things about the coupling

Submitted by stefanb on Wed, 07/06/2011 - 14:36

Hi everyone,

I've got some questions about the coupling.

1. If a density field is provided, the drag force is calculated in the correct way. What about the KSL-values? I haven't found anything about the density in the code and if I look at some references about KSL the equations to calculate it seems to be different. So maybe anyone can give me a hint where to find a description about the used KSL model (I think in a further version the Gidaspow-model was mentioned?).

2. Furthermore, is there any other reason than stability for the limitation of the KSL-value?
For my case, the KSL-limit is reached often and so I increased it without any problems.

3. Does anyone know a drag model on rotation for lots of particles? I know some but only for single particles. Or do you think it isn't necessary to keep the rotation and the resulting torque in mind?

4. What do you think of the used schemes? I prefer a combination of the GAMG solver for pressure and a smoothed solver for the other variables. This combination (which is similar to the schemes of e.g. CFX) seems to be more robust and faster if higher accuracy is needed. Which are your preferred schemes (I know that there is no general recomendation)? Or has anybody observed different behaviour?

5. Can anyone offer some literature for the virtual mass force? Of course, I can derive the equations from the source code but a more detailed look on it would be great.

Thanks in advance,
Stefan

Forums:

cgoniva's picture

cgoniva | Wed, 07/06/2011 - 16:10

Hi Stefan,

ad1) the current solver is designed for incompressible flow with constant rho. If you write a new compressible solver (or one which hast non-constant rho) you might have another look at Ksl (have not done that yet).
Currently Ksl is calculated as a sum of particle forces (Newton) / (relative velocity*cellVolume) = (N/m3)/(m/s) = kg/(m3*s)
In the solver it is used as implicit momentum source: Ksl/rho*Urelative = (m/s2) which is the unit of UEqn.

ad2)you can remove the limit. it is a relic which i recovered and removed a week ago on my version. You can simply overrule it by defining you own KslLimit in the dict.

ad 3)what exactly do you mean - magnus force? That would be doable, but transferring torque to the flow is IMHO only useful for resolved CFD-DEM?

ad4)Thats an interesting topic - I have not had the time to deal with it in detail. Though it would be nice to have a test-case comparing different solver settings in terms of robustness accuracy and speed. Any collaboration is highly appreciated.

ad5)A manual of a commercial CFD package might help (F*****). That's where I had a look when implementing

Cheers,
Chris

ckloss's picture

ckloss | Wed, 07/06/2011 - 16:25

>>3. Does anyone know a drag model on rotation for lots of particles?
There are papers by Cleary regarding fluid-included torque if that is what you mean, but I am not sure if there is a voidage closure in it- I think not
eg Chemical Engineering Science, Volume 66, Issue 3, 1 February 2011, Pages 231-240
But if this is relevant or not depends on which medium you are looking at.

Christoph

cgoniva's picture

cgoniva | Wed, 07/06/2011 - 17:07

3. Does anyone know a drag model on rotation for lots of particles? I know some but only for single particles. Or do you think it isn't necessary to keep the rotation and the resulting torque in mind?

Chemical Engineering Science, Volume 66, Issue 3, 1 February 2011, Pages 231-240
This accounts for momentum on a rotating particle due to flow. I do not think that the angular momentum is transferred from particles to fluid.

cheers Chris

stefanb | Thu, 07/07/2011 - 13:58

Thanks for the hints - helped a lot.

3) I just realised that for a dilute flow of particles the rotational kinetic energy is constant as long as no collisions occur (well, from the source it should be that way). In my case only dense flows are investigated so I think it is not necessary to have a look at an angular momentum transfer - but it might be useful to proof if it is really negligible. That's why I asked.

5) From the manual I conclude that the virtual mass is only needed if the density of the dispersed phase is much lower than of the continuous phase. Due to this I would concern that for grinding beads in a fluid the virtual mass could be neglected.

4) I have only experience with the schemes for pure CFD. In my actual case the fluid flow is not the important thing and, also, the CFD does not cost much computational time. So, I haven't got an idea how I could measure the influence of the schemes in my case. Furthermore, I am not sure which case would be the best to test it?
The only thing is that the combination of GAMG and Gauss-Seidel smoothed solver is more stable and faster with higher accuracy for single phase flows in a wide range of application I already tried.

Kind regards,
Stefan