspect that can still be done today.

A built in GT aero solver might use the panel method (breaks the shape up into panels and control points of flow tangency) on a simplified aero model, which could run under an hour, but I'm not really sure where it would realistically fall in terms of accuracy/time.

What sort of scale do the important flow effects exist on? That is, ignoring time taken for now, how coarsely can you sub-divide the car's surface like that and still get "accurate" results? I remember running a steady-state simulation (is that sufficient?) of half a "car" that was only crudely modeled, but it had all the right "features" in the solved flow field (or so I was told). It only took a matter of ten or twenty minutes, probably 8 years ago on typical academic "lab" computers of the time.

So the question is whether the sensitivity of that kind of thing is subtle enough between different shapes, and not necessarily if the absolute scale of forces it spits out are right (because you can massage them overall), just that between shapes, the relative differences are correctly represented. Again, it's very likely I don't know what I'm talking about here.

X-Plane using Blade element theory, which is basically as you described. The wing is broken up into multiple pieces, and the forces on those pieces are computed individually and then summed. This allows each part to see different conditions and to react differently to the flow. With the wing, there exist equations to are reasonable approximations to real behavior. For a car you would need to pre compute forces forces like you said. I wouldn't want to try and shortcut this process as it could easily lead to a lot of error since flow separation is so much more prevalent on a car than on a wing.

Yes, this is what I was saying before. You can model a wing reasonably well. To also model the effect of the wing on the car, you need the flow field around the car.

Good info, thanks. I wonder, then, if some "flow separation" model can be relied upon - should such a thing exist. There's all sorts of weird empirical relationships floating around from the time before engineers could even dream about "direct" simulation, but it'd require the input from someone with extensive knowledge of all of that. For example, much of it would have inspired "experiments" to take to the wind tunnels, just as simulation does today. I suppose the details of car aerodynamics are subtle and sensitive enough that the error could be large if you gloss over the wrong bits, but since Adrian Newey is Kaz's best mate now, maybe they can borrow an aerodynamicist to show them what to focus on!

In the auto industry I think codes are more standardized, while in aviation there are more proprietary codes. In either case, they will probably all agree very well on the level of detail that GT would use.

I meant more the marketing influence, and any data provided could easily be massaged by selecting the outputs in certain ways, even if PD stipulate a "recipe". It's probably preferable to just let them do that, rather than suffer the wrath of their lawyers because your panel-method shortcut "favours" certain shapes over others, as an arbitrary example.
But I take your point regarding level of detail. 👍

 

[TexRex]

Where is this diffuser you speak of?

Love that Alpine. Where was the shot taken? I used the shoulder bump just before the front straight on Trial Mountain to get a look at cars' undersides but it wasn't always capable of tortoising a car.

 

[Exorcet]

What sort of scale do the important flow effects exist on? That is, ignoring time taken for now, how coarsely can you sub-divide the car's surface like that and still get "accurate" results? I remember running a steady-state simulation (is that sufficient?) of half a "car" that was only crudely modeled, but it had all the right "features" in the solved flow field (or so I was told). It only took a matter of ten or twenty minutes, probably 8 years ago on typical academic "lab" computers of the time.

So the question is whether the sensitivity of that kind of thing is subtle enough between different shapes, and not necessarily if the absolute scale of forces it spits out are right (because you can massage them overall), just that between shapes, the relative differences are correctly represented. Again, it's very likely I don't know what I'm talking about here.

The scale depends on what you're looking for specifically. You can get a good idea of the flow away from the surface with a fairly course simulation, though in that case there is no guarantee that flow features you're seeing are correct. Vortices can form where there are none, or they can be the wrong size, etc. In general, more flow separation or sharp pressure gradients require more resolution. Separation is normal for cars, especially dedicated race cars with fins and wings everywhere. It would be a lot easier to get an idea of the aero around the Auto Union than it would be to do the same with the FGT or DTM cars. You can also get correct features for one shape, but be very much off on another. It's hard to judge much about the example you bring up. You could say that there are different levels of features, with high level details being easier to replicate. If you want to know what the flow really looks like around a car down to the surface, you'll be talking mm level resolution. If you're not interested in the surface, you can get away with inches of resolution. Going over a foot might make the differences between specific cars trivial. This is a really short answer by the way, as you could easily write papers on this of course.

