Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Infinite elements problem in AC/DC modul

Please login with a confirmed email address before reporting spam

Hi all,

I have problem concerning of "Infinite
Elements" in AC/DC modul. I know there are couple of discussion about this but this is still unclear to me. I have read the manual and I followed the instructions.

It is desired to make just a simple 2D model of the wire with high
potencial. The surrounding material is the air. Around the air the
Infinite Elements are imposed. I have printed a line graph of the norm of
electric field (the line goes from the edge of the wire to the external
edge of the model in radial direction). In the enclosured pdf file you can
find the diffrencies between the solution with and without the infinite
elements. The meshing can also be seen.

There are two basic problems: 1-The continuity of the curve, 2-The maximum
value of the norm of the electric field - it is about 4(5) times higher if
you don't use the infinite elements layer.

So, please explain to me where the mistake could be? Is it in bad Infinite
elements understanding or is their using just unsuitable for these kind of
problems?

What is realy interesting, as you could notice in the enclosure file, the
same problem appears also in the example model from the MODEL LIBRARY: "parallel wires". I have
made a similar line graph (for the norm of the magnetic field) and you can
see the similar discontinuity. Without infinite elements the line is
perfectly smooth. However, there is not so big difference between the
maximal values as in my model...

Please could somebody to clear up to me these kind of problems?

Thank you very much!

Katarina


8 Replies Last Post 31 ago 2013, 05:13 GMT-4
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 3 feb 2011, 06:53 GMT-5
Hi

for me the discontiunuity is normnal, as "infinite" elements has anpother mapping of "r".

A few precautions to consider with infinite elements, that I have noticed:
1) be sure you use cylindrical shapes and that the origine of the expansion for the Infinite elements is well in the nice symmetric shape of your model 0,0 is NOT always the right position, as you part might not always be around the origine
2) I would use less elements slightly smaller thickness and use block elements for the infinite element region (but I must admit I have forgotten the rationale why, it must be somewhere in one of the ACDC courses of COMSOL)

In 4.1 you have the egoemtry "layer" to make such an onion layer for msot geoemtrical shapes

I agre that the maximum value at the inner circle should be the same for each case, are you really imposing a voltage ? if not it might be linked to a resistivity/current issue

--
Good luck
Ivar
Hi for me the discontiunuity is normnal, as "infinite" elements has anpother mapping of "r". A few precautions to consider with infinite elements, that I have noticed: 1) be sure you use cylindrical shapes and that the origine of the expansion for the Infinite elements is well in the nice symmetric shape of your model 0,0 is NOT always the right position, as you part might not always be around the origine 2) I would use less elements slightly smaller thickness and use block elements for the infinite element region (but I must admit I have forgotten the rationale why, it must be somewhere in one of the ACDC courses of COMSOL) In 4.1 you have the egoemtry "layer" to make such an onion layer for msot geoemtrical shapes I agre that the maximum value at the inner circle should be the same for each case, are you really imposing a voltage ? if not it might be linked to a resistivity/current issue -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 3 feb 2011, 07:23 GMT-5
Hi Ivar!

The mapping in infinite area was the last try (as for meshing) I have done, because there was a discontinuity with FreeTriangulars as well. And in AC/DC Users guide I have read that it is strongly recomended sweep (or in 2D mapped? I suppose) meshing, so I tried....

1) I am sure there are cylindrical shapes. I have picked this choice. And the model is centered in 0,0 as well.

2) I have already used less amount of elements and there was always the discontinuity (but less visible, because of the course mesh ;) ). But anyway, I will try to make the infinite area smaller and try...

The potencial at inner circle is still the same (2000V), I have just disabled the infinite element layer.
What do you mean with "resistivity/current issue" ?

Btw., how can I easily do the "onion" layer in 4.1? I used differences in Boolean operations....

Thank you for the reply and greetings! Was nice to see you in Paris;)
Hi Ivar! The mapping in infinite area was the last try (as for meshing) I have done, because there was a discontinuity with FreeTriangulars as well. And in AC/DC Users guide I have read that it is strongly recomended sweep (or in 2D mapped? I suppose) meshing, so I tried.... 1) I am sure there are cylindrical shapes. I have picked this choice. And the model is centered in 0,0 as well. 2) I have already used less amount of elements and there was always the discontinuity (but less visible, because of the course mesh ;) ). But anyway, I will try to make the infinite area smaller and try... The potencial at inner circle is still the same (2000V), I have just disabled the infinite element layer. What do you mean with "resistivity/current issue" ? Btw., how can I easily do the "onion" layer in 4.1? I used differences in Boolean operations.... Thank you for the reply and greetings! Was nice to see you in Paris;)

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 3 feb 2011, 13:43 GMT-5
Hi

The discontinuity in the Inf layer is "logic", in my mind (but I do not have the equations to say so). And that the slope in the air volume changes (with or without INF layers) I also find normal.
But the voltage level at the inner ring should not change, so something else must be wrong somewhere, not sure like that what it could be, and do not have access to COMSOL for some days

I was woundering if you had a model where the voltage was defined solely by a current flow, also before the inner ring i.e. a spice extension, that could have explained why the absolute value at the ring seem to change so much

Indeed we meet many people, and learn a lot, at the COMSOL conferences, time to think about this years conference too ;)

--
Good luck
Ivar
Hi The discontinuity in the Inf layer is "logic", in my mind (but I do not have the equations to say so). And that the slope in the air volume changes (with or without INF layers) I also find normal. But the voltage level at the inner ring should not change, so something else must be wrong somewhere, not sure like that what it could be, and do not have access to COMSOL for some days I was woundering if you had a model where the voltage was defined solely by a current flow, also before the inner ring i.e. a spice extension, that could have explained why the absolute value at the ring seem to change so much Indeed we meet many people, and learn a lot, at the COMSOL conferences, time to think about this years conference too ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 4 feb 2011, 07:05 GMT-5
Ivar,

maybe I just see why it is discontinous. It is a kind of region where the outer boundary is placed in infinity and we see just a short length, so it has to be seen in this way.

