CFD - Computational Fluid Dynamics, Motorsport, Formula 1

[miqi23]

Yeah mate, I thought to get something going on this thread. It has been dead for such a long time now..

[alex1015]

I have FLUENT and Inventor available to me. Let's get something going here. I'm thinking a specific part of an F1 car and follow the changes over the season. This way we make the main model once and then follow it as it evolves and then run CFD calculations to see the corresponding changes in numbers.

[slimjim8201]

I have FLUENT and Inventor available to me. Let's get something going here. I'm thinking a specific part of an F1 car and follow the changes over the season. This way we make the main model once and then follow it as it evolves and then run CFD calculations to see the corresponding changes in numbers.

You have a PM. Contact Maureen to get CFdesign for your FSAE team. It is much better suited for design engineering. Then you can use it for pretty much whatever you want.

[miqi23]

Out of interest Slimjim8201, what is the maximum cell size model you can produce with CFDesign? Has it got a 64 bit platform yet?

[slimjim8201]

Out of interest Slimjim8201, what is the maximum cell size model you can produce with CFDesign? Has it got a 64 bit platform yet?

Good question. We've had an X64 version for years. The largest I've tested on a PC was about 15 million elements. The PC had 8 GB of ram and a single, dual core 2.4 ghz xeon processor. I believe one of the young bucks tried something with even more than that a few months back. Theoretically, our limit is the operating system's limit, which in the case of X64, is a LOT. The problem then is the solve times. Almost all of our users are operating on a single node. That is they are designing and simulating on a single PC. Not a Cray or some crazy UNIX cluster. Our distributed computing capabilities will take a huge leap with the next version, but even then, most of our users will still be running on a single node.

[alex1015]

You have a PM. Contact Maureen to get CFdesign for your FSAE team. It is much better suited for design engineering. Then you can use it for pretty much whatever you want.

Not to be a jerk but I don't see one, maybe I'm doing something wrong. I have FLUENT available through another means. I spent some time in a lab at my school and we do indeed have CFdesign. Personally though I am just a second year student so I haven't yet used it.

I just meant I will have FLUENT available to me this summer so that I can model some flows without having to be at school.

[slimjim8201]

You have a PM. Contact Maureen to get CFdesign for your FSAE team. It is much better suited for design engineering. Then you can use it for pretty much whatever you want.

Not to be a jerk but I don't see one, maybe I'm doing something wrong. I have FLUENT available through another means. I spent some time in a lab at my school and we do indeed have CFdesign. Personally though I am just a second year student so I haven't yet used it.

I just meant I will have FLUENT available to me this summer so that I can model some flows without having to be at school.

I apologize, it was another forum member that I sent a PM regarding our FSAE program. It's been a hectic week...

I wouldn't wish Fluent on any Mechanical Design Engineer. I would like that 6 months of my life back.

[AeroGT3]

You have a PM. Contact Maureen to get CFdesign for your FSAE team. It is much better suited for design engineering. Then you can use it for pretty much whatever you want.

I am going to go out on a limb here, because I don't know what kind of scripting is available with CFDesign. BUT, I would argue that Fluent is BETTER for design engineering because of the accuracy and the scripting available. My team has fully automated optimization programs using fluent and gambit scripting. We've also done the same for full car solutions, from meshing to submitting jobs the cluster, fetching them, and post processing/writing images and movies. We have people who've never touched CFD able to run this stuff afer 1-2 hours of training. But unlike most "design" CFD software, we have control over the mesh. I doubt that'd be possible with CFDesign?

[slimjim8201]

You have a PM. Contact Maureen to get CFdesign for your FSAE team. It is much better suited for design engineering. Then you can use it for pretty much whatever you want.

