[smajens]

Hey everyone!
I am planning to study the influence of different dune shapes on wave propagation and sediment transport in a natural bathymetry. As the wave refraction on the inlet of my CFD domain is somewhat important, this seems to be the perfect usecase for HDC / initialization with FNPF. As I want to transfer the quasi-steady state to my CFD calculation, I let FNPF run for 140s simtime, using the last 200s for HDC. You will find an illustration of the case in the appendix, with the CFD subdomain highlighted by a black frame.

Issue:
As my waves are super short (~25m), the required mesh is quite fine, leading to huge result directories. Following the suggestions from a previous post, I first tried writing the state file (FNPF) for the CFD subdomain at every iteration.
As this resulted in > 1Tb of data in the state files folder, I also tried setting P 42 0.2 instead of P 41, reducing the output to several 100 Gb. Chosing much higher values seems to be problematic, as the HDC input is seemingly interpolated linearly.
Unfortunately, even with access to a HPC I am running into memory problems while trying to perform HDC with these cases. To me it seems like DIVEMesh tries to load all state files into RAM for the HDC, greatly limiting the possible case size (besides taking about as long as the full FNPF simulation).
Is there any way to further limit the state file size in FNPF aside from limiting the output domain (P 43)?
It would be great to have an option like P 185 for fsf output, where one can set an interval at which the state files are written.
You can find my case files in the appendix.


I would be happy for any ideas!

Cheers,
Jan

[valgrinda]

Hi Jens,

maybe just send me an email (hans.bihs@ntnu.no) and you can explain your case a bit more in detail there. I think there could be a different approach to this.