[decenu]

Hi all,

I am working on relatively simple simulations using the Incompressible RANS solver paired with SST turbulence modelling.

I am trying to understand how the CFL number setting affects the results. Whilst I understand that the physical phenomenon should be steady state, SU2 runs a pseudo time iteration.

My questions are:
1. Can the CFL number cause divergence of a solution?
2. Can the CFL number directly affect the solution (convergence, accuracy and result)?
3. How does the adaptive CFL number change?
4. Any good practices for setting the CFL number?

It would be greatly appreciated if anyone could help out with these, in order to decrease my fear of solution variation/error/uncertainty in my calculations.

Thank you in advance!

[bigfootedrockmidget]

1. Can the CFL number cause divergence of a solution?

Yes. For steady state problems, you use a pseudo-transient approach in SU2. You are solving a time-dependent problem but SU2 uses a timestep that is large and can be different in every cell. The CFL number is a scaling of the (local) timestep computed by SU2. If it is too large, you get divergence.

2. Can the CFL number directly affect the solution (convergence, accuracy and result)?

If the problem has only 1 solution for your initial and boundary conditions, then you have found the numerical approximation of this solution when the solver converges, this is independent of the CFL.

3. How does the adaptive CFL number change?

You can read the description of all configuration options in the config_template.cfg

4. Any good practices for setting the CFL number?

SU2 will compute a large, but safe value of the time step. If SU2 did a 'perfect' job, then the CFL (for your incompressible steady problem) should be 1.0. If you have no clue, then start with 1.