I am going to go out on a limb here, because I don't know what kind of scripting is available with CFDesign. BUT, I would argue that Fluent is BETTER for design engineering because of the accuracy and the scripting available. My team has fully automated optimization programs using fluent and gambit scripting. We've also done the same for full car solutions, from meshing to submitting jobs the cluster, fetching them, and post processing/writing images and movies. We have people who've never touched CFD able to run this stuff afer 1-2 hours of training. But unlike most "design" CFD software, we have control over the mesh. I doubt that'd be possible with CFDesign?

CFdesign has full scripting capabilities and full control over meshing. It is intended for the mechanical/thermal design engineer. Someone coming from CAD, designing a new part or new system and they want to test various changes. BC changes, moving components around, adding/removing things, etc. It's all about the optimization process early in the design process. I respect the guys that write a code to run 500 simulations to hone in on the exact, 100% most efficient design, but early in the design process, an engineer needs an optimized solution, not a perfect solution.

That is not to say that CFdesign is not accurate. Underneath it all, most CFD codes are solving the same equations using different methods. I think the biggest difference is this:

A mechanical design engineer is building a new intake manifold. He's a CAD user most days, a CFD user once every month. He has four new intake manifold profiles and he needs to find out which one is the least restrictive, which one has the best flow distribution, and at the end of all that, he want's to take the best design and see the effects of opening and closing the throttle plate on internal flow fields. How long would it take a once in a month user to figure this out in Fluent (and Gambit)?

I like how you guys use Fluent, but someone had to set that up. You guys are taking a great traditional CFD tool and using a script to grant accessibility to mainstream people. CFdesign is accessible to mainstream people out of the box. Please don't get me wrong. I'm an ex Fluent user and I completely respect what it can do. The reality is this: CFdesign and Fluent are not direct competitors usually.

[miqi23]

Totally agreed that it does all the good stuff a Design Engineer needs from a Desktop point of view. It is also good to know that it has a 64 bit platform as well and you have done a 15 million cell model as the biggest case so far..

My question is - How long did it take CFDesign to produce that 15 million cell model? You said most cases are run on one node, is it possible to utilise all the nodes like on a Quad-core?

Indeed CFDesign is good for internal flows but how does it cope with external flow problems? Do you have any validation data to show how good it is for external aerodynamics applications?

[slimjim8201]

Totally agreed that it does all the good stuff a Design Engineer needs from a Desktop point of view. It is also good to know that it has a 64 bit platform as well and you have done a 15 million cell model as the biggest case so far..

My question is - How long did it take CFDesign to produce that 15 million cell model? You said most cases are run on one node, is it possible to utilise all the nodes like on a Quad-core?

Indeed CFDesign is good for internal flows but how does it cope with external flow problems? Do you have any validation data to show how good it is for external aerodynamics applications?

As the old adage goes, "crap in = crap out". External aerodynamics require a hefty mesh to capture what needs to be captured. The fact remains that without a stellar mesh quality, external mesh results won't be super accurate. We are primarily focused on upfront design engineering, the computers that MCAD guys use just plain aren't up to par with what is needed for a full external aero simulation. The capabilities are there, the scalability, the mesh requirements, the solver, the advection schemes and so on and so forth, but our users don't have machines for that sort of stuff on average. No matter what CFD tool you are using, a full formula car is going to require 30-50 million cells.

CFdesign takes advantage of everything necessary for this sort of stuff, but it's not our target user base.

[miqi23]

Ok good stuff, but how long did it take CFDesign to mesh that 15 million cell model?

[slimjim8201]

Ok good stuff, but how long did it take CFDesign to mesh that 15 million cell model?

45 minutes.

[connollyg]

Guys,

What makes you think 30-50 Million is good enough for F1? my understanding is that the top teams are talking 100 Million and above, and i know there has been talk of billion cell models!

G

[connollyg]

Ok good stuff, but how long did it take CFDesign to mesh that 15 million cell model?

45 minutes.

So how big does a model have to be to take 10+ hours on a 2K core cluster?

Rumour has it that more than one team is looking at core counts above this to try and reduce run times to around the 8 hour mark, the more backward teams are looking at weeks!

G