Dear Expert,
I am simulating the TFM of 3D geometry, but a floating-point error is coming and the solution is interrupted after some time. I solve the same problem in the 2D case but in the 3D case is not stimulating. please help us.
File size is too large so it is not uploading.
I attached the mfix file. please see it where I’m wrong, I also created the solver to simulate it.
lungu3D.mfx (11.4 KB)
When you say “File is too large”, did you use “Create bug report” or did you zip up the entire directory?
If you used “Submit bug report”, can you tell me what files are in the “too big” bug report so we can fix this? (Run unzip -l on the zip file and paste the file names)
The bug report is big around 160 mb, and MFix forum gives a limit of up to 100 MB file to upload here, how can I upload files here?
in the Bug Zip file Entir_domain.vtu 2653 KB, 3D.RES 65623 KB, and 3D_MESH.vtu 15342 KB. these three files are too large apart from that.
Ok, maybe the bug report script should exclude RES and vtu/vtp files. I will make that change. For now, the .mfix file should be enough. I’ll take a look when I get a chance. (There have been a lot of support requests lately so it might be a little while)
I did not get a floating-point error running this case but I did get a convergence failure
DT < DT_MIN. Recovery not possible!
Yes, then what should I do, please suggest? We need to reduce time step size or fine mesh is required?
Your initial condition is fully packed. It may run better if you start below maximum packing. It also may run better if you use ideal gas law and if you have non-zero initial granular temperature values. Or even use the algebraic model.
Dear Jeff.
I am simulating the 3D case but it takes a very long time to complete, for just a 6-second simulation it takes more than one week however in a 2D case it is completed in 6 hours. could you please tell me what should i do, can you check it on your end?
lungu3D.mfx (10.9 KB)
You should expect a 3D case to take longer to run than a 2D case. You most likely need to either get access to a cluster so you can run in parallel or coarsen the mesh.
I have a decent capacity system with 128 GB RAM, 8 GB Nvidia graphics, 12th Gen Intel(R) Core™ i9-12900, 2400 MHz, 16 core, 24 logical processors. Still, 3D simulation has been slow. Whereas the mesh size of 3D is 240000. Can you tell me why this is happening?
I don’t have sufficient information to provide a valuable answer. Are you actually using all 16 cores? what parallel partition did you use? What is the grid spacing in 2D vs 3D? What is the time step the 2D and 3D simulation run at (actual time step, not the initial DT you set)? Is the one week time for 3D the actual time it took or the estimate based on current state?
In 2d total mesh size is 6000 and 3d mesh is 240000, i did not how many core is used by mfix, the time step size is same in both the cases 0.0001 and 3d case already taken more then one week to simulate 8 sec.
If you went from 6K cells in 2D to 240K cells in 3D, than a 40-fold increase. If it took 6 hours in 2D, it should take 10 days in 3D, assuming everything else is the same.
I am guessing you are only using 1 core (serial). You either need to run this simulation in parallel or coarsen the mesh if you want to run in serial.