[Nek5000-users] Simulation blows up after running for approx. 53500 time steps!!

[nek5000-users at lists.mcs.anl.gov]

Try to add some divergence to the outflow boundary to accelerate the flow. You'll find an example in the turbJet example (see accl_outflow). This will work well if you're right and the problem is related to a significant backflow at the outflow boundary.

Some solver performance suggestions:

- lower your pressure tolerance (e.g 1e-5 is typically strict enough)
- your timestep seems to be quite low (check the CFL-number). If the field is fully turbulent and you use IFCHAR=.false. you're stable up to CFL ~ 0.5. If you use the characteristics scheme try to use a dt such that  CFL = 2 (make sure you set COURANT in the .rea file correct).


Stefan


On Apr 19, 2010, at 2:02 PM, nek5000-users at lists.mcs.anl.gov wrote:

> Hello,
> 
> I am running a DNS simulation of backward facing simulation (BFS) at
> Re=3000. The Reynolds # is based on step height (h=1) & free stream
> velocity (ux = 1.0). The BFS geometry has 5184 elements and lx1=ly1=lz1=8.
> I have attached the *.rea and *.usr files. I am running the case on 512
> processors.
> 
> The case blows up after approx. 53500 time steps with dt=5e-4. After
> analyzing the case in VIsit, it looks as if the simulation blows up as the
> vortex reaches the outflow boundary for the first time. I think, the
> normal outflow boundary condition is not sufficient. I have attached a
> shortened version of the logfile for reference.
> 
> The movie for the simulation can be seen at the following link.
> 
> http://www2.uic.edu/~hkanch1/movie.mpeg
> 
> Any suggestions on how to fix the problem would be helpful.
> 
> Thanks for all the help.
> 
> Harish.<bfs3d.usr><logfile1>_______________________________________________
> Nek5000-users mailing list
> Nek5000-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users