But I still don't understand this difrence between the maximum values... There is just a potencial 2000 at the inner circle and the outer potencial is set to 0. Nothing else...

Have to do the analytical solution.

If you have some other guesses, please share ;)

Greeting, Kate
Ivar, maybe I just see why it is discontinous. It is a kind of region where the outer boundary is placed in infinity and we see just a short length, so it has to be seen in this way. But I still don't understand this difrence between the maximum values... There is just a potencial 2000 at the inner circle and the outer potencial is set to 0. Nothing else... Have to do the analytical solution. If you have some other guesses, please share ;) Greeting, Kate

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 4 feb 2011, 08:22 GMT-5
Hi
I agree, discontinuity is understandable, but there should not be any significant differences in the height.

Now what are the electric potential at infinitiy, or at the boundary without the inf value ? your system being axisymmetric you should be able to define it as a 1D-axi case. By having a magnetic isolation at the external boundary you have a global potential, how is this linked to a ground in yuor model ?

I'll exepect a 1/R^2 decay of E, no?, in the radial direction

I would have to fetch my books again, but today I'm 2000km away from my bookshelf, to far off to reach it, what about a look at i.e. en.wikipedia.org/wiki/Electric_field

--
Good luck
Ivar
Hi I agree, discontinuity is understandable, but there should not be any significant differences in the height. Now what are the electric potential at infinitiy, or at the boundary without the inf value ? your system being axisymmetric you should be able to define it as a 1D-axi case. By having a magnetic isolation at the external boundary you have a global potential, how is this linked to a ground in yuor model ? I'll exepect a 1/R^2 decay of E, no?, in the radial direction I would have to fetch my books again, but today I'm 2000km away from my bookshelf, to far off to reach it, what about a look at i.e. http://en.wikipedia.org/wiki/Electric_field -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 10 feb 2011, 08:51 GMT-5
Hi, sorry for late answer. I have some ideas about these problems now. So I would like to share it. It seems to be perfectly logic, but I just have looked at it from another point of view before...

As for the "knee" in the graph:
The infinite element domain is effectively scaled geometry and the outer boundary lies at infinity.
The effective radial coordinate inside the infinite elements grows much faster than the geometric radial coordinate. So we are moving in another scale.
And that also affects the value of the field. Because it is different if you put zero potencial to infinity or if you put it to much closer point, let's say 1um from the wire (d=0.05um)... So it is completely different elecetric field. It should be stronger, if you put zero potencial closer to the wire.

Btw. I would expect also the decay of 1/r^2, but if I solve it analyticaly, not even the possibility of the charged cylinder (1/r) even sphere (1/r^2) doesn't match with the comsol calculation. That still remains to be a mystery for me...
(because, I mean, 2D model of this kind of geometry could be both a cylinder or a sphere...)

ciao, katerina
Hi, sorry for late answer. I have some ideas about these problems now. So I would like to share it. It seems to be perfectly logic, but I just have looked at it from another point of view before... As for the "knee" in the graph: The infinite element domain is effectively scaled geometry and the outer boundary lies at infinity. The effective radial coordinate inside the infinite elements grows much faster than the geometric radial coordinate. So we are moving in another scale. And that also affects the value of the field. Because it is different if you put zero potencial to infinity or if you put it to much closer point, let's say 1um from the wire (d=0.05um)... So it is completely different elecetric field. It should be stronger, if you put zero potencial closer to the wire. Btw. I would expect also the decay of 1/r^2, but if I solve it analyticaly, not even the possibility of the charged cylinder (1/r) even sphere (1/r^2) doesn't match with the comsol calculation. That still remains to be a mystery for me... (because, I mean, 2D model of this kind of geometry could be both a cylinder or a sphere...) ciao, katerina

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 10 feb 2011, 14:52 GMT-5
Hi

I follow and agree on most, apart from the ending sentence, in 2D its not a cylinder or a sphere that you are looking at, that would be 2D-axi, with a loop length of 2*pi*r. Quite a different topology from the 2D of 1[m] long (by default) section. (ok if you draw a circle in 2D its a long cylinder you are looking down the axis, and not perpendicular to the axis as in 2D-axi)

--
Good luck
Ivar
Hi I follow and agree on most, apart from the ending sentence, in 2D its not a cylinder or a sphere that you are looking at, that would be 2D-axi, with a loop length of 2*pi*r. Quite a different topology from the 2D of 1[m] long (by default) section. (ok if you draw a circle in 2D its a long cylinder you are looking down the axis, and not perpendicular to the axis as in 2D-axi) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 31 ago 2013, 05:13 GMT-4
Hi, all
I realize Infinite Element as its radius increases or decreases compared with a charged body, the calculation result is changed. In this case, the radius of how many times it is often compared to simulation objects?
Hi, all I realize Infinite Element as its radius increases or decreases compared with a charged body, the calculation result is changed. In this case, the radius of how many times it is often compared to simulation objects?

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.