Coupling, very slow temperature convergence

This forum is dedicated to Syrthes related issues, as the Syrthes tool does not have its own forum.
Post Reply
aminer2k
Posts: 7
Joined: Tue Aug 06, 2024 11:10 pm

Coupling, very slow temperature convergence

Post by aminer2k »

Hello,
I have a CS 8.03 / Syrthes 5.0 coupled thermal problem. The coupled solution is very slow.
-About 300k cells
-The solid problem solves in ~10 seconds in Syrthes, 1 core
-The fluid problem solves in ~60 seconds in CS, 1 core
-The coupled problem, CS - Syrthes will not stabilize, after even 3+ hours (1 core, or 4 cores/4 cores) The coupling for thermal is very slow to converge. It slowly settles towards what I think is the solution

Per the tutorial, it appears that time stepping must be fixed, not adaptive. If I increase the time step, the fluid solution will diverge.

Changing CDTVAR from 1 to 10 does not change anything, I thought this was to speed up the T solution vs the fluid solver. I do not see other relaxation / or controls to speed this thermal solution up. There are options for specifying solvers for T. Any guidance for this issue?

files attached. The interface will not allow me to attach a .syd
Thank you
Attachments
setup.xml
(9.42 KiB) Downloaded 650 times
Yvan Fournier
Posts: 4164
Joined: Mon Feb 20, 2012 3:25 pm

Re: Coupling, very slow temperature convergence

Post by Yvan Fournier »

Hello,

If this should converge to a steady-state solution, the time steps do not need to be identical between the fluid and solid domains. In this case, I would recommend the pseudo-steady algorithm on the code_saturne side (with a local time step), as this will optimize for the CFL number and usually provide a good compromise between stability and speed of convergence.

Best regards,

Yvan
aminer2k
Posts: 7
Joined: Tue Aug 06, 2024 11:10 pm

Re: Coupling, very slow temperature convergence

Post by aminer2k »

Great, thank you. I will try that.

From your comments, it sounds like the total steps and step size need to be the same only for transient problems.
Much appreciated.
Post Reply