Test-case-1.mfx (12.5 KB)
TEST-CASE-1.msh (1.9 MB)
Hello,
I tried to build a simple case in MFiX. The solid particles are supposed to move below and settle in the bottom. But the movie shows that the particles almost disappear at the end. Kindly have a look at the attached files and let me know where I made wrong. I followed the tutorials
3.6. Three-dimensional DEM hopper — MFiX 24.2.3 documentation
I look forward to hearing from you where I am doing wrong in setting up the model.
Then I need to set up flow through this funnel whose .STL file is attached.
I look forward to hearing from you.
Thanks and Regards,
Debashis
GEL et al Funnel.STL (62.6 KB)
Hello Dr. Fullmer,
Kindly suggest where I did wrong in setting this up.
After this, I need to set up this model
This will be based on the attached paper
Gel-et-al-2018-MFiX-V&V.pdf (3.6 MB)
I look forward to your suggestions.
Thanks, and Best Regards,
Sincerely,
Debashis
Seeing as how I just simulated a hopper recently with MFIX-Exa (see here) and in the spirit of college basketball’s open transfer portal season, I thought I’d mock this up real quick to see if it might entice you to switch teams. Here are the inputs I came up with after skimming the paper:
inputs.txt (4.7 KB)
geom.csg (1.0 KB)
ascent_actions.yaml (1.2 KB)
This hits just over the 75k particles Avinash used. Of course, he ran this on the original Joule back in 2017 or earlier… but still, he reported then
the time-to-solution can be brought down to a wall-clock time equivalent of 2 days for a hopper
system having 75,000 particles on 16 cores.
I let it rip on one H100 
and it ran the 10s simulation time in under half an hour 
. I put monitors for the mass above and below the stem,
top_mass.csv (14.7 KB)
bot_mass.csv (14.8 KB)
stem_flow_rate.csv (17.1 KB)
which look like this:
I also put a mass flow monitor at the bottom of the stem. If you compare that to a simple center difference of the total mass monitors, you get a good match but the time-step-wise flow rate obviously has a lot more noise, which is expected, I think.
I also tried to animate this w/ the ascent file above:
This looks incredibly fishy. The particles are supposed to be colored by velocity magnitude ranging from 0 to 1 m/s. (@RobertoPorcu I ran this on the 25.04 release I built to test today, I wonder if something is off being passed over to ascent due to your particle container trimming? Maybe? Hopefully?)
Finally, I’ll note that if you don’t use the tangential history term in the DEM model, this runs in like 12 mins flat on one H100. And the results are nearly indistinguishable too (
@jmusser). They are the blue/orange/cyan colors that pair with the black/red/green colors in the figures above. In all cases the discharge rate is about 2.4 g/s. Do the authors say what the measured value was for this case? I don’t see it, but now I want to know if I’m in the ballpark or not.
Hello Dr. Dietiker,
Thank you.
I will implement your suggested changes and let you know.
I am a new user of MFiX and I plan to use it for simulating Lunar Regolith flow and flow in Dry Powder Inhalers (DPI). Currently I am trying to use it for lunar regolith flows.
Gel-et-al-2018-MFiX-V&V.pdf (3.6 MB)
I am trying to build a model for this attached case. I have also attached a paper.
I am curious to know if there is already any existing MFiX model file for this that I can use to start with.
Kindly let me know.
I use MFiX-DEM on a WINDOWS cluster.
I look forward to hearing from you.
Thanks and Best Regards,
Sincerely,
Debashis
Hello Dr. Fullmer,
Thank you very much for your reply. The animation looks very good. But I need to calculate the angle of repose of the piling. It seems like they are all settling down and flattening out and no conical shape as shown in the paper.
However, I do not have MFiX-EXA. Seems like I need to have a Linux HPC cluster for MfiX-Exa. I currently run my simulations using MFiX-DEM on a Windows 11 Cluster. It runs well.
Can I install Mfix-EXA on Windows cluster ?. Please let me know.
It is possible to set up this conical hopper in Mfix-DEM ? That will be very helpful.
I need to simulate this case and also the case explained in this attached paper.
Otto-et-al-Regolith-Flow-2018.pdf (3.9 MB)
I am building the solidworks model for the geometry and then will try to set up the model.
Please let me know if these can be done in MFiX-DEM or I need MFiX-Exa.
I look forward to hearing from you.
Thanks and Best Regards,
Sincerely,
Debashis
MFIX-Exa is an HPC code. But you can install it on a Windows machine using WSL and following my ubuntu instructions. Not sure how that would work on an Windows cluster. Frankly “windows cluster” sounds like an oxymoron to me.
Fishy particle velocity was indeed an ascent indexing bug due to changes in the particle container. (thanks @RobertoPorcu).
BTW, I see at the end of this paper:
The next step in the process would involve performing the experiments to identify important control variables affecting the QoIs.
I guess those experiments never happened?? Incredibly unsatisfying not knowing if this discharge rate prediction is accurate or not.
