[CFD Newbie]

Hello everyone,

Just wanted to ask people with background in aero cfd if they had used (and their experience with) Powerflow from company Exa. They claim to have 14 of the top 15 car manufacturers in the world as their customers, but I know that Audi switched from Powerflow to Openfoam.

1. Is Powerflow better than traditional aero-simulation software aka Star-CD, Fluent or Cfx?
2. Is LBE better / more accurate than Navier Stokes for aero-simulation?
3. If you are one of Powerflow's commercial users (for example in one of these car companies) - what software did you use before Powerflow? Are you also using their aero-acoustic/thermal features? What do you think of them? Also, how critical is Powerflow to your design process? Can you survive using what you used earlier? Can you use Openfoam like Audi - if so why dont you use that?

Thanks very much!

Regards,

Wannabe Aerodynamic Expert

[cfdnewbie]

Is LBE better / more accurate than Navier Stokes for aero-simulation?

This question is somewhat vague, are you referring to the equations themselves or the codes based on them?
As long as the continuum assumption holds, NS is as good as LB, since NS is essentially derived from LB for a very small Knudsen number.

Can you use Openfoam like Audi - if so why dont you use that

Just from my experience with industry: not only Audi but also the biggest suppliers of these companies (companies even bigger than Audi) are shifting from commercial to open software - mainly because of parallel licensing issues.

[CFD Newbie]

Originally Posted by cfdnewbie

This question is somewhat vague, are you referring to the equations themselves or the codes based on them?
As long as the continuum assumption holds, NS is as good as LB, since NS is essentially derived from LB for a very small Knudsen number.

Thanks cfdnewbie - I guess I unintentionally copied your name
I was referring to the codes based on the equations - sorry for the confusion. The company claims that their LBE code is more accurate and reliable in simulating air flow for vehicles than traditional methods

Just from my experience with industry: not only Audi but also the biggest suppliers of these companies (companies even bigger than Audi) are shifting from commercial to open software - mainly because of parallel licensing issues.

Thank you. I am not familiar with parallel licensing and would be grateful if you could elaborate on that a little bit if possible - why are companies moving towards opensource and why not just use traditional solutions such as fluent? Is Powerflow just a 'nice to have' that these vehicle companies use in addition to software they were using prior to powerflow or did powerflow replace fluent? Could you see other major car companies go the same route as Audi?

Many thanks!

[julien.decharentenay]

Interesting discussion.

I am under the impression that most major automotive company uses more than 1 CFD software. This would (in my opinion) due to the large variety of application (external aerodynamic, thermal management, combustion, exhaust, noise, etc) as well as historical reasons.

[CFD Newbie]

Thanks Julien.

Would you happen to have first/second hand experience using Powerflow or other aerodynamic simulation software in your experience? How do they compare?

Thank you!

[RodriguezFatz]

Solving Boltzmann equation is (physically) only better than the fluid equations, if you don't expect Maxwellian distribution functions for the different particles. Since this assumption (Maxwellian) is often even justified in nearly vacuum applications (e.g. plasma), I don't get the point why you should not use fluid equations under atmospheric pressure.

The only reasons I can think about is, that Boltzmann equation is numerically i) faster to solve and ii) more stable.
I've got the feeling it is not i), so the reasonable range of applications for Boltzmann solvers seem to be restricted to ii).

Feel free to convince me of something else.

[CFD Newbie]

You are probably right but I would defer to experts on this.
do you have experience using such software? How do they compare?

Thank you.

[RodriguezFatz]

No I don't... but in plasma simulation.

[cfdnewbie]

Originally Posted by RodriguezFatz

The only reasons I can think about is, that Boltzmann equation is numerically i) faster to solve and ii) more stable.
I've got the feeling it is not i), so the reasonable range of applications for Boltzmann solvers seem to be restricted to ii).

Could you elaborate on both points a little bit, please? Especially, why would you say that they are more stable? you mean in terms of physics, or numerically?

Thanks!

[RodriguezFatz]

These were just some ideas, what could be... I have no clue if they are correct or not. And I ment numerically.

[arjun]

Originally Posted by RodriguezFatz

The only reasons I can think about is, that Boltzmann equation is numerically i) faster to solve and ii) more stable.
I've got the feeling it is not i), so the reasonable range of applications for Boltzmann solvers seem to be restricted to ii).

Feel free to convince me of something else.

The classic lattice boltzman may indeed be much faster than the traditional navier stokes solver that you see. (It also depends on which Navier Stokes Solver you are talking about, those who use fft for pressure equation could compete very well).

I am not convinced about the (ii) that is about LBM being more stable. As someone who is implementing LBM , I see LBM being many time more unstable than the Navier Stokes solvers I have written and came across.

Not sure what Exa does for their implementation but so far I am not able to get myself convinced that it could be as stable as SIMPLE .

Further I have been working on finite volume lattice boltzmann and it is not only slow it is also very unstable.

[RodriguezFatz]

Alright, so why should someone use it for "common" cases (not some special things)?

[arjun]

Originally Posted by RodriguezFatz

Alright, so why should someone use it for "common" cases (not some special things)?

What do you mean by special things? I have problems even with simple things like flow around cyliners even at moderate Reynolds numbers of say 1000 to 10000 or so.

[RodriguezFatz]

Well, I thought there might be some special cases, where you get fast and simple solutions... your respond sounds like there aren't.

[arjun]

Originally Posted by RodriguezFatz

Well, I thought there might be some special cases, where you get fast and simple solutions... your respond sounds like there aren't.

There are, at least Exa is able to get them.

The point is I am not able to get it and very likely that a normal joe would also not be able to get.

Anyway I am working on it and see if it could be applied to general polyhedrals like finite volume solvers. Even though I have working code on polyhedrals, so far stablity is one reason I would not use it with confidence i could use say fluent or starccm.
I will try and see if it could be made as stable as our navier stokes solvers. Who knows what future holds.

[kyle]

I've been evaluating switching to an LBM code for a heavy truck aerodynamics application after two years with Star-CCM+. I'm dealing with highly separated flow where steady-state approximations are usually not very helpful, and traditional DES/LES simulations are prohibitively computationally expensive at the geometric detail that I require. LBM is certainly less accurate for predicting skin friction or heat transfer, but I am largely interested in measuring pressure drag caused by large unsteady vortices buffeting against downstream geometry.

I'm hoping LBM can give me an "acceptable" LES simulation faster, even if the skin friction is slightly less accurate than what I would get with a traditional mesh and solver.

[Paulh]

We've found that both CCM+ and PowerFlow significantly underpredict the aerodynamic drag of heavy vehicles.

[CFD Newbie]

Hi Paul,

What software do you use for heavy vehicles then?

Thanks

[Totalsim]

Interesting discussion.
I believe the main advantage is the speed of transient simulations.

I've never done a back to back but I should think there are limitations as I've not heard of F1 clients using it.

FYI - we use DES within OpenFOAM for our transient simulations. For certain bluff applications its necessary for the level of accuracy you need. http://www.youtube.com/watch?v=TXMPE5mtXcw

[Paulh]

We use both STAR-CCM+ & PowerFlow.