Constant Residual of Turbulence Kinetic Energy and Dissipation Rate
 

Hi all,

I am having an issue when running a simple 2D axisymmetric model of annular flow. The flow if fully turbulent (Re approx. 7E5) and I am using the k-epsilon model with intensity (3%) and hydraulic diameter (0.009 m) as parameters and standard wall functions. The mesh is well refined and structured with a maximum aspect ratio of 3.

I am employing a mass flow inlet BC and 0 Pa pressure outlet. Both walls are stationary with prescribed temperature.

When I run the model, all parameters converge nicely except for k and epsilon. They are simply constant. Also, I get the issue that the turbulence viscosity ratio is limited to 1E5 (default limit set by Fluent).

I have done some searching online yet the only advice is to refine mesh and correct the turbulence parameters. I believe my mesh and BCs are fine so I'm kind of stumped here. Any help would be greatly appreciated...


Attached below are snaps of the mesh and residuals:

Attachment 58309

Attachment 58310

1) your mesh is very crude. I'd advise to refine it. Use at least 20 cells in the height direction.

2) how do you have 2 walls? You are using the axisymmetric model right - that means the center should be a symmetry, not a wall. If you have walls on both sides, use 2D-planar

1) I refined it further and the issue is still not resolved. I ran a similar model a few months ago and it wasn't much more defined than this. The first layer thickness at the walls comes from a yplus value which I calculated as 0.3 mm.

"2) This is an annular flow (between two tubes) meaning there has to be two walls. The line of symmetry is the x-axis, which is not visible from the screenshots.


I have spoke with someone who advised me to use a k-omega model, as it's more appropriate than k-epsilon due to this being a "two layer model". In the k-omega model, the yplus value should <1 meaning I will need to refine more at both walls.

y+ is a dimensionless variable, so it can't be 0.3mm; how did you calculate that?

Anyway, the residuals should converge for k-e too, if it works for k-omega. Did switching to k-omega solve the problem?
I also am not sure if it's allowed to set up an annular flow problem like this - I can remember from when I was a FLUENT TA that for, axisym pipe flow simulations, small offset from the symmetry axis would lead to errors when they started solving, but I don't know if in that case the supposed axis was actually set to wall. Did you try running it as planar just to check whether it would converge?

Viscosity limit error can be resolved by increasing maximum ratio (turb. viscosity/viscosity) by 1 or 2 order. simultaneously one can also try to reduce under relaxation factor of the relevant solver up to 0.01 for few iterations until properties of interest become stable again.