Exhaust Blown Floor - Forward Exhaust Exit

[hollus]

How long do your CFD's take?

[PNSD]

For the simulation, do you make your own meshes then or does the program do it automatically? I thought floworks worked slightly differently...

[hardingfv32]

Does increased flow entering under the front of the side pod help or hurt the aero properties of the side pods?

Brian

[shelly]

Shelly i hear you, but what I am saying is that the exact same mesh can run on a slower system.
Users only limit their mesh because of sanity. Time and your electrical bill is the enemey with a slower computer. Worse if the CFD is being used commercially.

If i a run a 1 million cell mesh on a super computer, I can run 1 million cells on a slower one. One will take an hour the other may take a week.
Results will be the same if settings are exactly the same.

But if i try to economize on time, i may cut down cell count on a slow computer and also other refinements to try to make the time equal. That is where the slow computer fails, when you turn down the calculation details to compensate time wise.
That's the compromise, especially in F1 where time is critical with resource restrictions ect.

The same mesh with the same solver will give you the same result (ignoring compilers' details). But you need a powerful computer to run a big enough mesh; and you need a decent quality mesh also.
So if your mesh is not big enough you can not make reasonable predictions, and you can not say where the pathlines go.

[ringo]

So if your mesh is not big enough you can not make reasonable predictions, and you can not say where the pathlines go.

How long is a piece of string?

How can you know the threshold for reasonable and unreasonable?

In this case, mesh density is a bigger concern than mesh size. It provides reasonable results, as was shown form real time pictures and video of the R31 at both high and low speeds.
It's not a coincidence the results look striking similar to the real thing.

On the floor sucking in air:


In the USF1 shot, you see the flow being squeezed. The exhuast flowing to the diffuser wont like this as it will be directed head on into the tyre. It will be shut out from the diffuser by the denser slower flow.

In the second shot...
Make what you will of it, but notice what the turning vane is doing.

[shelly]

With "big enough" I was not clear. I mean a big enough mesh file, with a big number of elements, so that they are small enough to have sufficient resolution.

How big? Depends on which type of mesh you are running (tri, penta+tri hybrid, hexa, poly?) With penta+tri a reasonable threshold is 50-100M cell for half car model I think.

"striking similarity with real thing" is still to be found.

The two images you posted are intersting, especially the second; but we are seeing only two path lines, we do not see more pathlines in the bargeboard zone or vector plots under the diffuser.

[hardingfv32]

What is the criteria for reasonable predictions? A good correlation with windtunnel results or track data?

Brian

[ringo]

How long do your CFD's take?

Depends on many things. I find complex geometry takes longer.
I never did one longer than 12 hours(energy bill ). Sometimes it seemingly will go forever and you can save the calculation and take the results you have now and continue later.
If you find the numbers are converged and don't seem to be making any meaningful changes you can stop.

But i think the topic has diverted to criticizing the CFD. We should be focusing our energy on finding out what those renualt engineers said exactly.

We don't have any quotes from them, why is that?

If this thing is targeted at the diffuser according to them, why is it going there colder and slower than the other designs?

Colder and slower is not desirable, and would defeat the purpose of retarded ignition which promotes high speed and hot exhaust.

If renault find the car is falling back will they switch to a redbull design?

These are the questions that need to be asked!

[ringo]

What is the criteria for reasonable predictions? A good correlation with windtunnel results or track data?

Brian

Track data in the form of youtube videos in the rain or dust.

[hardingfv32]

1) My question was not an effort to "to criticizing the CFD". Why do you provide such a weak response to my question? Is the CFD questionable in this case?

2) The engineers view our discussions with amusement. Their input is of almost no value unless supported by some form of logic.

3) I for one can make much better evaluation if I know there is agreement on the subject optimum aero design of the side pods. If I know what is best for the side pods then the the location of the exhaust becomes a question of compromise.

So what makes the side pods work in the grand scheme of the aero design?

Brian

[PNSD]

How long is a piece of string?

How can you know the threshold for reasonable and unreasonable?

Ermm, grid convergence?

In CFD, if you can not afford the time, then you are wasting your time IMO.

Grid size and density is by far the biggest and most important variable, and grid convergence is the only way of knowing whether a mesh is suitable or unsuitable apart from how accurate the result is ofc.

[flyboy2160]

while ringo is doing a most excellent job putting down his detractors with not only logic and calculations, but also with wit and humor, i'm going to chime in in his defense against his detractors ( i can hear it already: a ringo "fanboi." )

you people bitching about his analysis are forgetting that he's NOT running a 10 day tergazillion flop /sec solution on a supercomputer to get the "final solution" just before making the 1 million dollar tooling to fabricate the car.

he's doing something akin to the most efficent thing to do at the start of a design analysis: a quick and dirty first look using limited resources to see if the idea somebody sketched out on a napkin makes any sense and is worth pursuing with more time and more money or is hopelessly bogus.

i suspect many of those bitching have either never run a real cfd themselves or are the old fashioned run-up-the-score types who are against any quick and dirty cfd because it threatens their nobody-else-knows-how-to-do-it jobs. (i have run cfd for money and always get into trouble with the run-up-the-score guys.)

[shelly]

@flyboy: I can not understand from your post if you are serious or ironical.

If you are serious, I think you have missed the point, which is: if your cfd is not run with a good level of accuracy, its answers are not correct.

If you are ironical, your languge and ton eare too heavy, and in the end not appropriate

[mep]

I have to agree with flyboy. Poor ringo has a really hard stand here. Some guys attack him wherever they can. Until now he is the only one who has put up any CFD work to this discussion. You guys should be much more respectful to each other. It seems some guys in this forum (not just this topic) just fight to get right and prove the other wrong.

[marekk]

I respectfully disagree we can assume RANS run on half-car, 1 milion cell mesh with very simple F1-like model and youtube video as grid convergence proof can give us (almost) the same end results as multi iterations LES run on full 3D, at least hundreds of bilions of cells mesh, real F1 CAD model, boundary set to at least 5 car's lengths, 3 width's and 3-4 hight's, and grid convergence criteria based on real world track/ wind tunel data combined with years of experience and knowledge of operators.

And you simply can't run those problem sizes on on your few GB of RAM PC - people don't live that long.

That said, i'm still open to any arguments other then "boy you have on idea".