code_saturne User's Forum

Hi all,

I have puzzling problem. If i reduce the time step for my LES simulations the velocity diverges. If increase the time step back to its original value (2e-4), no divergence occurs and I can get past the point where the solution diverges with the smaller time step. I want to reduce the time step as the instantaneous cfl number goes above 1 (maxes out at around 2) with the original value.

I am using Code-Saturne 5.2.0 and I have attached the xml, listing files and subroutines for the simulation which diverges.

Any advice you can give would be much appreciated.

Kind regards,

Kenneth

Hello,

This is strange, but could be linked to several factors, possible the Rhie & Chow filter, or some other instability (in some LES cases adding a small amount of upwind, say 2% can help avoid some issues while not adding significant error). Also, your max CFL constraint leads to a very small mean CFL, which is less risky but can lead to such issues with the Rhie & Chow filter (which is used to avoid checkerboarding with colocated finite volumes but introduces a time-step dependent error term).

I mentioned your case to a turbulence specialist so he might take a look.

Best regards,

Yvan