Transient Phase Change Solver

Numerical methods and mathematical models of Elmer
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: Transient Phase Change Solver

Post by Roland »

Hi,
As discussed in my former post, Kevin, I removed the air domain (which is of no use for the moment) in the simplified crystal_growth model (with the former free surfaces considered as walls for the moment, in order to test only the moving of the phase change interface).
Here attached is the corresponding zipped GUI Elmer project.
When looking at the Paraview animation it seems better but unfortunately the triple point does not really follow the curved boundary 1 (which is a wall for the moment) but seems to go downward in the vertical direction. Is this perhaps due to the fact that, in the phase change interface, only MeshUpdate 2 (vertical mesh node update) is updated with PhaseSurface and not MeshUpdate 1? Still it comes again perhaps from the fact that I am not sure of the proper use of the Transient Phase Change Solver....
Thanks in advance to give your opinion about that. Any help is wellcome!
Roland
Attachments
CG_03_Elmer.zip
(915.03 KiB) Downloaded 27 times
arved
Posts: 27
Joined: 06 Aug 2020, 15:05
Antispam: Yes

Re: Transient Phase Change Solver

Post by arved »

Hi Roland,

I just had a quick look at your sif-file. By default, the Variable PhaseSurface has only one degree of freedom. With the additional line

Code: Select all

Variable DOFs = 2
you can make it 2D and set both dofs of MeshUpdate

Code: Select all

Mesh Update 1 = Equals PhaseSurface 1
Mesh Update 2 = Equals PhaseSurface 2
You also have to adjust the initial condition

Code: Select all

PhaseSurface 1 = Real 0.0
PhaseSurface 2 = Real 0.0
I just tried, unfortunately it crashes. I hope that helps anyway!

Best regards
Arved
Attachments
case.sif
(8.05 KiB) Downloaded 25 times
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Transient Phase Change Solver

Post by kevinarden »

I viewed closely at the movement near the triple point in paraview. Not only does the triple point not stay on the curve, none of the points on the curve stay on the curve. It does not appear that the nodes on the curve is adhering to the Normal-Tangential Mesh Update = True condition. Target boundary 7 is but target boundary 3 is not. Maybe there is a problem with Normal-Tangential Mesh Update = True for a curved boundary.

Boundary Condition 7
Target Boundaries(2) = 1 7
Name = "BoundaryCondition 12_free_surf_liq"
Normal-Tangential Mesh Update = True
Noslip wall BC = True
External Temperature = 293
Mesh Update 1 = 0
Heat Transfer Coefficient = $h_fs
End
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: Transient Phase Change Solver

Post by Roland »

Hello Arved and Kevin,
Thanks a lot, Arved, for your help! I will try what you suggest about using PhaseSurface as a 2 DOF variable (I must admit that the doc is not very clear about that...).
Thanks also to you, Kevin. As you say it seems not normal that the triple point does not follow the curved wall boundary 1 in normal-tangential coordinates with MeshUpdate 1 = 0....
Again the problem comes perhaps (probably...), as already mentionned, from the fact that the triple point belongs to 3 boundaries (the phase change interface 6, the crystal vertical surface 5, and the liquid surface 1-7 ), meaning that it must "obey" to 3 BC's, which makes that this is perhaps the reason why this triple point does not "know" what to make... As a specialist of crystal growth (which I am not!) have you perhaps, Arved, an opinion about that (I must say that in the Comsol models there seems to be no problem)?
So what could be done so that the triple point more or less follows (at least it is my feeling...) the curved boundary 1 (which is a wall for the moment and which will again be the liquid meniscus later) ?
We keep us informed.
Roland
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Transient Phase Change Solver

Post by kevinarden »

It appears you are correct that a node that shares a boundary will not follow both boundary conditions. This is complicated by the fact the triple point has three BCs, it will only follow one. I tested normal-tangent BCs, using only heat equation, no phas change, and it doesn't appear to work on curved boundaries. It works on sloped boundaries, but as you say when a node is shared by two boundaries, it only follows one not both.
original mesh with sloped lines. The mesh update in this test is independent of any other variable, it is just updated with variable time, so this is just a test of the mesh update solver only.
undeformed.png
(182.86 KiB) Not downloaded yet
updated mesh, the node at the interface only follows 1 boundary.
updated.png
(184.19 KiB) Not downloaded yet
The nodes at the intersection follow the slope of the first boundary but ignores the second.
Same thing happens on the curved boundary test, but the other nodes on the curve. do not seem to follow the curve.

In other codes I have used the boundary conditions at an intersection of boundary conditions appeared to be combined. In Elmer it appears that one or the other is followed. This may explain why Comsol is different at that point.
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: Transient Phase Change Solver

Post by Roland »

Hi,
Ok Kevin for these interesting tests. My questions:
1/ when a point is submitted to 2(or 3) BC's at the same time, which of these BC's does the point choose? And how make that the point then moves according to this unique BC?
2/do you think that there could be a "trick" for forcing the triple point for instance to follow only the meniscus(which could be a simple bent line instead of a curved one(like on your PNG drawings), since after several time steps (as it is a liquid surface when the real free surface BC is again set) it will deform according to the capillarity law?
Thanks in advance
Roland
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Transient Phase Change Solver

Post by kevinarden »

1> My understanding is the last one defined is used.
2>I can rum some tests
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: Transient Phase Change Solver

Post by Roland »

Yes ok Kevin, this would be nice if you could bring some help! As you can understand I would absolutely like to have this crystal growth model work, even on a simplified way in a first state.
We keep us informed and still many thanks
Roland
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Transient Phase Change Solver

Post by kevinarden »

This attempt came close, the triple point follows the slanted line, but the point above it on the vertical line stays with the vertical line. Ideally it would transition from the vertical line to the slanted line, which is not so easy, but I have some ideas to test.
t0.png
t0.png (104.66 KiB) Viewed 381 times
ta.png
ta.png (106 KiB) Viewed 381 times
tc.png
tc.png (93.81 KiB) Viewed 381 times
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: Transient Phase Change Solver

Post by Roland »

Ok Kevin, this is very interesting.
Please keep me informed when you have further things about that.
Thanks again
Post Reply