Hello all,
I am getting considerable difference in results between resolved CFDEM and resolved with dynamic mesh refinement. My case as shown in is particles flowing through a slot/constriction. The number of particles left in the domain is plotted against time for the two methods in . Both the simulations are identical (with same random seed and initial position of particles) except that one is resolved with static mesh and the other is resolved with dynamic mesh.
I have tried 2 mesh sizes for the resolved with static mesh (4 cells per particle dia and 8 cells per particle dia). I am not able to understand why should there be such remarkable difference in the results as I was expecting them to converge. However, in the static mesh the particles bridge sooner but in the dynamic mesh they do not bridge at all.
Is there any limitation to using Resolved with IB for particle-particle bridging type simulations? It will be great to know your suggestions/comments!
Cheers,
Som
| Attachment | Size |
|---|---|
 Simulation case | 145.9 KB |
| results.jpg | 37 KB |
cgoniva | Thu, 02/07/2013 - 17:21
hi!
hi!
do you use the latest version with parallel IB?
which voidfraction model do you use - are the particles captured correctly on the fine mesh?
Cheers,
Chris
Somnath | Tue, 02/12/2013 - 04:18
Hi Chris,
Hi Chris,
Thanks for your reply. To answer your questions: yes, I was using the (then) latest coupling version 2.4.4 and voidfraction IB.
However, the problem I figured out was with the Courant number. Initially, I was just making sure that the Co<1 cocnsistently throughout the simulations. However, for the resolved cases, I saw that a Co < 0.5 is necessary to get converged results for the static cases (I tried 4 cells/dia, 8 cells/dia and 10 cells/dia for each 5%, 10% and 20% particle concentrations respectively). For the dynamic meshes at 5% and 10% particle concentrations, Co<0.5 gives converged solutions but at 20%, even lower Co is required. I am not sure why this is but am guessing it may be because of very fine meshes being generated at particle interfaces at high concentration.
Another observation is that, to simulate bridging concentration in dense suspensions, as is my case, 4cells/dia seems to be giving the same results as 8 cells/dia.
Thanks and would love to hear if you have any comments!
Regards,
Som
cgoniva | Thu, 02/21/2013 - 09:18
Hi,
Hi,
>>as is my case, 4cells/dia seems to be giving the same results as 8 cells/dia.
do you mean 4 un-refined cells, which are then locally adapted? (otherwise I'd guess the mesh is too coarse)
Cheers,
Chris