Diverging solutions for natural convection flow at differentially heated vertical wal
 

Hi guys,


this is my very first thread in this forum.

Currently I am working on natural convection flow in a closed cavity with two differentially heated, opposing, vertical side walls using Boussinesq approach using OF v2006.
I use the chtMultiRegionFoam, because for one of the walls, the temperature profile is a function of hight. The solid wall represents a thin copper plate.
So far so good, my simulation run quite good for wall hights until 7.5m... but when I furtherly increase my wall hight to 10m, the solution diverges (exploding Courant-Number/shrinking time steps). I already limited my temperature fields in fvOptions, so no non-physical temperature values anymore.

I run those simulations in 2D, so front and back patches are set as empty (2D, because number of cells would fastly exceed my HPCs computation capacities, while maintaining y+~1)

I guess my problem lies with the pressure. But I am not able to find my mistake.

All walls got "fixedFluxPressure"-BC for p_rgh and "calculated"-BC for p of course. InternalField values I sat to 0, because my equation of state is Boussinesq (I tried 1e5, but it didn't change anything of course) Thus I needed to define a pRefPoint and a pRefValue in fvSolutions. I tried setting pRefPoint to the very top of my geometry und to the very bottom - still getting same error. pRefValue is still 0 (also tried 1e5 here).
I use kOmegaSSTSAS, but also tried laminar for checking if it's a matter of the turbulence model, but it seems not to be the problem.



My error message is basically saying, that my time steps became much too small for getting a solution. Also number of iterations for p_rgh increase significantly together with rapidly decreasing timeStep size.


All of my walls have noSlip conditions in U and wallfunctions for turbulence:
alphat --> compressible::alphatWallFunction
nut --> nutkWallFunction (for low- and highRe)
k --> kLowReWallFunction (for low- and highRe)
omega --> omegaWallfunction (for low- and highRe)


For Temperature I have "fxedValue" on the right vertical wall and the typical "compressible::turbulentTemperatureCoupledBaffleMi xed"-BC for fluid-solid-interface on the right vertical wall. Top and bottom walls are set to "zeroGradient".

As I said before, my simulatios ran pretty good and fast with lower vertical wall hights. But as I increase wall hight and cavity width (proportionally with an aspect ratio of hight/width=10), keeping mesh resolution nearly constnat (still maintaining y+~1), the simulation won't run anymore.

As an additional information: the physical behavior is a natural convection loop, bacause the left solid wall has a higher Temperature than the fixed temperature right wall.



Two attached pictures show the Temperature field of the complete geometry (in this case hight: 0.384m, cavity width: 0.038m, copper plate thickness: 9.5mm) and the velocity field with mesh resolution of the top part of the geometry.



Attachment 84537

Attachment 84538


I'd be really happy, if anyone could help.


Greetings
David