I tried to do a Particle Settling Simulation using Coarse Graining and a custom Particle Size Distribution. Unfortunately it doesn’t work and I’m immediately getting an error message. It says: Killed. Primary job terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.
Maybe someone could have a look? I attached my files and maybe someone can give me a hint, what’s wrong. I tried to move the psd-file into the build folder as it is shown in one of the tutorials. I also tried in the main directory. But both cases do not work.
I also have a question regarding particle size distributions in general. They can be used with coarse graining in MFiX. Which method is used by MFiX? Is the coarse grain factor constant or is the coarse grain diameter constant? Or can this be set in MFiX?
Hi @Fritzi. Sorry for the delayed response.
You seem to have attached the file geometry.stl twice. Instead of attaching individual files, you can go to the main menu, select “Submit bug report” and MFiX will create a ZIP file with all relevant inputs and diagnostic information, including logs.
I believe that geometry_0001.stl is just a copy of geometry.stl (?) so I tried copying the file and running the case. Watching top, I can see that the memory usage increases until the system runs out of available memory and the oom-killer (oom=out-of-memory) kicks in. In the system log:
[67968.234479] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=c3,mems_allowed=0,global_oom,task_memcg=/c3,task=mfixsolver_dmp:,pid=13087,uid=103
[67968.234502] Out of memory: Killed process 13087 (mfixsolver_dmp:) total-vm:17652532kB, anon-rss:12175336kB, file-rss:2560kB, shmem-rss:2176kB, UID:103 pgtables:30440kB oom_score_adj:0
[67970.392260] oom_reaper: reaped process 13087 (mfixsolver_dmp:), now anon-rss:224kB, file-rss:204kB, shmem-rss:160kB
You need to either increase the available memory (possibly using more DMP nodes) or decrease the number of particles.
It’s also possible that there is some sort of memory leak. It’s a good idea to try running jobs in serial mode (no DMP) to see if the problem is DMP-specific.
I defer to @jeff.dietiker for the question about coarse-graining.
When using a PSD with Coarse grained DEM, you can either
Set a constant statistical weight. In this case the PSD is scaled uniformly. Say the statistical weight is 8, then the CGP diameters will be twice that of the real particles (with a PSD).
Set a constant CGP diameter. In this case, the statistical weight will be different for each CGP, and the CGP will have a uniform diameter.
These 2 options are available in the Solids> materials pane:
There seem to be an issue when specifying a constant CGP diameter with a PSD. The seeding is also using the specified baseline diameter which in your case makes the lattice spacing too small. This may be why you are running out of memory. We are looking into it. Thanks for the report.
thank you very much for having a look at my case. I created a bug report and attached it to this posting. I now tried the first suggested option by Jeff and set the Statistical Weight to W=8000000 = (200³). I need to use those high Coarse-Grain-Factors (like 200) as I’m trying to simulate an industrial sized fluidized bed.
I’m now getting the warning of unsuccessful seeding and there are way less particles generated than I calculated myself. I also tried changing the initial condition from particle count to volume fraction, but it still doesn’t work.
I’m really looking forward to hearing from you.
Thank you very much in advance and best regards
Please copy the attached file to you project directory and rebuild the solver, and let us know if this helps.
Your PSD has a fairly wide range of size with a max/min ratio of 58. If you use a constant stat weight, it will be tough to get a successful seeding because the seeding lattice is based on the largest diameter.
I could get a successful seeding with a constant cgp diameter of 0.01m, which generated 1.7 Million particles. I am still working on the des grid fix with PSD. For now just set the number of cells in the des grid to be the same as the fluid mesh.
It may be beneficial to clip your custom PSD so you don’t have such a wide range of particles.
Thank you very much for your help. As soon as my current simulations are done, I will try this. But I’ve got one more question: Does MFiX require a PSD based on volume or number? I couldn’t find the answer in the documentation. Thanks in advance!
