[Creg]

Whilst running multiple unsteady simulations recently, for some reason my solution files show no signs of unsteady behaviour, despite what I would think are appropriate settings in my cfg file.


To double check this I ran the SU2 tutorial for the unsteady naca0012 from the provided restart file and the flow did indeed appear unsteady in the following steps up until convergence. However, when running the exact same config file but with no restart the flow solution never shows any signs of unsteady flow up until convergence.


I have reinstalled SU2 and rerun said cases but this does not seem to fix the problem.



Does anybody have any clue as to what the problem could be?

[TKatt]

Hi Creg,

I have similar issues and unfortunately no good solution. But some questions which might help corner the issue (which I am doin rn):

1. In the simulation from dreestream conditions: Do the residuals stall after a while i.e. all residuals and output quantities are exactly(!) the same over inner iterations and time iterations.
2. Running a steady simulation until some sort of convergence and using that as a restart solution worked for me

3. While at steady simulations: Increasing the timestep to very large values should result in the same as the steady sim BUT I noticed when lowering the timestep then to the original one there is a (distinct) point where the result switches back to a bit different one.


I am a bit unsure whether I am facing a physical or numerical "feature" or if there is a bug inbetween.

Best, Tobi

[TKatt]

Can you please check whether you use FGMRES as LINEAR_SOLVER?

Please check if the linear solver converges by adding LINSOL to HISTORY_OUTPUT. If LinSolRes is 0 at some point that indicates that there was no solution update (and with that where will be no transient behavior ). If that is the case please increase the LINEAR_SOLVER_ITER. In your case that is currently 5 so check if 20 or so works better. If you then look at LinSolRes again that should not be zero and ideally below that


After that you can still restart to use your preferred config options. The initial transient seems to be tough on the solver.



Let me know if that was helpful



Best Tobias

[raviramesh10]

Dear TKatt and Craig,

I had similar issues when running a DDES 2nd order simulation as well, and the solution suggested by TKatt, i.e., using a converged 2nd order steady state simulation, worked in my case. You could also try using the LINEAR solver, as that would be a good estimate as to what exactly is going wrong from a numerical perspective.

Alternately, depending on the approach that you are using to resolve the flow field (LES/URANS/DDES), you can vary the numerical limiters as well, so that the solution remains stable for a (relatively) shorter period of time.

Hope this helps. Good luck with your simulation.

Ravi

[jdp810]

Are you sure that you should see vortices?

From the config file I see your marker is a cylinder. If it is a cylinder, the Reynolds number you are setting do not produce vortices. At least I think that at Re > 400 the vortices disappear.

[bigfootedrockmidget]

Originally Posted by jdp810

Are you sure that you should see vortices?

From the config file I see your marker is a cylinder. If it is a cylinder, the Reynolds number you are setting do not produce vortices. At least I think that at Re > 400 the vortices disappear.

No, there is a Hopf bifurcation at Re~50 and a second longitudinal frequency appears at Re=180 and then a third at Re>230. You will then get more and more frequencies until you have a continuous spectrum at around Re>1500. The von Karman street will still be visible at Re=400, just with more frequencies compared to e.g. the Re=100 case.

[jdp810]

Well, I thought that from 400 on you won't see them, but it's good to know what's the actual number I know little about that topic, do you have a good paper/study that talks more about the Hopf bifurcation and cylinders?

Anyway, the Reynolds number of your simulation is 1.2e6, so there's no chance to see any vortices...

[bigfootedrockmidget]

Here are some nice recent simulations of the turbulent vortex shedding in the regime Re =2.5e5 - 8e5:

https://www.sciencedirect.com/scienc...15000557#f0080