Simulation error on very first iteraiton (Complex urban geometry)


#1

Hi everyone!

So I have been attempting to run this particular simulation for awhile now as you can see here and under “SST Case F Mesh 1”. Run 1 was a geometry with slightly less nodes (30 mil) and Run 2 uses the more fine mesh (40 mil). Both runs end at around the first iteration for whatever reason and I have not been able to get them started no matter what I do.

I will list down the things that I have tried but it would be important to mention that the other runs such as “SST Case F Mesh 2 (Low)” and so on have been relatively successful despite them being almost as fine as the new meshes and with a larger domain. Those other “successful” runs also have somewhere between 100-200 illegal cells whereas the new mesh has zero so yes it pretty strange that those are able to run well but the new mesh which is of better quality per say is unable to run.

The error does not give a message and I don’t seem to be able to garner anything from the error log. Initially, I assumed my custom inlet values were causing the issue but it was clearly not as when set to a normal inlet with a low velocity (10m/s) the same exact error occurs.

Things that I have done to possibly rectify the issue:

  • Made a new mesh that is less fine
  • Set the non-orthogonal corrector to 2
  • Smooth solver for the numerics
  • Cell limited gauss linear for the numerics
  • Initialization with potential flow false
  • Simple and hierarchical decomposition algorithms
  • Different end times and write intervals
  • Slip wall for the sides instead of symmetry

All the above attempted fixes all yield the same error at exactly the same time (around 100 ish minutes).

I went ahead and also re-meshed the “successful” run with the same domain size but without things like the boundary layer inflation to reduce the illegal cells but “The machine ran out of memory” despite using 32 cores.

Some further suggestions would be much appreciated!

Cheers everyone!

Regards,
Barry


#2

Hi Barry!

Well you already tried a lot of options and I could run the simulation on my SimScale account with 96 cores if you like and see if the issue persists. Let me know if you like the idea and I will give it a spin then.

Cheers!

Jousef


#3

Hey Jousef!

You could try to see if that resolves the issues. I previously had no problems with running it on 32 cores so this is rather strange.

Cheers!

Regards,
Barry


#4

Hi Barry!

96 cores did the job :slight_smile: I will share the project with you :wink: Let me know if you need any other settings and I will run those for you, no problemo!

Best,

Jousef


#7

Hey Jousef!

Thanks for the help! Looks like it is confirmed that it is a memory issue :frowning: I think the storage space may have been exceeded for the 32 GB runs now for whatever reason despite a less fine mesh.

Cheers!

Regards,
Barry


#8

Hi Barry!

We will soon post some information about the RAM allocation for instances. Already in touch with Alex (@afischer)!

Cheers!

Jousef