Elmer Integration (GSoC) - Activity Log

[thschrader]

As I remember there is a limitation of Elmer-FC-FEM regarding 2D-Meshes. I think it is a limitation of Elmer-Grid, but I am not 100% sure. So, it is not possible to simply recreate the original 2D-model from the Elmer-tutorials.

You are right, I used a 3D-FC-model too. Its the same in the cfd-of-wb (openfoam), you must "construct"
the original 2D-case as 3D. I aborted the calculation after more than 1 hour runtime, it seems that the elmer-solver does not
converge on the 3D-calculation (the 2D-case converges after 5 min...). It would be helpful to see the residuals when running elmer
in FC like in the cfd-of-wb. However, I am not very familiar with elmer, maybe I did something wrong. On the other hand:
in the tutorial itself they talk about convergence problems...
I will use a "real world problem" (3D-flow around a cylinder) as I did in cfd-wb with openfoam.
it's a long long way to tipperary
For me the most interesting tutorial is nr. 9:

FluidStructureInteraction.JPG (58.77 KiB) Viewed 2424 times

noConvergence.JPG (49.58 KiB) Viewed 2424 times

[thschrader]

Did some testing with different meshes.
gmsh: the solution in fem-wb/elmer only converges when using 1st-order elements
(like gmsh does in cfd-wb by default).
I only get the solution from the tutorial when using hex-meshes (cfmesh/snappyhexmesh).
When using tetrahedrons, the velocity-field is completely different and I
have no idea why.

test.JPG (83.72 KiB) Viewed 2387 times

[HoWil]

I aborted the calculation after more than 1 hour runtime, it seems that the elmer-solver does not
converge on the 3D-calculation (the 2D-case converges after 5 min...).

Firstly, please check your solver settings like the Preconditioning, Solver-Type or order of the solvers (priorities). There are many important things which can be forgotten/overseen. I attached some quick screenshots of the most interesting parts.
Secondly, always double-check the material-properties in python console as listed in the code-box here https://forum.freecadweb.org/viewtopic. ... 10#p223408 . I had several times problems with the material-dialogue and I am still not sure if everything is solved jet (check "App.ActiveDocument.FluidMaterial.Material").

It would be helpful to see the residuals when running elmer
in FC like in the cfd-of-wb.

There is no plot implemented jet, but you can see the residuals in the solver-output as text. This is in 99% of my cases enough to get an idea if everything runs fine.
BR,
HoWil

[HoWil]

Did some testing with different meshes.
gmsh: the solution in fem-wb/elmer only converges when using 1st-order elements
(like gmsh does in cfd-wb by default).
I only get the solution from the tutorial when using hex-meshes (cfmesh/snappyhexmesh).
When using tetrahedrons, the velocity-field is completely different and I
have no idea why.
test.JPG

One reason could be that in the tutorial a flow profile is used for the input (Velocity 1 = Variable Coordinate 2; Real MATC "6*(tx-1)*(2-tx)") .
But there has to be something else missing! Please compare the screenshot form my previous post https://forum.freecadweb.org/viewtopic. ... 10#p223408 (and the attached one). It fits in my opinion the cfd-wb plots very well. My tip... wrong material or too coarse mesh.
BR,
HoWil

Bildschirmfoto von »2018-04-01 21-39-27«.png (207.67 KiB) Viewed 2379 times

[thschrader]

HoWil, thanks for the hints.
I used the solver settings from your post above,
but without simulating the heatflow. In openfoam you can
run multiphysics, but (at the moment) it is not possible to do this
with cfd-wb in FC. One reason could be, that my model is much bigger.
When loading the tutorial file step.grd in elmer-GUI, you can see at the bounding box,
that the channel is 12 m long and 2 m high. Your model is 1,4 m long. But it does not
explain, why I get different results when using hex-mesh / tetra-mesh in cfd-wb alone.
I will have a look at it
Thomas

[thschrader]

But it does not
explain, why I get different results when using hex-mesh / tetra-mesh in cfd-wb alone.

Aaaah...
Because it makes sense to set the output to the last calculated time-step,
not to the first...
Elmer-gmsh: should be the same "problem". I get only one result-file, not a iteration
series like in cfd-wb. Seems I have wrong settings in the output-solver.

first.JPG (39.64 KiB) Viewed 2349 times

last.JPG (39.13 KiB) Viewed 2349 times

elmer_sif.JPG (22.58 KiB) Viewed 2349 times

[thschrader]

@HoWil:
OK, I got it. I rebuild your 1,4 m channel and did a elmer run without heat,
using your solver setting. I checked the result with cfd-wb (gmsh). Nice.
It is no problem to use a constant velocity inlet, after 300 mm you have
a parabolic velocity-profile.
I will try to rebuild elmer tutorial nr.8 (fluid interaction
with elastic obstacle) in FC.
Thomas

elmer.JPG (144.2 KiB) Viewed 2282 times

cfd_wb_gmsh.JPG (176.28 KiB) Viewed 2282 times

[HoWil]

@HoWil:
OK, I got it. I rebuild your 1,4 m channel and did a elmer run without heat,
using your solver setting. I checked the result with cfd-wb (gmsh). Nice.
It is no problem to use a constant velocity inlet, after 300 mm you have
a parabolic velocity-profile.
I will try to rebuild elmer tutorial nr.8 (fluid interaction
with elastic obstacle) in FC.
Thomas

Great

[thschrader]

OK, I tried to rebuild elmer tutorial nr. 8 (fluid-structure interaction)
with FC, but I have problems with the tutorial itself.
How must I interpret the mesh elastic modulus in the sif?
I found nothing in the tutorials or using google.
The tutorial does not run when using physically reasonable values for the structure,
for instance using the values for concrete. The solver does not converge.

mesh.JPG (23.7 KiB) Viewed 2161 times

elmer_tutorial8.JPG (56.15 KiB) Viewed 2161 times

[HoWil]

OK, I tried to rebuild elmer tutorial nr. 8 (fluid-structure interaction)
with FC, but I have problems with the tutorial itself.
How must I interpret the mesh elastic modulus in the sif?
I found nothing in the tutorials or using google.
The tutorial does not run when using physically reasonable values for the structure,
for instance using the values for concrete. The solver does not converge.
mesh.JPG
elmer_tutorial8.JPG

Hi Thomas,
I am not familiar with this problem/tutorial, but it could be that these settings are related to an "moving"/adaptive mesh.
If it is so that this is not supported by Elmer in FC jet.
BR,
HoWil