Transient Phase Change Solver

[kevinarden]

it depends on what criteria you want to remesh on. It appears in your case it is mesh update.

[Roland]

Yes, you are right, I will try to add the automatic remeshing to the MeshUpdate solver.
I keep you informed.
Still many thanks and happy new year!
Roland

[Rich_B]

Hello,

You may want to look at this post about adaptive refinement:

viewtopic.php?p=24869#p24869

Rich.

[Roland]

Hi,
Thank you Rich for these informations. According to this, only some solvers accept the Adaptive Mesh Refinement...?
As you probably saw it, I would need the Adaptive Mesh Refinement in the MeshUpdate solver. Do you think that the Adaptive Mesh Refinement could be accepted by the MeshUpdate solver, or at least the Adaptive Remesh?
Additionnaly do you know how to set the remeshing triggering, meaning how to set the remeshing criteria?
For instance what means the following command line:
Adaptive Error Limit = 0.1?
Thanks in advance for your help.
Roland

[Rich_B]

Hello Roland,

Sorry, I can't answer your questions about Adaptive Mesh Refinement. The point of sending a link about previous forum posts is that AMR has some limitations in Elmer. If a particular solver isn't coded for AMR, then it won't work at all.

My only suggestion would be to refine your mesh manually, you may spend less time than trying to get AMR to work for you.

Rich.

[Roland]

Hello Rich,
Yes I think you are right. As you suggest it I will try to work with a finer mesh without AMR. Just one thing: do you know if some others work about this crystal growth process with Elmer and if this some critical point of managing the triple point has been already modeled ?
Again thanks for your help and we keep us informed.
Roland

[Roland]

Hi,
I solved the crystal_growth model again with a finer mesh (without automatic remeshing,as you suggested it, Rich).
It starts well (see the here attached results at t=0s and t=0.1s) but at about 0.16s the solver stops and displays following error message:
*****************************************************************************************************************
MeshSolve: -------------------------------------
MeshSolve: MESH UPDATE SOLVER:
MeshSolve: -------------------------------------
MeshSolve:
MeshSolve: Starting assembly...

MeshSolve: Assembly done
MeshSolve: Set boundaries done
10 0.5050E-02
14 0.3301E+39
ERROR:: IterSolve: Numerical Error: System diverged over maximum tolerance.

Note: The following floating-point exceptions are signalling: IEEE_INVALID_FLAG IEEE_DIVIDE_BY_ZERO IEEE_OVERFLOW_FLAG IEEE_UNDERFLOW_FLAG IEEE_DENORMAL
STOP 1
*****************************************************************************************************************
Here attached is also the zipped GUI project.
It would be very nice if someboby could take a look at it and give me his opinion about how this could be improved.
Thanks in advance for any help!
Roland

[Roland]

...Sorry, I sent the t=0.1s.png attached file twice in my former post!
Here attached is the t=0s.png file.
Roland

[kevinarden]

The reason for the solution terminating is mesh distortion. The mesh update, without adaptive mesh, is causing the elements to become so distorted that a meaningful element matrix can not be formed. This is perhaps caused by the boundary conditions 7, 8, an 9. They all come together at a single point, so what happens at that point? In Elmer the last defined BC wins at a BC intersection. Which means that single point at the intersection of BC 7, 8, and 9 is defined by BC 9, BC 7 and 8 have no effect on it. BC 9 updates the mesh in both X and Y and causes the point/node to rise up and to the right. BC 8 has no update mesh 1 defined, so those nodes are not moving in the X, mesh update 2 is defined, but the corner point is following BC 9 not BC 8.

The fix could be as simple as having the current BC 8 be BC 9, and the current BC 9 be BC 8, depending at what your intent is for the intersection of those 3 BCs, and your intent for mesh update on the current BC 8 (line 4).

meshdistortion.png

(551.37 KiB) Not downloaded yet

[Roland]

Hi Kevin,
Thank you very much for these very interesting comments. I will try to invert the BC8 and BC9 command lines and see what happens.
You put the finger precisely on that critical problem of the managing (and modelling) of the triple point on which 3 BCs apply which is obviously conflicting.
Concerning the BC8 on the solid/liquid phase change interface, as you said it, only mesh update 2 is forced to the PhaseSurface variable. I did this just by intuition since I don't master really this Transient Phase Change Solver:
1/is it possible to set both mesh update 1 and 2 with the PhaseSurface variable ?
2/ is it possible to control the position of the triple point with a given setting in the Transient Phase Change Solver?
3/ are there some Elmer examples/ tests where the managing of the triple point is treated in such a crystal growth process?
Many thanks in advance for you help.
Roland