CFD - Computational Fluid Dynamics, Motorsport, Formula 1

[MadMatt]

Yes, or you can also use the Cooper body which is very similar. Its major characteristic is a diffuser:



Ignore the sting that was for correlation with wind tunnel experiment. SAE paper (if you got access) related to this body is: "The aerodynamic performance of automotive underbody diffusers", number 980030.

On a side note, I just got a new computer with 32GB of RAM and although I haven't tried to push it at its maximum yet, I am confident I can mesh models of more than 30 millions cells easily. The problem then is the computational time. With 20 millions cells model in steady state, it goes at around 2 iterations per minute, so would take between 2 and 3 days to reach convergence. Just to give some numbers there.

I will post more when I get the chance to try this.

[KeithYoung]

Yes, or you can also use the Cooper body which is very similar. Its major characteristic is a diffuser:

http://img19.imageshack.us/img19/5903/l4o5.jpg

Ignore the sting that was for correlation with wind tunnel experiment. SAE paper (if you got access) related to this body is: "The aerodynamic performance of automotive underbody diffusers", number 980030.

On a side note, I just got a new computer with 32GB of RAM and although I haven't tried to push it at its maximum yet, I am confident I can mesh models of more than 30 millions cells easily. The problem then is the computational time. With 20 millions cells model in steady state, it goes at around 2 iterations per minute, so would take between 2 and 3 days to reach convergence. Just to give some numbers there.

I will post more when I get the chance to try this.

At the SCCA Runoffs this weekend I talked to a guy that designed his own cars. He said if he tried to use the maximum mesh that his computer would crash in post.

[pwlucas]

What mesh sizes are you guys running? I run anywhere from 5 million for simple case and have gone up to 25 million. I do not do open wheel cars however.

[MadMatt]

Full car from 5 to 20 millions.

[pwlucas]

Full car from 5 to 20 millions.

Assuming this is half car using symmetry? or full car in yaw?

[hollus]

Maybe this is a bit of a stupid question.
When making these CFD analyses, I assume that the conditions of the air entering the "virtual tunnel" and exiting it have to be given as boundary conditions. In that case, how is total drag actually "measured" (calculated)?
Or maybe only the "incoming" air has to be defined this way?

[shelly]

Drag is calculated as surface integral of pressure force + friction force on the body, projected in the horizontal direction

[hollus]

So boundary conditions (air speed, direction, temperature and density) are specified both at tunnel entry and tunnel exit?

[shelly]

Yes.

[hollus]

Thanks.
So how is the actual difference calculated / decided? My instinct says that the loss of momentum of the air crossing the tunnel (determined by said boundary conditions) should match losses due to the car and losses due to the "rolling road". Said otherwise, the total loss of energy of the air should match "car drag" and "road drag" (from the air that the car has set in motion). But one of those terms is one of the things one is trying to "measure"... Is there some sort of iterative change of the boundary conditions until this matches? Are they matched by hand? By fudge factor?
Also, is the "exiting air" boundary condition uniform over the whole end-of-tunnel plane? That would not be a very realistic boundary condition.
Feel free to ignore me, just musing about why CFD sucks so bad at quantifying drag.

[PNSD]

CFD is great at calculating pressure drag, and pressure forces. CFD is just not great at calculating viscous drag and forces, but then is also related to the mesh size.

For bluff body aerodynamics CFD is pretty fantastic

[MadMatt]

Full car from 5 to 20 millions.

Assuming this is half car using symmetry? or full car in yaw?

The numbers I gave are for the full car yes, as you would use directly for yaw simulations for example. If you take half car and use symmetry, you save half the number of cells (assuming you don't want to simulate yaw), but the results are a bit different (especially due to the rotating wheels). On the other hand it will take much more time to reach convergence, assuming your computer can handle 20 million cells models!

But if you just have to do straight line simulations, using full car is not really vital. Hope it helps!

I think CFD sucks at calculating drag because all the small surface imperfections/surface finish is never modeled properly in the software. Either because the software doesn't allow you to, or either because the engineer is not bothered to look into that, otherwise I think the results would be much better. And they are not THAT bad. It is common sense to say, but CFD results alone cannot be taken for granted.

[shelly]

it depends on what you mean for normal PC. 16gb Ram? 32-64? Openfoam is a free software whose perfrmance is alredy good enough for complete car cfd (search internet). I suggest you learn how to use it; then you can get good results even on a 16gb ram pc; tht's alredy enough for rans on half-car with reasonable level of detail.

[MadMatt]

Hello guys,

I have 500hours of "power-on-demand" simulation available for Star-CCM+. I am not going to use these as I have license at work, so I am willing to sell this. I am just asking for a bit of money so I can buy a nice book so if you are interested, make me an offer in private!

It is valid until the 22nd of March 2014.

[Owen.C93]

Never new they did that type of license. I heard they charge per CPU core though.