[Ruonin]

Hi folks,

I'm implementing a Low Re k-ep model and, as mentioned in the forum on several occasions, I have to refine the mesh heavily near the wall.

I have a boundary layer, and I have used the CFD tools for accessing the distance for the first layer. In my case it's 5E-6 m, which I think is very close to the numeric minimum value of the software.

My question is this: as I refine the mesh near the wall, I get a problem in the convergence. The k and ep variables just stall at a certain residual value.

Has anyone experienced this?

Best Regards,

Rui Silva

[Mohd123]

Maybe you have reached convergence if stall at high residual value.

At what residual value it stalls before refining and after refining?

[Ruonin]

Thank you for your reply Mohd123,

Before refining I get to about 10^-3 residual (standard for COMSOL) and after refining the value is about 10^-2.

Could it be that the boundary layer mesh is to fine and causing numerical diffusion?

Best Regards,

Rui

Originally Posted by Mohd123

Maybe you have reached convergence if stall at high residual value.

At what residual value it stalls before refining and after refining?

[Mohd123]

The best thing I suggest to do to make sure that mesh is totally independent of solution is to do a Grid Independence Study, so that you keep refining the mesh (from course to fine) until you reach a contant solution and the more you refine the mesh the solution will be constant.

Originally Posted by RSilva

Thank you for your reply Mohd123,

Before refining I get to about 10^-3 residual (standard for COMSOL) and after refining the value is about 10^-2.

Could it be that the boundary layer mesh is to fine and causing numerical diffusion?

Best Regards,

Rui

[FMDenaro]

Originally Posted by RSilva

Thank you for your reply Mohd123,

Before refining I get to about 10^-3 residual (standard for COMSOL) and after refining the value is about 10^-2.

Could it be that the boundary layer mesh is to fine and causing numerical diffusion?

Best Regards,

Rui

Hi,
for any consistent and stable discretization, in no way a grid refinement can cause an increasing in the numerical diffusion! Try check your iteration parameters