Heating a metal wire

Numerical methods and mathematical models of Elmer
SnappyTitan
Posts: 8
Joined: 14 May 2020, 07:45
Antispam: Yes

Re: Heating a metal wire

Post by SnappyTitan » 17 May 2020, 09:23

Hi Peter,

Thank you for your response. I think I follow your explanation. Though, I will have to look more into the suggested methods as I'm not very familiar with them.

On the topic of convergence, I'm running into an issue now with my new setup. Basically, I extended the length of the wire from 8mm to 16mm and BCs to 850K with everything else remained the same. The heat solver won't converge. I have played around with different combinations of the Solver method, system preconditioning, and relaxation factor as suggested in the manual but nothing seems to work. The convergence history always periodic. One example is below. I set the maximum output to 50 but I didn't see anything weird; though I couldn't make sense with a lot of them. Could you or someone provide some hints on how can I approach this?



Thanks much!!

unv file: https://drive.google.com/open?id=1SZ9gt ... SqbsU-5r_D
Convergence Historyhttps://photos.app.goo.gl/azJDESJkmWJfVV8a7
Attachments
case.sif
(3.34 KiB) Downloaded 10 times

kevinarden
Posts: 618
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Heating a metal wire

Post by kevinarden » 17 May 2020, 12:13

Yes I agree the heat equation has to have a boundary condition, but a specified temperature conducts into the metal which is not the intent in this case.

raback
Site Admin
Posts: 3670
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Heating a metal wire

Post by raback » 17 May 2020, 17:41

Hi

Probably problem is thermal conductivity being function of temperature. Add some relaxation. For example, set both Solvers to have "nonlinear system max iterations = 1" and "nonlinear system relaxation factor = 0.5".

-Peter

SnappyTitan
Posts: 8
Joined: 14 May 2020, 07:45
Antispam: Yes

Re: Heating a metal wire

Post by SnappyTitan » 17 May 2020, 20:02

kevinarden wrote:
17 May 2020, 12:13
Yes I agree the heat equation has to have a boundary condition, but a specified temperature conducts into the metal which is not the intent in this case.
Hi Kevin,

My goal for this study is to compare how well the model fits to the published results : thermal trending, resistance, etc... I'm not sure if this is the right way to go about it since this is my first actual project excluding the tutorials. But as you and Peter suggested, I will need to find a different way to go about it for the next step since the BCs aren't known in advance.

SnappyTitan
Posts: 8
Joined: 14 May 2020, 07:45
Antispam: Yes

Re: Heating a metal wire

Post by SnappyTitan » 17 May 2020, 20:16

raback wrote:
17 May 2020, 17:41
Hi

Probably problem is thermal conductivity being function of temperature. Add some relaxation. For example, set both Solvers to have "nonlinear system max iterations = 1" and "nonlinear system relaxation factor = 0.5".

-Peter
Hi Peter,

I setup a script to loop through the parameters. The values that produced convergence are
HeatSolver: Nonlinear System Max Iteration= 5
Nonlinear Relax Factor = 0.1
StatCurrentSolver: Nonlinear System Max Iteration = 1-5
Nonlinear Relax Factor =0.1-0.7

However, only the heat equation converged after >200 iteration steps. The static current oscillated about the relative change = 0.1. This is confusing to me. I thought the simulation only stopped when both solvers converged or the steady state Max iteration reached the limit.
Attachments
case.sif
(3.35 KiB) Downloaded 11 times

Post Reply