hi,

Is CFX good enough for LES? or is fluent better in terms of memory capability and speed?

i heard passing remarks from collegues that fluent was better but nobody was able to substantiate..!!

can any one tell me whether or not CFX is good enough??

Hi,

Neither codes are ideal for LES as, in general, dedicated LES codes are explicit solvers whereas CFX and Fluent are both implicit. As LES uses very small timesteps then you are usually in the region of stability for explicit solvers and they are much faster and lower memory than implicit solvers.

CFX has the SAS turbulence model which can be very useful in some types of the LES simulations, eg bluff bodies. Check it out and see if it is useful for your case.

CFX is likely to use more memory than fluent as it is a coupled solver versus a segregated solver. Whether CFX or fluent is faster for your simulation is hard to say - the only thing for sure is neither codes are ideal.

Glenn Horrocks

hello

glenn ... one sales man from ansys tell me that cfx and fluent are both explicit codes..... and also star cd ...

are you sure about that?

best regards hayate

Yes, Glenn is correct.

Hmm,

Glenn, usually your answers are so well thought out....

Stability for explicit solvers with advection terms only is one thing. Stable courant number scales with 1/dx, where dx is the smallest mesh length scale.

The problem is that stability when including diffusion terms, which is most LES calculations, is much more restrictive than advection. The stable courant # scales with 1/dx^2.

Implicit solvers are insensitive, or possibly far less sensitive to this numerical fact.

So, explicit is not necessarily the best solution. It's probably fine if you run a uniform mesh.

Explicit will kill you thought when:

- you refine the mesh. The required timestep goes down as well 1/dx^2. So, if you make the mesh a factor of 2 finer you need a timestep a factor of 4 smaller, making your calculation a factor of 4 more costly.

- real industrial case meshes are almost never nice and uniform. Mesh spacing varies all over the place so the smallest cell sets the time step requirements.

So, for all intensive purposes implicit solvers are the only solution for practial LES calculations. For academic cases like flow around a cylinder or other simple geometries you can probably get away with an explicit solver.

[derz]

Don't forget, there is also implicit LES and explicit LES, which is different from implicit and explicit solving.

Implicit LES, like that used in CFX, implicitly filters the field via a simple spatial average which occurs during the discretization (which I believe ends up as a top-hat filter), while normal explicit LES applies actual filter functions to remove errors at higher wavenumbers...

Explicit filtering opens a can of worms, as the more things you bring into the solution, the more interplay there is between everything, and more likely things can go wrong (despite having the potential to be more accurate). Thus, thankfully, CFX and most commercial solvers stick with ILES, which is more robust.

[evcelica]

Originally Posted by Omer
;82705

Yes, Glenn is correct.

Yes, Glenn is ALWAYS correct.

[ghorrocks]

I tell my wife that and she does not believe me. I don't understand why.....

I guess it is because a good LES model does not get the dishes washed any faster (or more accurately).