heat: interior boundaries how to

[Gerard]

I have calculated the exact same case in a software (Unorm) that I am familiar with. I have put both the results in one picture. In Unorm you see how the isotherms deflect (extreme because aluminium vs. plastic) around the thermal bridge, I believe this is correct. This is just a simple example for learning Elmer. But in Elmer I get a wrong result, so something is wrong in my settings but I can't get my head around it. Thanks both for your help and patience!

[raback]

Hi

I think you need adiabatic BCs for the middle section up/down. Adiabatic are the natural BCs for FEM so just do nothing there.

-Peter

[Gerard]

Hi, I am really stuck on this one. I made a pptx with my simple workflow and added the project file. As you will notice the result of the Elmer heat flow calc is wrong. Now I believe Elmer is much smarter than this beginner, so what am i doing wrong? If someone could have a quick look, I would be very grateful!

[kevinarden]

If the top boundary is being held at 293 Kelvin then why isn't the entire top boundary 293K in the Unorm solution? That is what a boundary definition is in Elmer, that whole boundary is held at 293K. It must mean something else in Unorm. For the top boundary to not be a constant 293K and the bottom boundary to not be a constant 263, there must be heat crossing those boundaries in the Unorm solution. In Elmer no heat is defined as crossing those boundaries. If the temperature boundaries are fixed and no heat is entering or leaving the system, then the steady state solution is a average temperature distribution through the thickness regardless of conductivity. That is the solution Elmer is presenting. Unorm is doing something different across the boundaries.

[kevinarden]

Also in your gmsh there are coincident nodes between the two bodies, in Elmer this means they do not transfer heat to each other. The coherence command is used in gmsh to eliminate the coincident nodes. after eliminating the coindident nodes

I change the conditions to have a constant boundary temperature of 293 at the top, but let heat radiate out of the bottom to an external temperature of 263. I also set and initial body temperature of 263, not defining an initial condition means the body would start at 0.

You start to see the thermal bridge I one direction, in Unorm it appears heat is flowing across both boundaries.

result.png

(57.97 KiB) Not downloaded yet

case.sif

(2.15 KiB) Downloaded 163 times

[raback]

Hi,

Ok, two problems
1) Boundaries 5 6 7 should all be adiabatic
2) The mesh consists of two separate parts.

The 2nd problem may be fixed by joining with ElmerGrid "-merge". Rather you should address it in Gmsh and have the blocks share the edge. Not create two separate edges.

Wall.PNG (24.24 KiB) Viewed 2715 times

-Peter

[kevinarden]

Yes i changed to both walls being adiabatic
I would be interested in how your sif looks

result2.png

(60.41 KiB) Not downloaded yet

[Gerard]

You are right, this could be one of my stupid mistakes: U norm automatically takes into account an average overall heat Resistance for the air layer on both sides of the wall (R=0.13 m2.K/W interior, R=0.04 exterior). So if you leave these out of the calculation the isotherms remain horizontal? Or in other words, I have to introduce radiation / convection etc. on both sides to get a realistic result? Thx for you help!

[kevinarden]

heat has to flow across the boundary, the only way I know in Elmer to do that is radiate it across the boundary, other experts may know a better way. I suppose you could model a body of air one each side and set R values for each body if you need those particular values, or determine what the equivalent emissivity at the boundary would be for those R values.

[Gerard]

Thanks kevinarden and raback for your comprehensive answers and advice! I will study your suggestions (and do my physics homework:-)