Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Problem in solving eddy current in micromagnetic model

Please login with a confirmed email address before reporting spam

Hello there,
i am trying to investigate the effects of an oscillating permanent magnet
over a copper plate using comsol and the ac/dc module. After reducing the
degree of freedom and the model from a 3D to a 2D axialsymmetric there
still are several problems on solving this model. I expect an eddy current
in the plate (and further an re-induced current in the magnet, which
should damp the oscillation - this seems to me utopic at the moment). The
model seems to be solved correct, but the problems so far are:
-While the magnet moves out of his initial condition there seems to be an
undefined zone between the boundaries and the moving mesh/analyzed area.
What is it about?

-The area (area 2) that should be vacuum characteristic is not been
solved, only painted white. Here there should be no induced current -->
green!
-A general problem seems to me the function of the animation, which is
often not comparable with the solutions at different times in the "plot
parameter"
-I am not a brutal beginner in comsol, but would be thankful for an
general assesment of the model or an improvement suggestion!

A rapid response would be appreciated.


Yours sincerly


5 Replies Last Post 3 févr. 2011, 10:10 UTC−5
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 6 janv. 2011, 04:52 UTC−5
Hi

Depending on what you are looking at (reference frame or ALE) you will see the geometry move and still have the background initial wire-frame lines present, I suppose this is what you observe as "gap", if not its indeed probably an issue with a badly constrained BC. And then you should block the external boundaries and impose the oscillation on the triangular Boundary too, that is perhaps generating your gaps ?

I agree that the animation is always a bit tricky to get to synchronized with the solver sequence, you need to play with the settings (I cannot remember anymore which one to propose ;)

There seem to be a nice buildup of Eddy currents during the first rises then it dies out, you still see the delayed sign changes. Is this related to the rather high frequency you are using for your oscillations ? Have you tried a frequency parametric scan ?

Have you checked the skin depth w.r. the 20 um layer ? it looks 2-3 times too thin

for me copper has a conductivity rather around 6E7 and not 1E6 but that should not change the main issue I agree

So probably its the missing imposed motion moving magnet BC in the ALE that are bothering you

And one more ;)
by having an oscillating driver function and a transient time stepper, you should use an intermediate Time stepping - Time-step and not the "free" one as then you will loose the oscillations rather rapidly.
The free transient solver expects a smooth type asymptotic solution convergence and no oscillations. This could well explain why your original solution "smears" out => wrong: its really the initial turn on pulse as you start with a step with your sinus (no 2nd order derivative smoothing) so the first swing seem to be numerical. Pls verify carefully

using a 1-cos() and starting with a motion in one direction from "0" will probably take away this initial peak, also a stationary solver to set-up the fields before the first time step, and use that as initial conditions could also help

--
Good luck
Ivar
Hi Depending on what you are looking at (reference frame or ALE) you will see the geometry move and still have the background initial wire-frame lines present, I suppose this is what you observe as "gap", if not its indeed probably an issue with a badly constrained BC. And then you should block the external boundaries and impose the oscillation on the triangular Boundary too, that is perhaps generating your gaps ? I agree that the animation is always a bit tricky to get to synchronized with the solver sequence, you need to play with the settings (I cannot remember anymore which one to propose ;) There seem to be a nice buildup of Eddy currents during the first rises then it dies out, you still see the delayed sign changes. Is this related to the rather high frequency you are using for your oscillations ? Have you tried a frequency parametric scan ? Have you checked the skin depth w.r. the 20 um layer ? it looks 2-3 times too thin for me copper has a conductivity rather around 6E7 and not 1E6 but that should not change the main issue I agree So probably its the missing imposed motion moving magnet BC in the ALE that are bothering you And one more ;) by having an oscillating driver function and a transient time stepper, you should use an intermediate Time stepping - Time-step and not the "free" one as then you will loose the oscillations rather rapidly. The free transient solver expects a smooth type asymptotic solution convergence and no oscillations. This could well explain why your original solution "smears" out => wrong: its really the initial turn on pulse as you start with a step with your sinus (no 2nd order derivative smoothing) so the first swing seem to be numerical. Pls verify carefully using a 1-cos() and starting with a motion in one direction from "0" will probably take away this initial peak, also a stationary solver to set-up the fields before the first time step, and use that as initial conditions could also help -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 janv. 2011, 07:22 UTC−5
Hello,
At first i want to thank you for the rapid response in this matter. So far
i have updated my model, but there still are some open-end questions.

-How do I solve at first the static magnetic field and afterwards the
transient moving ALE? Maybe this would let the unresolved problems
dissapear
-How do I make a frequency parametric scan? I didnt find something about
this in the comsol guides. Is there any link/description?
-There are still not the eddy currents which i expect to emerge. At a time
step it seems to be realistic but if you look on the development over the
imposed times, the eddy current oscillte with a higher frequency than the
magnet does.
-I got the advice from the Support that my model is not in the ALE-Frame.
How do I have to understand this? Support has not responded so far...
-And again: I would be very, very grateful for ideas for improvement and a
general assessment of my model. =)

King regards

Petro
Hello, At first i want to thank you for the rapid response in this matter. So far i have updated my model, but there still are some open-end questions. -How do I solve at first the static magnetic field and afterwards the transient moving ALE? Maybe this would let the unresolved problems dissapear -How do I make a frequency parametric scan? I didnt find something about this in the comsol guides. Is there any link/description? -There are still not the eddy currents which i expect to emerge. At a time step it seems to be realistic but if you look on the development over the imposed times, the eddy current oscillte with a higher frequency than the magnet does. -I got the advice from the Support that my model is not in the ALE-Frame. How do I have to understand this? Support has not responded so far... -And again: I would be very, very grateful for ideas for improvement and a general assessment of my model. =) King regards Petro


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 janv. 2011, 11:00 UTC−5
I assume you have the same problem like me (see my question). Your magnetic field (arrows) does not change in the plot. I think it is not recalculated for every step of the moving mesh. I tried to solve that with an assembly / identity pairs. But still it doesn't work.

Does anybody have any clue how to solve this?

Clemens
I assume you have the same problem like me (see my question). Your magnetic field (arrows) does not change in the plot. I think it is not recalculated for every step of the moving mesh. I tried to solve that with an assembly / identity pairs. But still it doesn't work. Does anybody have any clue how to solve this? Clemens

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 janv. 2011, 09:24 UTC−5
Hi

I'm wondering now, to solve for the Eddy currents do you not need the Lorentz velocity terms, which means you need to solve for "V" too ?

An I'm noticing: I'm have trouble now "reverting "to 3.5 ;)
--
Good luck
Ivar
Hi I'm wondering now, to solve for the Eddy currents do you not need the Lorentz velocity terms, which means you need to solve for "V" too ? An I'm noticing: I'm have trouble now "reverting "to 3.5 ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 3 févr. 2011, 10:10 UTC−5
Thank you Ivar! Aren't the eddy currents (should be in phi) already considered in Aphi (azimuthal induction currents, vector potential)? Do you have an idea why the magnetic field is constant regardless the relative position?

Thanks again!

C
Thank you Ivar! Aren't the eddy currents (should be in phi) already considered in Aphi (azimuthal induction currents, vector potential)? Do you have an idea why the magnetic field is constant regardless the relative position? Thanks again! C

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.