Menu

Discussion Forum

New Discussion
Filter

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 The information provided may be out of date.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.

A problem in electromagnetics.it seems to ba a bug of COMSOL

Posted Jan 14, 2015, 9:05 a.m. EST Low-Frequency Electromagnetics Version 4.4 16 Replies

Wenzhi Yang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

As we all know, if there is a current in conductor, there will be a static magnetic field around it.If i simulate it using a 3D numerical model， i can get the static magnetic field. However, when i simulate it by a 2D axisymmetric model to reduce the computation , the magnetic field equal 0 ! Both model use the "elecric and magnetic field"and the boundary condition is same. WISH SOMEONE CAN HELP ME .THANKS A LOT.
16 Replies Last Post Jan 21, 2015, 6:28 a.m. EST
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 14, 2015, 10:43 a.m. EST
Hi

if your single domain is "the" conductor you will not see much, until you put some "air" around it, and you connect the current such that it can loop around to close both electric and magnetic fields.

So I'll say this is not a "bug" just that your Maxwell equations cannot be solved with those domains and boundaries of your model

--
Good luck
Ivar
Hi if your single domain is "the" conductor you will not see much, until you put some "air" around it, and you connect the current such that it can loop around to close both electric and magnetic fields. So I'll say this is not a "bug" just that your Maxwell equations cannot be solved with those domains and boundaries of your model -- Good luck Ivar
Wenzhi Yang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 14, 2015, 10:02 p.m. EST
thanks very much about your attention.However,i have built a similar numerical model in 3D and got the magnetic field, so there is no need for the "air" and a comlete current circuit .Also ,i have bulit this 2D axisymmetric model using COMSOL 3.5, and there is a B_phi which is what i need.
thanks very much about your attention.However,i have built a similar numerical model in 3D and got the magnetic field, so there is no need for the "air" and a comlete current circuit .Also ,i have bulit this 2D axisymmetric model using COMSOL 3.5, and there is a B_phi which is what i need.
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 15, 2015, 3:59 a.m. EST
Hi Wenzhi,

I have to second Ivar that your setup looks unphysical, at least in FE terms. You might consider to upload your models, to let us see the details of your settings. You should clear solutions and meshes to keep it small.

And maybe you want to tell us more about the objectives of this work.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Hi Wenzhi, I have to second Ivar that your setup looks unphysical, at least in FE terms. You might consider to upload your models, to let us see the details of your settings. You should clear solutions and meshes to keep it small. And maybe you want to tell us more about the objectives of this work. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Wenzhi Yang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 15, 2015, 9:09 a.m. EST
hi，
Thanks so much about your reply. I have up the model about 3D and 2D axisymmtric. I bulit a very simple physical model . Just give one side a electric ptential, the other side is ground. The magnetic field is what should be care about. But the results is different, and in 2D axisymmtric model we cannot get the B in phi direction.
thanks！
wenzhi yang
hi， Thanks so much about your reply. I have up the model about 3D and 2D axisymmtric. I bulit a very simple physical model . Just give one side a electric ptential, the other side is ground. The magnetic field is what should be care about. But the results is different, and in 2D axisymmtric model we cannot get the B in phi direction. thanks！ wenzhi yang

Attachments:

Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 15, 2015, 5:29 p.m. EST
Wenzhi,

I had a look into your models and although the setup is a bit unusual, I think the 3D solution is correct. The potential and ground BCs should work because they are at the model border.

As far as I see, the 2D axial model should be equivalent. I would suggest to forward this to the COMSOL support.

Would be interesting what the final result will be.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Wenzhi, I had a look into your models and although the setup is a bit unusual, I think the 3D solution is correct. The potential and ground BCs should work because they are at the model border. As far as I see, the 2D axial model should be equivalent. I would suggest to forward this to the COMSOL support. Would be interesting what the final result will be. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Henrik Sönnerlind COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 16, 2015, 2:16 a.m. EST
Hi,

You are using the option "Out-of-plane vector potential".

