thermal expansion

General discussion about Elmer
Rich_B
Posts: 421
Joined: 24 Aug 2009, 20:18

Re: thermal expansion

Post by Rich_B »

Very nice! Z displacement of 10 mm after 20 timesteps.
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: thermal expansion

Post by Roland »

Hi,
Many thanks to you, Kevin and Rich!
I will see your modified model, Kevin, and try to understand how your mesh changes leads to the right solution!
Say that, as I don't master very well Gmsh, and to save time, I made the geometry and mesh with Comsol and imported it in Elmer.
Perhaps (and probably) the best thing is to make the geometry and mesh with Gmsh and then import it in Elmer. I will try this also in order to see if it gives the correct solution. But it is a bit tricky not to know in advance if an imported geometry/mesh will give the right solution or not...
We keep us informed and again thanks for your efforts!
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: thermal expansion

Post by Roland »

Hi Kevin,
I run your case.sif of your EM_HT_LE file with your Gmsh modified and refined mesh.
The Z displacement is correct, about 10 mm (like in Comsol) and corresponds to the bending of the bi-material plate (the 2 materials having 2 different thermal expansion coefficients).
But the in-plane X and Y displacements, of also about 10mm are not correct and much exagerated. Comsol gives about 1 mm which seems correct according to a rough following analytical calculation:
- The temperature increasing, over 20s, is about delta(T)=365-293= 72 K, which can be rounded to delta(T)=100K for simple calculation
- If the thermal expansion coefficient is alpha then the strain (relative length increase = dl/l) is alpha*delta(T). Thus:
1/For the Aluminium: expansion coefficient: alpha_alu= 2.6e-05 --> strain_alu = 2.6e-05*100 = 2.6e-03 = 0.26% (=dl/l)
2/ For the insulating coating: expansion coefficient: alpha_coating = 5e-05--> strain_coating = 5e-05*100 = 5e-03 = 0.5% (=dl/l)
This means that in the XY plane the X and Y displacements (corresponding to the X and Y length increasing) are about 0.3% to 0.5% of the initial length, meaning 0.5%*200mm = 1 mm. This is much less (10 times!) than the values given by Elmer.
Thanks to give your opinion about that when you have a little moment.
I am wondering if the result given by the elastic solver is very sensitive to the mesh...?
We keep us informed.
Thanks in advance for your help
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: thermal expansion

Post by kevinarden »

Many LE solvers, and I am not familiar with Comsol, do not like unrestrained degrees of freedom, If you have a body like the plate that is free to move unrestrained in the X, Y, or Z, and the rotations about those axis, then technically the motion can be infinite with very little load. In actual it is numeric computation problem, so not all solvers handle it automatically. Some codes implement an inertial relief, allowing you to analyze a structure in space with no load in some directions, but some solvers need the 6 rigid body motions restrained. I believe elmer wants each rigid body motion restrained for LE. You already have all edges restrained Z, If you picked one edge and restrained X and picked another edge to restrain Y, then the plate body could not float in space in the X, Y, or Z, nor could it arbitrarily rotate about any of those axis. Some LE solvers will actual quite and report an error if there is a rigid body mode, in Elmer it seems to cause a poor convergence, and occasionally strange results. My recommendation with the LE solver in Elmer is to make sure the body is restrained in all 6 DOF directions, X, Y, Z and Rotation about X, Y, Z. This can usually be done without impacting the desired deflection shape of the body. As before, sometimes I just target a few nodes to take out the possible rigid body motion.
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: thermal expansion

Post by Roland »

Hi Kevin,
Yes I can understand that the LE Elmer solver needs to restrain all the directions (and rotation axis) for a correct convergence.
But fisrt in the GUI you can only access the faces and not the edges. Is there a trick to restrain an edge (or also a point) in the GUI?
Secondly the plate is just set for instance on a flat table, meaning that, when it bends, the lower 4 edges are free to glide horizontally (in X and Y directions on the table) which I set with a zero Z_displacement BC (only for Z!). I guess that if you restrain the 4 edges in the 3 X,Y,Z directions (same with rotation axix) the solution will not be the same as if only blocking the Z direction...?
What do you think about that?
Thanks in advance for your help
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: thermal expansion

Post by kevinarden »

If you know a node number, you can in the free test input of the GUI on the boundary condition tab type

Target Nodes(1) = node number

to target a node vice a surface.

In this case you can pick 1 of the faces in the XZ plane and restraint X, The plate will be free to grow in the X direction, but not move rigidly in the X direction. You can then pick 1 of the faces in the YZ plane and restrain Y. The plate will be free to grow in the Y direction, but not move rigidly in the Y direction. This along with the Z restraint on all edges will also prevent any rotation about the axis.

The result should be the same, since the plate can expand in each direction.
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: thermal expansion

Post by Roland »

Hello Kevin,
Ok thank you I will try this.
I will also try to make the geometry not with Comsol(the problem comes perhaps from the import from Comsol...) but directly with Gmsh and see if the problem remains or not.
As I am not very clever for making the geometry(and the mesh)with Gmsh, have you some experience with Freecad? Is it more convenient than Gmsh? Is there a direct connection possible with Elmer?
Thanks in advance for you help!
Rich_B
Posts: 421
Joined: 24 Aug 2009, 20:18

Re: thermal expansion

Post by Rich_B »

Hello Roland,

Regarding your questions about gmsh and freecad, I'd like to mention that the document GetStartedElmer.pdf includes detailed, step by step, instructions on using each one of those two programs to create Elmer models.

Rich.

Edit: Sorry, I misspoke, GetStartedElmer includes detailed instructions for gmsh and Salome, but not Freecad.
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: thermal expansion

Post by kevinarden »

I have used Salome, gmsh, and freecad, and my preference is in that order. Salome is the most powerful geometry creator and mesher but harder to learn. gmsh is good for simple geometries. freecad seems more geared for cad drawings not fea models, although many people use ot for fea models. All three have documentation, tutorials, and forums like this.

As RIch said the Document
https://www.nic.funet.fi/pub/sci/physic ... dElmer.pdf

Chapter 9 is a good start.
Roland
Posts: 226
Joined: 12 Apr 2018, 11:29
Antispam: Yes

Re: thermal expansion

Post by Roland »

Thanks a lot!
I will take a look and we keep us informed
See you
Post Reply