[quarkz]

Hi,

I am simulating hypersonic double cone external flow k-omega SST which diverges after 5000+ time steps to ~ 1e-6.

It worked when my y+ = 50. Now y+ = 1.

May I know what's the likely problem?

I tried a few things:

1. change from vanleer to upwind in the interpolation scheme - stability greatly improves but still diverges around 0.005. Not sure why since from the contour plots it seems to have reached steady state. However, being a 1st order mtd, the wall heat transfer does not compare well with expt.

2. changing from bounded Gauss limitedLinear to bounded Gauss upwind - slight improvement.

3. using wall function for k, nut - slight improvement but I shouldn't be using any wall function, is that so? Since y+ = 1.

4. changing from smoothSolver to PCG or BICCG - slight improvement

I also have another qn. For omega, should I use omegaWallFunction or fixedValue? I saw somewhere in the forum that I must always use omegaWallFunction. Is that so?

I hope to use a 2nd order interpolation scheme. I'm thinking of Gamma. Is there other recommendations?

Thanks

[hyFoam]

Hi,


- Is your temperature field bounded between [T_min; T_max]?


- If you run the exact same simulation but with a laminar setup, does it crash as well?



Thanks

[quarkz]

Hi hyFoam,

If I use the standard rhoCentralFoam, I can only reach 1.74279e-05, with initial dt = 1.e-10. Now I managed to get it to run longer, maybe up to 2.e-3 by changing the initial internal velocity to half of the inlet velocity, and also doubling the grid in the y direction.

For hy2Foam, I turned off most setting because the enthaply is not that high. I do get temperature out of range:

^[[1;31mAttempt to use rho2ReactionThermo out of temperature range 838 times during this iteration.
^[[1;31m Thigh: 10000 < 6e+06^[[0m

but I managed to run to 2e-4. Similarly, doing the 2 changes mentioned above I managed to run longer, similar to the above.

I am not sure why it still diverges after 2e-3, because contour plots show that it seems to have stablized.

I didn't run the laminar case because previously it gave very unsteady and incorrect results.

Thanks anyway.