[beluiz93]

Hi guys,

So, I'm having a really big problem with my LES (Smagorinsky model) simulation. Basically, the Courant number is getting higher and higher and I saw in paraview that there is a point around the surface of my sphere that the velocity is exploding. This happens in the very beginning of my simulation. Does anyone know what might be happening?

I'm using a relaxation Factor of 0.5 in all variables and my initial time step is around 10e-8 is order to try to get a more stable simulation.. I'm also using the pimpleFoam solver.

Any help will be very appreciated! Thanks in advance!

[Simjoh]

Are you trying to start your simulation from a zero velocity field? If so it might to initate the with a steadystate solution fom either potentalFoam or simpleFoam.


Running LES with PIMPLE is not correct in my oppinion. Try to switch to PISO and keeping you courant below 1.

[beluiz93]

Originally Posted by Simjoh

Are you trying to start your simulation from a zero velocity field? If so it might to initate the with a steadystate solution fom either potentalFoam or simpleFoam.


Running LES with PIMPLE is not correct in my oppinion. Try to switch to PISO and keeping you courant below 1.

Hello Simjoh! Thanks a lot for the reply!

So, I started the simulation with the InternalField equal to my Inlet Velocity (internalField uniform (0.146 0 0)). I saw that initiating
with the steadystate solution it might make things a bit more stable with my LES simulation. However, I'm quite new to OpenFOAM and I still don't know how to implement that in my simulation. I mean..how can I start the flow with my steadystate solution?

Sorry if this a newbie question. Thanks a lot again!

About the pimple or piso, it's possible to do LES with pimple. It's not common but it's possible. By now, I'm not concerning about that because the error might be regarding my mesh or that initial field as you said..

[anishtain4]

To get a steady simulation change your solver in system/controlDict, then change the end result directory name to 0 and change back the controlDict to PISO or whatever else you want for the LES simulation.

Simon:
Can you please share it with us why PIMPLE is not good for LES simulations?

[beluiz93]

Originally Posted by anishtain4

To get a steady simulation change your solver in system/controlDict, then change the end result directory name to 0 and change back the controlDict to PISO or whatever else you want for the LES simulation.

Simon:
Can you please share it with us why PIMPLE is not good for LES simulations?

Hello Anishtain! Thanks for the reply!

I think I got it what you mean..So I basically run my simulation in steadystate mode until lets say..100 seconds and then I pick the last time folder (100 in that case) and rename it to 0. After that I can run normally the LES simulation with pimple or piso..Am I right?

Thanks a lot!

[Simjoh]

Originally Posted by anishtain4

To get a steady simulation change your solver in system/controlDict, then change the end result directory name to 0 and change back the controlDict to PISO or whatever else you want for the LES simulation.

Simon:
Can you please share it with us why PIMPLE is not good for LES simulations?

Sorry but I don't have the books or knowledge to in detail convince you the diffrence in a theoretical way.

So I will just do some argumentation that you might or might not like.
As far as I know SIMPLE and PISO where the first algorithms around, atleast among the ones implemented in openfoam today. The SIMPLE algorithm is for stead state símulations and PISO for time dependent problems. The downside with the PISO algorithm is that it is numerically unstable and you need a short timestep in order to keep the algorithm stable.

Taking a short timestep is quite a waste for time averaged procedures such as Raynolds Avareaged Navier Stokes (RANS) or even URANS. When you are not interested in the details. Therefor the PIMPLE algorirhm was developed so that you can take a larger timestep while marching in time and still have a stable solution procedure.

For LES however you are probably interested in the details and then you would not want to smear out the result with the PIMPLE algorithm.

Since the PIMPLE and PISO algoritms are very similar you can chose to run the PIMPLE arlgortithm in PISO mode by changing the fvSolution

OuterCorrectors         2;//1 will give PISO mode

This link gives you some more details.
https://openfoamwiki.net/index.php/O...hm_in_OpenFOAM

What is your Re?
And how big is your domain?

[Simjoh]

Originally Posted by beluiz93

Hello Anishtain! Thanks for the reply!

I think I got it what you mean..So I basically run my simulation in steadystate mode until lets say..100 seconds and then I pick the last time folder (100 in that case) and rename it to 0. After that I can run normally the LES simulation with pimple or piso..Am I right?

Thanks a lot!

Doing this way is OK, but you BC will probably be overwritten. So if you have fluctutations on your inlet BC these will be frozen in time. You bprobalby would like have some fluctuations on you inlet in order to introduce some turbuelnce.

[beluiz93]

That's great Simon! And that's exactly what I'm doing. I'm running PIMPLE in PISO mode.

My Reynolds is ~10.000. And my mesh has around 2 million elements with the surface of the sphere made with prismatical elements and outside part of the sphere (the internal area of the cube) is done with tetraedical elements. The prismatical part is just because I need some important information around the sphere like the boundary layer been removed throughout time with my high Reynolds simulation and other important stuff.

About that part of getting the last time folder in the steadystate and putting as my initial field for my LES.. You're totally right. I won't have any fluctuations in the beginning.. And that's not good for LES.. Any ideas of how can I input that condition? Because my initial file will be a bunch of velocity values. I don't think I can adapt that :/..

[Simjoh]

Originally Posted by beluiz93

About that part of getting the last time folder in the steadystate and putting as my initial field for my LES.. You're totally right. I won't have any fluctuations in the beginning.. And that's not good for LES.. Any ideas of how can I input that condition? Because my initial file will be a bunch of velocity values. I don't think I can adapt that :/..

I've only implemented som fluctuations using the groovyBC using some normal distributed fluctuations. This is however not a very scientific approach.

If you have time and want to do it better then I did I suggest you to read through this PDF.

http://www.tfd.chalmers.se/~hani/kur...esentation.pdf

[beluiz93]

Got it. I never heard about that groovyBC before. Sounds interesting!

Thanks! This PDF seems nice. Although it looks like a bit of challenge to implement that vortex mode in openfoam.

[Simjoh]

In order to get groovyBC you got to install swak4Foam it is a really good package that makes a lot of the work with opefaom a bit easier.

Good luck!

[beluiz93]

Nice! I'll take a look at that in more details then. Thanks a lot man!

Cheers!