Determination of physical timescale
I am facing some convergence issues with my steady state simulations. In order to get convergence, I followed the steps in CFX manual and at the link:
and used both Autotimescale and physical timescale options. I reduced timescale factor for the Autotimescale option and the actual timescale for the physical timescale option and got convergence by both.
However, while trying to estimate the physical timescale, I followed the steps in the above link and plotted the streamlines with time as the variable as shown in the attached figure. The fluid residence time estimate from streamline seems to be order of 1e-1. However, to get convergence, I had to reduce the timescale to as low as 1e-4 i.e. three orders of magnitude lower.
So is the fluid residence time not a good estimate of the physical timescale?
The fluid time scales is just an estimate of what is required for convergence. Some need more, some less. The time scale required for convergence depends on numerical stability factors.
When you cannot run at the fluid time scale it is a sign of a numerically unstable simulation. Try to improve mesh quality, check your physics is correct, use double precision numerics or use a better initial condition.
Thanks Glen for your reply.
|All times are GMT -4. The time now is 18:00.|