Elmer Discussion Forum

Hi,
I have convergence problems with a steady incompressible flow (inlet, outlet and walls) and have tried the turbulence model k-epsilon with different solver settings (tolerances, Linear Solvers and relaxation factors) and mesh densities.
But I have still the problem and get in the plot only a small layer of velocity at the inlet. The rest of the fluid is not moving.
Maybe someone has an idea what the problem is. I think I am setting up the whole simulation wrong.

I am using ElmerGUI with Windows 10 64bit.

Kindly regards
Markus

Here is the convergence and velocitiy plot

If it is turbulent flow than a steady state may not exist.
In that case you can use transient analysis.
Mesh size matters, it may be too coarse of a mesh.
Problem size matters, yours is 400 meters by 600 meters.

I scaled it down to 0.4 x 0.60
I meshed it coarse, I would go any coarser.
This is the last step before it diverged (13 Steady State iteration)

Hi ,
thank your for your time. The scale factor of my elmer model is correct with 0.4x0.6 - the attached CAD Model (exported STEP from Salome) was wrong.
I attached the last mesh with its Density of 1.3e6 nodes and 6.9e6 elements. I think it is fine enough to get a result.

In your Model you have the same "step" of velocity at your inlet and I think this is physically not possible.

Do you really think it could be a turbulence problem with the steady state solving? If you reduce the velocity at the inlet its getting a little bit better, but its still problematic. And with this dimensions and only v_inlet=0.01 mm/s I thought it to be laminar.

In my opinion the case is equal to to the curvedPipeGUI-tutorial?!

I also tried to run a simulation with viscous layers at the walls (meshed through salome), but Elmer doesn´t accepted the resulting element type.

Kindly regards
Markus

The velocity input is of 0.01 is 0.01 m/s. One way to test if is the flow is to use a really high viscosity, if it converges you can start decreasing it until it diverges. I believe the inlet size on the curved pipe tutorial is 0.01 m. at 0.4 m a velocity of 0.01 m/s has a Reynolds of about 3000, which is in the turbulent range, In the pipe tutorial the Reynolds number is about 100.

Yeah, I think I got it with a viscous layer mesh and some modification in the ke-solver like
http://elmerfem.org/forum/viewtopic.php?p=22933

Thank you for your support .