Steady state simulations should be sufficient most of the time. Highly unsteady flow isn't desirable and auto makers will tend to eliminate it, though the cars in GT may go far past what the engineers were thinking the car's limit would be.

I haven't really used a full scale panel method simulation, but with CFD tools like in SolidWorks you wouldn't want to go to a coarseness measured above centimeters or inches.

Good info, thanks. I wonder, then, if some "flow separation" model can be relied upon - should such a thing exist. There's all sorts of weird empirical relationships floating around from the time before engineers could even dream about "direct" simulation, but it'd require the input from someone with extensive knowledge of all of that. For example, much of it would have inspired "experiments" to take to the wind tunnels, just as simulation does today. I suppose the details of car aerodynamics are subtle and sensitive enough that the error could be large if you gloss over the wrong bits, but since Adrian Newey is Kaz's best mate now, maybe they can borrow an aerodynamicist to show them what to focus on!

There are models for flow separation that can be added to potential flow solvers, I think it might still take a bit too long to solve for your average player, but I lack experience with them so I can't give a solid estimate. Newey's input would be valuable for a solver or some kind of simplified equation based model, but I still think the best method would be to apply general trends to cars based on their overall shape. Maybe this can be refined with information about their specific geometry, like height, width, etc.

 

[Griffith500]

The scale depends on what you're looking for specifically. You can get a good idea of the flow away from the surface with a fairly course simulation, though in that case there is no guarantee that flow features you're seeing are correct. Vortices can form where there are none, or they can be the wrong size, etc. In general, more flow separation or sharp pressure gradients require more resolution. Separation is normal for cars, especially dedicated race cars with fins and wings everywhere. It would be a lot easier to get an idea of the aero around the Auto Union than it would be to do the same with the FGT or DTM cars. You can also get correct features for one shape, but be very much off on another. It's hard to judge much about the example you bring up. You could say that there are different levels of features, with high level details being easier to replicate. If you want to know what the flow really looks like around a car down to the surface, you'll be talking mm level resolution. If you're not interested in the surface, you can get away with inches of resolution. Going over a foot might make the differences between specific cars trivial. This is a really short answer by the way, as you could easily write papers on this of course.

Steady state simulations should be sufficient most of the time. Highly unsteady flow isn't desirable and auto makers will tend to eliminate it, though the cars in GT may go far past what the engineers were thinking the car's limit would be.

I haven't really used a full scale panel method simulation, but with CFD tools like in SolidWorks you wouldn't want to go to a coarseness measured above centimeters or inches.




There are models for flow separation that can be added to potential flow solvers, I think it might still take a bit too long to solve for your average player, but I lack experience with them so I can't give a solid estimate. Newey's input would be valuable for a solver or some kind of simplified equation based model, but I still think the best method would be to apply general trends to cars based on their overall shape. Maybe this can be refined with information about their specific geometry, like height, width, etc.

Cracking answers, thanks for taking the time to explain all of that (and indulge my silly questions)! 👍

So the short answer, which you started with, is that true simulation will not be in the game any time soon. But either trivial models or highly simplified models can give some of the "expected" behaviour "we" are after.

That probably seems like I'm stating the obvious to some, but it's taken a little while for what you said to sink in:

If PD really wanted to move away from generic aero performance from aero parts, they would probably need to work with manufacturers and use their wind tunnel/CFD data to come up with specific aero behavior for each car. I think in the sim's current state, a good compromise is using actual equations on wing shape and position with generic modeling for the flow field around the car based on body type (coupe, sedan, hatchback, etc).

Sounds like it should be "lively" enough, depending on how one implements the generic flow-field part. Now someone just needs to do it!