Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Boundary layer mesh gives problems in iterative solvers?

Please login with a confirmed email address before reporting spam

Hi all,

I am trying to solve a 3D turbulent flow problem with k-eps model. I have found that when I try to solve a model meshed with boundary layer mesh the iterative linear solvers find it very tough to coverge. I have tried different preconditioners and solvers, but the problem remains. If I solve them with direct solvers there is not such a problem. I also found that if the aspect ratio of the last element of the boundary layer (next to tetraedral elements) is not very high it sometimes converges.

There is anything in the solvers that can help in this regard? Or is it that COMSOL iterative solvers are just not good enough for this flow problems with boundary elements?

P.S.: I have checked this problem also in very simple 2D model (just a straight tube)

2 Replies Last Post 17 mar 2010, 04:43 GMT-4
Jim Freels mechanical side of nuclear engineering, multiphysics analysis, COMSOL specialist

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 16 mar 2010, 10:54 GMT-4
Have you tried enabling the realizability constraint in the k-e model physics properties ? Also, if you have any problem corner areas that are present in your model, you could try corner smoothing. What stabilization settings do you have ?
Have you tried enabling the realizability constraint in the k-e model physics properties ? Also, if you have any problem corner areas that are present in your model, you could try corner smoothing. What stabilization settings do you have ?

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 17 mar 2010, 04:43 GMT-4
Hi,

Thanks for answering so fast.

No I have not tried to switch on those two options. Actually when I read the manuals I did not understand very correctly their functions. Actually there are many corners, yes.

As to stabilization I just activate streamline diffusion and crosswind diffusion (0.1 as default). I am afraid of isotropic diffusion because I found in many laminar cases that continuity is not longer conserved for coarse mesh and parameters above 0.01-0.02.

I would also like to ask which iterative solvers, preconditioners do you usually prefer for this type of problems. The magnitud of my problem is:

Round, square pipes of around D=0.8m-1.2m. Lengths around some 10-15m. Elbows, section changes. I am currently solving these cases with around 200.000 mesh elements (around 2-3M DOFs) on an 8GB RAM machine.

Should the mesh next to the walls be finer for lower wall offset values? Or can the wall offset values be set independent of the mesh size next to the boundaries? I also thought about using other element types. Could it help?

Regards.

P.D.: Should a flow with high Re (very turbulent) be possible to be solved with the laminar model and not very fine mesh, or should it only be possible with turbulent models? Is the laminar model the one used to solved the full scales of turbulences (in 3D transient)?
Hi, Thanks for answering so fast. No I have not tried to switch on those two options. Actually when I read the manuals I did not understand very correctly their functions. Actually there are many corners, yes. As to stabilization I just activate streamline diffusion and crosswind diffusion (0.1 as default). I am afraid of isotropic diffusion because I found in many laminar cases that continuity is not longer conserved for coarse mesh and parameters above 0.01-0.02. I would also like to ask which iterative solvers, preconditioners do you usually prefer for this type of problems. The magnitud of my problem is: Round, square pipes of around D=0.8m-1.2m. Lengths around some 10-15m. Elbows, section changes. I am currently solving these cases with around 200.000 mesh elements (around 2-3M DOFs) on an 8GB RAM machine. Should the mesh next to the walls be finer for lower wall offset values? Or can the wall offset values be set independent of the mesh size next to the boundaries? I also thought about using other element types. Could it help? Regards. P.D.: Should a flow with high Re (very turbulent) be possible to be solved with the laminar model and not very fine mesh, or should it only be possible with turbulent models? Is the laminar model the one used to solved the full scales of turbulences (in 3D transient)?

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.