k omega model converge problem

[Jundi He]

Hi engineers:

I have a simple steady state simulation which is a annuli flow, with inlet on the top surface and outlet on the bottom. The outer wall is 300 degree C and the inner wall is 400 degree C. When I use k epsilon model, the result converge well in a short time. But when I switch the turbulence model to k omega SST, the result has a oscillation. This is the temperature and kinetic energy variation of the monitoring point during iteration of k omega SST. Anyone know what's the problem? Thanks!

Best Regards!
Jundi He

[Jundi He]

And this is my xml file of the code saturne case. Many thanks!

[Yvan Fournier]

Hello,

I'm not quite sure what can cause this. It would be interesting to check with version 5.0 beta, as some monor fixes to k-omega clipings were done recently (and do not appear in other versions). But no clippings on k or omega seem to appear in your "listing", and your CFL number sems OK, so if you still have the issue in version 5.0, we may try to take a look at this case.

Best regards,

Yvan

[Jundi He]

Hi:

Thank you. I solved my problem by refining the near wall mesh, it seems that k omega SST requires a fine near wall mesh. I have another problem, when I choose the turbulence model, there is 3 options of wall function, one scale and two scale wall function, what is the definition of these type of wall function? I try to google it but didn't get the result. Can I define my user wall function? Thank you!

Best Regards!
Jundi He

[Luciano Garelli]

Hello,

You can find information about the turbulence model and wall functions in both, the theory and users guide of CS. Also, you can check the files(cs_wall_functions.c and cs_wall_functions.h) about the implementation of this laws.

Additionally, you are solving a heat transfer problem so you need to check your thermal bounday layer. CS has two options (iwalfs option) which are explanied in the B-clptur routine section of the theory guide.

Regards,
Luciano

[Jundi He]

Hi:

when my simulation includes wall roughness, the simulation crush due to a very low friction thermal scalar (tstar) at the rough wall, but when I lower the thermal roughness but only use a normal dynamic roughness, the simulation can run and converge well. My question is, what is the definition of thermal roughness, and the friction thermal scalar, there seems to be little detail information on the user manual or the theory guide. Thank you!

Best Regards!
Jundi He

[Jundi He]

Sorry, for the last post, I already switch from k omega SST to k epsilon, with the impose of the wall function. I guess the effect of the roughness is to modify the wall function near the rough wall. And I don't understand the thermal roughness, and don't know why I get a low friction thermal scalar at the rough wall. Thanks!

[Luciano Garelli]

Hello,

I was reading your last listing file and note the following things. Issue has been detected during the mesh quality check, also you have a message that all your cells have a mass source term, it is correct?

At the outlet you are using a free inlet/outlet BC without a external head loss definitions (warning message in listing), Why are not using an outlet BC?

Regards,

Luciano

[Jundi He]

Thanks for the advise, now I can make the simulation converge. It seems that the outlet head loss is not necessary for the simulation, and yes I have a mass source in different vertical positions. About the thermal roughness and dynamic roughness, when I use the scalable wall function, the near wall mesh can be very fine, but the first mesh still can't be lower than the thermal roughness, which will cause the simulation crush by some reason.

Now I have another interesting question about steady state simulation in code saturne: when my simulation is steady state, the time step can not influence the convergence, but when I change the time step from 0.01s to 0.1s, the simulation crush. When the time step is 0.01s, the simulation can converge well. Why does the time step influence the result? Thanks for the time.