Since the B-field in your model is in the azimuthal direction, you must use 'In-plane vector potential' or 'Three-component vector potential'. In addition, you probably need to add 'Gauge Fixing for A-Field'.

Regards,
Henrik
Hi, You are using the option "Out-of-plane vector potential". Since the B-field in your model is in the azimuthal direction, you must use 'In-plane vector potential' or 'Three-component vector potential'. In addition, you probably need to add 'Gauge Fixing for A-Field'. Regards, Henrik
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 16, 2015, 5:05 a.m. EST

Hi Henrik,

when I checked Wenzhi's 2D model I also changed the vector component setting to 3 component but didn't think about the Gauge fixing, although the solver message is quite typical for it.
This brings me to the question what criteria decide whether the Gauge Fixing is required or not. I had quite a couple of discussions about that, also with COMSOL people during trainings, but there was no clear answer for it. It remains a trial and error thing. And even when the solver converges one can't be sure about the validity of the result. Then looking at the E-field or A-field sometimes shows a noisy picture that indicates the Gauge Fixing might be required.
Of course one would like to avoid it, particularly in the 3D case, because it doubles the memory demand.

Are there clear criteria? Or do we need to try and fail and finally succeed?

Thank you
Edgar
--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Hi Henrik, when I checked Wenzhi's 2D model I also changed the vector component setting to 3 component but didn't think about the Gauge fixing, although the solver message is quite typical for it. This brings me to the question what criteria decide whether the Gauge Fixing is required or not. I had quite a couple of discussions about that, also with COMSOL people during trainings, but there was no clear answer for it. It remains a trial and error thing. And even when the solver converges one can't be sure about the validity of the result. Then looking at the E-field or A-field sometimes shows a noisy picture that indicates the Gauge Fixing might be required. Of course one would like to avoid it, particularly in the 3D case, because it doubles the memory demand. Are there clear criteria? Or do we need to try and fail and finally succeed? Thank you Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Wenzhi Yang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 16, 2015, 7:45 a.m. EST
oh, I get it. It's so nice you give me this advice. Thank you very much.
oh, I get it. It's so nice you give me this advice. Thank you very much.
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 16, 2015, 2:49 p.m. EST
Hi

isn't the requirement for Gauge A-field fixing very similar to the cases of defining at least a "small" spring constant on a structural model if there is no other "fixed" BC, or defining at least one fixed p=0 pressure point in fluid model.
Working one "stage up" in the integral domain (i.e. A field), one add a "floating" constant to the dependent variables, that one need to strap down somewhat.

So for me a careful BC, number of dependent variables and unknowns study is required to identify when a Gauge fixing point/node/equation is needed or not

I must take a closer look at a solution of a B field without "air", that puzzles me, except if there is enough other material to allow to close the B loop ;)

--
Good luck
Ivar
Hi isn't the requirement for Gauge A-field fixing very similar to the cases of defining at least a "small" spring constant on a structural model if there is no other "fixed" BC, or defining at least one fixed p=0 pressure point in fluid model. Working one "stage up" in the integral domain (i.e. A field), one add a "floating" constant to the dependent variables, that one need to strap down somewhat. So for me a careful BC, number of dependent variables and unknowns study is required to identify when a Gauge fixing point/node/equation is needed or not I must take a closer look at a solution of a B field without "air", that puzzles me, except if there is enough other material to allow to close the B loop ;) -- Good luck Ivar
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 16, 2015, 4:32 p.m. EST

Hi Ivar,

you are probably right and there is no simple straightforward criterion for the Gauge Fixing requirement. I guess I am not doing enough analysis.

Regarding Wenzhi's model, the current has only a z-component, so we get a purely azimutal B-field, that can well exist with the given BCs.
I am not sure if this could be realized. It might be something like a normal conducting wire embedded into a superconductor. Wouldn't this confine the field inside the embedded wire? Who knows what Wenzhi is inventing there ;-).

