LES simulation with conjugate heat transfer in code_saturne 6.0.0

[xiaoxue123]

Hello,

I'm trying to run a LES simulation with conjugate heat transfer. The version I'm using is code_saturne-6.0.0.
The simulation is restarted from a previous run in which the solid domain was not included. My problem is that I find in the case with conjugate heat transfer considered, the temperature field is fixed (N_iter for temperature field is 0 as reported in the listing file) while the velocity field is changing. I haven't figured out why the temperature field becomes constant with CHT considered. Fluctuations in temperature may be damped by the solid but it should not be totally constant. Could you help to have a look? Attached please find the listing file, the xml file (named cylinder.xml), the subroutines which I used to define the properties of solid, enable internal coupling and initialize the solid domain.

Thanks,
Xiaoxue

[Yvan Fournier]

Hello,

Your setup seems incomplete. You also need to declare an internal coupling (see internal_coupling example in cs_user_parameters-base.c). Otherwise, things remain uncoupled.

If you get spurious fluctuations of non-coupled variables in the solid part, you may need to use the cs_mesh_quantities_set_has_disable_flag() and cs_internal_coupling_tag_disable_cells() functions but this might still have bugs in v6.0.0 (v6.0.4 has fixes, so is recommended).

Best regards,

Yvan

[xiaoxue123]

Hello Yvan,

I actually had internal coupling in the simulation. The cs_user_parameter.c, cs_user_initilization.f files were not successfully attached to my first post, sorry about that. They are now attached here.

The spurious thing I found is that, in the listing file the temperature field seems to be totally quiet, the output of temperature at every time step seems to be the same. It is a LES simulation, the velocity field does have fluctuations as shown in the listing file. Do you have any idea of what's going on with the temperature?

Thans,
Xiaoxue

[Yvan Fournier]

Hello,

Could you post the run_solver.log and performance.log files ? I'm not sure which info is logged for the internal coupling (I would need to check in the sources), but there might be some elements in the coupling location step, allowing to see if the mapping from the fluid to/from the solid parts are done correctly.

Also, using "or" instead of ", " in the selection criteria might be safer (from memory, both should work, but we only use the " or " syntax, so there might be a selection bug. You can also check the selection by defining a postprocessing mesh with the same criteria, and visualizing it).

Finally, trying a similar setup an a RANS computation (where I am sure the internal coupling works) may help.

Regards,

Yvan

[xiaoxue123]

Hello,

The run_solver.log, performance.log and setup.log are uploaded.

I'll try to run a RANS simulation then get back to you.

Best regards,
Xiaoxue

[Yvan Fournier]

Hello,

The performance.log isnot complete, so it seems the computation crashed or was killed. The "plate_*" surfaces are interior faces, where I would expect them to be on the boundary once the automatic preprocessing has been done? I am not 100% sure here, as the log may be done before the mesh is split along the solid boundary, but I would expect them to appear as a boundary (though they should initially be interior faces as here).

Could you try using cs_internal_coupling_add_volume instead of cs_internal_coupling_add ? Since its use is simpler/more automated, it might solve the issue.

Best regards,

Yvan