Page 1 of 2

Heating a metal wire

Posted: 14 May 2020, 07:59
by SnappyTitan
Hello,

I'm trying to study the thermal profile of a metal wire.

Setup:
1) I created a cylinder using Salome and I meshed the cylinder into three parts: the two faces at the end of the cylinder and the whole body.
2) In Elmer, I used the two end faces as my boundaries (ground and high).

The simulation ran successfully without any warnings or errors. However, the result in Paraview showed that the whole wire was at 0V, 0 joule heating, etc..

Troubleshooting so far:
1) I verified in ElmerGUI that there were indeed 3 boundaries as expected.
2) There was only 1 body as expected.

I have attached my .sif and .unv below.

Thanks


Update:
So instead of creating a sif file and editing it, I set all parameters (equations, materials, etc..) in ElmerGUI and created a sif file. This seems to work. I compared the sif I wrote before with the new one and I didn't see any differences. Weird!!! :? :?

Re: Heating a metal wire

Posted: 14 May 2020, 12:30
by kevinarden
viewtopic.php?f=7&t=7021&hilit=joule+heating

May help to read through this post for calculating the joule heating of the wire

Re: Heating a metal wire

Posted: 15 May 2020, 08:55
by SnappyTitan
Hi Kevin,

Thank you for the link. I was able to make some progress. Elmer yielded the values for effective resistance and volume very close to the ones I calculated. The thermal profile has the same trending as expected from experimental data. However, the value is almost 500K lower with Elmer result. I have verified all input parameters and they seem correct. Do you have any suggestions?

Re: Heating a metal wire

Posted: 15 May 2020, 12:10
by kevinarden
Looks like the 800 temperature condition on the ends conducts though the metal wire making the whole wire 800 since it is a steady state solution. Since metal is a good heat conductor that would be the expected steady state condition for an enforced temperature boundary. Either the joule heating is not adding anything to the heat equation or it is small.

I took the 800 boundary condition off and the temperature from joule heating was 2.E14, which is certainly not small.

Re: Heating a metal wire

Posted: 15 May 2020, 18:06
by SnappyTitan
Hi Kevin,

800K at the ends of the wire is what I expect them to be when applying 7A. I have attached the plot I'm referencing below.

Also, how did you remove the temperature BCs? I've tried that and kept getting trivial results.

Thanks!

Re: Heating a metal wire

Posted: 15 May 2020, 18:52
by SnappyTitan
I figured it out. I(diot) used the diameter value for the radius.The results are correct now! Thanks :)

Re: Heating a metal wire

Posted: 15 May 2020, 23:43
by kevinarden
I just deleted temperature=800 line in the sif file.

Re: Heating a metal wire

Posted: 16 May 2020, 00:54
by kevinarden
I think the heat equation is going to take the 800 degree boundary condition and conduct heat into the wire, until the whole wire gets to a steady state of 800 degrees. If you only use the heat equations that is what happens.

I took the specified temperature off and set it to radiate heat out of the boundaries to an external temperature of 300 and let the joule heating be the only heat source.

It resulted in 900K.

I had to increase the number of steady state iterations to converge.
case.sif
(3.32 KiB) Downloaded 235 times

Re: Heating a metal wire

Posted: 16 May 2020, 06:57
by SnappyTitan
Hi Kevin,

With the BCs and the wire diameter of 0.256 mm (originally, I mistakenly read for 0.512 mm), I was able to get a profile similar to the experimental one. 800K on the ends and 1200K at the center. This brought up another question. What is the best way to setup the model when temperature BC is not provided. There is obviously a difference with and without BCs.
Also, when I removed the temperature BCs, the heat equation didn't converge(thermal2).

Re: Heating a metal wire

Posted: 16 May 2020, 20:32
by raback
Hi

Temperature BCs are essential to the solution of the heat equation. If you don't know them that is usually sign that you should extend the model to include a larger domain where you know the boundary conditions or use heat transfer coefficients to idealized farfield temperature.

Heat equation does not converge if you don't apply BCs for the simple reason that the solution is not uniquely defined if there are no boundary conditions. It is not a deficiency of the code, or linear solver strategy, it is just how the world is.

-Peter