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Hi Ivar, you are probably right and there is no simple straightforward criterion for the Gauge Fixing requirement. I guess I am not doing enough analysis. Regarding Wenzhi's model, the current has only a z-component, so we get a purely azimutal B-field, that can well exist with the given BCs. I am not sure if this could be realized. It might be something like a normal conducting wire embedded into a superconductor. Wouldn't this confine the field inside the embedded wire? Who knows what Wenzhi is inventing there ;-). Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Zhi Zhang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 18, 2015, 10:30 p.m. EST
Excuse me, I to simulate the electromagnetic induction heating, the current is 35 [A], current frequency is 35 KHZ.What physical fields should choose?
Excuse me, I to simulate the electromagnetic induction heating, the current is 35 [A], current frequency is 35 KHZ.What physical fields should choose?
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 19, 2015, 2:07 a.m. EST
Hi

At 35 kHz your wavelength is 3E8[m/s]/35E3[Hz] ~30 km so most probably your device is far less in size, hence the ACDC module is appropriate. Then as you have current flowing and magnetic fields interacting I would propose to use MF (MEF would add more complexity and solve slower).

The best is to study the 2-3 first ACDC Model library examples

--
Good luck
Ivar
Hi At 35 kHz your wavelength is 3E8[m/s]/35E3[Hz] ~30 km so most probably your device is far less in size, hence the ACDC module is appropriate. Then as you have current flowing and magnetic fields interacting I would propose to use MF (MEF would add more complexity and solve slower). The best is to study the 2-3 first ACDC Model library examples -- Good luck Ivar
Jens Krause
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 19, 2015, 3:58 a.m. EST
Hallo Edgar and others,

on the subject of gauge-fixing in 3D I have the following strategy: I think that in principle you need it
because the stiffness matrix is singular without it. However when using an iterative solver there is a
preconditioner called "SOR gauge" that finds a valid solution even without gauge-fixing.
Hence I always use gauge-fixing, when using direct solvers or different preconditioners.

Regards

Jens
Hallo Edgar and others, on the subject of gauge-fixing in 3D I have the following strategy: I think that in principle you need it because the stiffness matrix is singular without it. However when using an iterative solver there is a preconditioner called "SOR gauge" that finds a valid solution even without gauge-fixing. Hence I always use gauge-fixing, when using direct solvers or different preconditioners. Regards Jens
Wenzhi Yang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 21, 2015, 5:06 a.m. EST
hi Edgar, Ivar and others,
Thanks again for your help before. I am simulating a flow influenced by Lorentz Force and have a problem. Would you mind spare a little time to give me some advice ?
Liquid metal is good conductor. When i put a current into the liquid metal flow, a magnetic field will be induced and the Lorentz force thereof will influence the flow.
Because “magnetic and electric fields” is not suitable for a time-dependent flow , the COMSOL support had adised me to replace it with "magnetic fields"coupled with"electric currents". However, the results( magnetic B ) are completely different.
For simplication, i give a simplified model below.
Thanks.
wenzhi yang



hi Edgar, Ivar and others, Thanks again for your help before. I am simulating a flow influenced by Lorentz Force and have a problem. Would you mind spare a little time to give me some advice ? Liquid metal is good conductor. When i put a current into the liquid metal flow, a magnetic field will be induced and the Lorentz force thereof will influence the flow. Because “magnetic and electric fields” is not suitable for a time-dependent flow , the COMSOL support had adised me to replace it with "magnetic fields"coupled with"electric currents". However, the results( magnetic B ) are completely different. For simplication, i give a simplified model below. Thanks. wenzhi yang
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 21, 2015, 5:39 a.m. EST
Hi Wenzhi,

you didn't couple mf and ec. You have to add an External Current Density to mf that takes the current density calculated in ec.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com
Hi Wenzhi, you didn't couple mf and ec. You have to add an External Current Density to mf that takes the current density calculated in ec. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology http://www.emphys.com
Wenzhi Yang
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago Jan 21, 2015, 6:28 a.m. EST
oh,nice. You are so great ! Thank you very much.
oh,nice. You are so great ! Thank you very much.

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.

Suggested Content