Dear all,
I want to simulate a large particle moving through a (cylindrical) tube. However, as a cylindrical tube uses a non-uniform mesh, the CFD-DEM coupling does not work properly.
In my case (see attachments) I use the following setup. In the center of the tube I use hexahedron (cubical) cells and near the wall I use cylindrical-like cells. Now two scenarios occur when I release a particle in the tube:
1) When released in the center, the CFDEM code runs smoothly
2) When released off-centered, including both cylindrical cells and hexahedron cells OR only cylindrical cells, the CFDEM code aborts quickly with the following error:
=== cfdemCloudIB::getDEMdata() === particle rotation not considered in CFD
nr particles = 1
[tud206828:16593] *** An error occurred in MPI_Waitall
[tud206828:16593] *** on communicator MPI_COMM_WORLD
[tud206828:16593] *** MPI_ERR_TRUNCATE: message truncated
[tud206828:16593] *** MPI_ERRORS_ARE_FATAL (your MPI job will now abort)
I was wondering whether anyone has seen this before and if you would know of a way how to solve this?
Best,
Boris
| Attachment | Size |
|---|---|
 cross-section.png | 34.27 KB |
| sideview.png | 14.14 KB |
| cross_section_plus_particle.png | 66.47 KB |
| cross_section_plus_particle_centered.png | 68.45 KB |
bnaar | Wed, 03/30/2016 - 15:25
Specified the problem
Dear all,
After some research I found that the error occurs in following lines of "/derived/cfdemCloudIB.C":
// update voidFractionField
alpha.internalField() = voidFractionM().voidFractionNext().internalField(); // there might be a better approach, see cfdemCloud.C
alpha.correctBoundaryConditions();
More specifically, the error occurs at the "alpha.correctBoundaryConditions()" statement.
I hope this helps giving feedback to my problem.
Kind regards,
Boris Naar
alice | Wed, 03/30/2016 - 16:15
Hello Boris,
Hello Boris,
you already identified an important point in the first line of your post: the model currently used for determining the voidfraction field is based on the assumption that the mesh consists of uniform hex cells. If you want to apply the solver to a non uniform mesh, this has to be accounted for in the voidfraction model.
Best regards,
Alice
bnaar | Wed, 03/30/2016 - 16:30
Dear Alice
Dear Alice,
Thank you for your response! Would you have any suggestions on which part of the code (within the IBVoidFraction.C model I assume) I can best change? I will then attempt to improve the code for non-uniform meshes.
Kind regards,
Boris
alice | Mon, 04/11/2016 - 17:04
Hello Boris,
Hello Boris,
the part where uniform cells are assumed starts at line 139 within the file IBVoidFraction.C. For each vertex point 1/8 of the volume is regarded and weighted according to the "degree of coveredness" of the line between cell centre an vertex. (This turned out to be cheap and sufficient for our applications at the time being, but causes troubles in your case).
Best regards,
Alice