Having trouble with starting mfix exa

How can I set the boundary condition for solid inflow?

from (now deleted) second post…

Hi, I’m trying my best to use mfix-exa. But I have 2 questions.

  1. I runned the Li-Janssen simulation but the UDF code does not work at all. There was no reaction. What should I do to make the reaction happen?
  2. Is it possible to do the solids inflow for boundary condition? There are no informations in the document.

Thank you always.

@JunsungYeo you don’t need to make a new post to ask the same questions. I saw the old post, but I had a deliverable due 8/1 that consumed my time and then I had to prioritize other things I had neglected, such as my own student. For 1) if it didn’t work, then you didn’t compile it into the code. One thing you can do is overwrite the source file to make sure that it’s being compiled in. You shouldn’t have to do this, but I would just to be sure. The file to overwrite is in mfix/src/usr/. For 2) yes you can, but I don’t think I have anything off hand to share. I’ll need to set something up and test it.

Sorry that there is not much for you to work off of, but it’s a new code, we’re just getting started on user support items/features. On my TODO list is a PIC manual that someone like you could follow along with. I just haven’t had the time/haven’t prioritized it yet. You just have to realize that there will be some growing pains as an early adopter of a new code with limited support. If that doesn’t fit your timeline/schedule, you might want to consider sticking with Classic for the time being.

We do have a monthly users meeting on Mondays at 11 EST. I’m guessing that’s a terrible time for you, but if you would like to attend and ask the developers questions directly, we can share that meeting link.

1 Like

@JunsungYeo sorry for the delay… I set something up last week that had a PIC parcel mass inflow and worked as intended. the inputs are attached for your reference.
inputs.txt (17.2 KB)

1 Like

Thank you so much.
I have another question. Is it possible to set the thermal conductivity as a temperature dependent?
Thanks again for your support.

currently it is just constant. do you have a need for T-dependent conductivity? do you have a model for that? it is something we could consider for future development if there is a need for it. if you can give modeling references maybe @rangarad could put in an issue for you. (that’s no guarantee that it would be implemented, but it gives us something to track.)

1 Like

Hi.
What is the unit of homogeneous and heterogeneous reaction rate in mfix_usr_reactions_rates_K.H code?
Is it mol/m^3 s & mol/s each?

I’m pretty sure it’s (mol/s) for both. Maybe @RobertoPorcu could confirm.

1 Like

Hi,
it would be mol/(s m^3) for homogeneous reactions, and mol/s for heterogeneous reactions.
Note that we have recently made a change in the reaction type nomenclature. We do not call the reactions as homogeneous or heterogeneous anymore, but we define them as “eulerian” or “lagrangian”. Eulerian reactions are fluid to fluid phase homogeneous reactions and have units of mol/(s m^3). Lagrangian reactions can be either fluid<–>solids heterogeneous reactions or solid to solid phase homogeneous reactions that happen inside a single particle and have units of mol/s.

2 Likes

But does the code for mfix exa work like that?
Recently, I made Eulerian equation using unit of mol/(m^3 s).
With it, the result of reactions was extremely fast, and the error occurred.
What I had supposed was, I tried to unify the unit with Lagrangian reaction rate which is mol/s, by multiplying cell volume.

Please review this and give me your answer about it.

Yes, MFIX-Exa works like that.
mol/(m^3 s) for Eulerian reactions
mol/s for Lagrangian reactions
How large is the timestep size? Did you try with a smaller timestep?

1 Like

Yes I tried with the smaller timestep but it was not the matter.
The Eulerian reaction
CH4 + H2O → CO + 3H2
was the problem.
The molar fraction of CO and H2 exploded from almost zero to 25% and 75% just in 0.01 sec.
So I fixed the code (multiplying cell volume) and then the result seems to be reasonable.

I see. Just to double-check, are you defining the eulerian reactions in the “EulerianReactionRates” class and the lagrangian ones in the “LagrangianReactionRates” class in mfix_usr_reactions_rates_K.H ?
The code expects the eulerian rates in mol/(m^3 s). If multiplying by cell volume is preventing the code from crashing, likely it’s because the eulerian reaction rate is too large, and multiplying by cell volume (which tipically is a very small value) reduces the eulerian reaction rate and the code is happier… but likely you’re not getting the physically correct behavior

1 Like

Thanks for your kind explanation.
Yes I sorted the reaction types.
I can upload my code and input block after weekend.
Im also struggling with other errors so it will be really helpful if you or other developer check my works.
Thank you.

Sounds good, I can give a look at your code and input block after the weekend

biomass_combustion.zip (160.8 KB)
Please check my code.
I’m not sure what’s the problem.


image
image
These are my references for the chemistry equations.

Thank you!

Hi, can I use 2 gpus for one simulation without nvlink?

we’ve scaled the code out to 65536 GPUs and run on NVIDIA, AMD and Intel GPUs (or whatever AMD and Intel want to call their accelerators)

1 Like