SaveLine tangential component

Numerical methods and mathematical models of Elmer
Post Reply
zeljkoT
Posts: 24
Joined: 14 Nov 2018, 12:55
Antispam: Yes

SaveLine tangential component

Post by zeljkoT »

Hi

I'm successfully using the SaveLine solver to extract the electric field on specific boundaries (see case: SaveLine on multiple boundaries)
I checked the docu and came across the "Save Flux" keyword. This one enables to get the normal component.

If I use those keywords within the sif file:

Code: Select all

! SaveLineEnorm
Solver 2
  Exec Solver = After all
  Equation = "SaveLine shield"
  Procedure = "SaveData" "SaveLine"
  Filename = "boundary_lines_shield.dat"
  Output Directory = ./results
  File Append = Logical False
  Save Mask = String "line shield"
  Save Flux = Logical True
  Flux Variable = String "Potential"
  Flux Coefficient = String "Relative Permittivity"
End
then the electric field on the boundary of the high potential side gets extracted. In addition, there is another file called *.names with the column names of the extracted data (see image in appendix).

My understanding is that:

[*] "Flux 1" == normal component
[*] "Flux 2" == tangential component

But what is "Flux normal" ???

Thx in advance and cheers
Attachments
Screenshot 2023-08-17 184207.png
Screenshot 2023-08-17 184207.png (16.1 KiB) Viewed 421 times
raback
Site Admin
Posts: 4832
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: SaveLine tangential component

Post by raback »

Hi

I think it saves cartesian components and the dot product with normal.

Note that the most accurate way to to get fluxes is to compute reaction forces (in this case induced nodal charge) by "calculate loads". If you want to know the charge distribution then divide this by the nodal weight. See test case "WeightComputation".

-Peter
zeljkoT
Posts: 24
Joined: 14 Nov 2018, 12:55
Antispam: Yes

Re: SaveLine tangential component

Post by zeljkoT »

Hi Peter

Thx for your response.

Ah yes, of course, Flux1 and Flux2 are in Cartesian components. Thx.
To get the tangential component on an insulator surface, I would simple do Ft = sqrt( E1**2 + E2**2 ) - Fn with the extracted data from SaveLine solver.
Do you think, this is not accurate enough? In the appendix you can see the fluxes by normal and tangential components.

Regarding your second point: You mean, to divide the "induced nodal charge" by the nodal weights ? I'm not sure if I understand what you mean exactly. Can you explain it in more detail pls ?
I did check the test case you mentioned and included the export of:
  • potential loads
  • potential weights
  • potential boundary weights
  • relative permittivity
So what do you suggest how to proceed?

Thx in advance and cheers
Attachments
Figure_1.png
(50.48 KiB) Not downloaded yet
Screenshot 2023-08-21 121623.png
Screenshot 2023-08-21 121623.png (23.68 KiB) Viewed 395 times
raback
Site Admin
Posts: 4832
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: SaveLine tangential component

Post by raback »

Hi,

Any method to compute the flux over boundary will include only the normal component since the tangential component of the flux will not cross the boundary. If you study the weak form of Poisson equation you see that what pops out is the normal flux and for that you can only set BCs. Both these methods involving SaveLine relate to the flux over surface. There is no option to obtain flux(es) along tangential components. You might use "FluxSolver" to compute the heat flux involving all components.

-Peter
zeljkoT
Posts: 24
Joined: 14 Nov 2018, 12:55
Antispam: Yes

Re: SaveLine tangential component

Post by zeljkoT »

Hi Peter

Right, thank you for the explanation. Only the normal component, understood.

But just to close the first loop, let me ask again what did you mean with the "nodal weight" ?:
If you want to know the charge distribution then divide this by the nodal weight. See test case "WeightComputation".
What do I need to divide with?

thx a lot in advance.
raback
Site Admin
Posts: 4832
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: SaveLine tangential component

Post by raback »

Hi

You divide the "varname load" (provided as a library feature) with the "boundary weight" to go from discrete to distributed flux.

-Peter
Post Reply