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 The information provided may be out of date.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.

[Please help] Magnetic-structural coupled analysis, result not good,model attached

Please login with a confirmed email address before reporting spam

This is part of my thesis. I'm trying to simulate the magneto-structural coupled analysis on a magnetic gear. The attached file is partial of the model. If you have any advice, please see the model and give me a reply.:) Thanks.

The cylinder consists of permanent magnets and steel, which can be treated as a rotor. The single bar is also made of steel. It is static analysis. I just want to modulate the maximum of the single steel rod's deflection.

The three physics I used for coupled analysis is : AC/DC no current, Structural, ALE.

The FEA result on steel rod deflection is 10% of the experimental result. I'm wondering if I'm setting the wrong mesh.

Furthermore, I want to do the coupled analysis manually, yet I've not figured out how to make it.

Thanks.


13 Replies Last Post 21 juin 2016, 19:31 UTC−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 18 avr. 2013, 14:07 UTC−4
Hi

Clear the mesh can be a cause, but also do you have enough free "air" around to allow the field lines to loop nicely ? that can easily give factors of "2".
But there are other reasons too, it could be caused from sharp corners on the iron or magnets, these make numerical singularities, and when you integrate the forces over the boundaries, you can easily get very wrong results. To avoid this use filleted corner, when possible, but this also increases the mesh density hece time to solution.

It's good to be able to compare measurements with the FEM theory, as like that you learn a lot about traps you should stear around ;)

--
Good luck
Ivar
Hi Clear the mesh can be a cause, but also do you have enough free "air" around to allow the field lines to loop nicely ? that can easily give factors of "2". But there are other reasons too, it could be caused from sharp corners on the iron or magnets, these make numerical singularities, and when you integrate the forces over the boundaries, you can easily get very wrong results. To avoid this use filleted corner, when possible, but this also increases the mesh density hece time to solution. It's good to be able to compare measurements with the FEM theory, as like that you learn a lot about traps you should stear around ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 8 mai 2013, 14:11 UTC−4
Hi Ivar,

Thanks for your advice.:)

I tried to simplify my model and cut the geometry into half along the symmetry boundary. Then I tried to refine the element size. The result is strange, almost 10% of the experimental data.

1. Increase the volume air that surrounding the whole part does not truly increase the deformation of the part.

2. Change the boundary condition from beam fixed end to simple supported does not change the result. What I did is change the fixed condition(x=0, y=0, z=0) to prescribed displacement(x=0, y=0, z is not zero)

3. I doubt the ALE was wrong. I only set the air to be deformable. The single bar has deflection due to the magnetic force. I think the single bar is also deformable, when setting ALE. I'm trying to running this. Should I use "deformed geometry" instead of ALE? I found the settings are almost the same.

4. Is there any way to do one way coupling analysis manully. For example, 1st step: Couple magnetic force to Solid mechanics model. 2nd step: Couple deformation in Solid mechanics model with AC/DC model. THEN repeat the former steps as iteration. I mean if I can ignore the ALE and use something else to express the change of mesh.

Thanks for your reply:)
Hi Ivar, Thanks for your advice.:) I tried to simplify my model and cut the geometry into half along the symmetry boundary. Then I tried to refine the element size. The result is strange, almost 10% of the experimental data. 1. Increase the volume air that surrounding the whole part does not truly increase the deformation of the part. 2. Change the boundary condition from beam fixed end to simple supported does not change the result. What I did is change the fixed condition(x=0, y=0, z=0) to prescribed displacement(x=0, y=0, z is not zero) 3. I doubt the ALE was wrong. I only set the air to be deformable. The single bar has deflection due to the magnetic force. I think the single bar is also deformable, when setting ALE. I'm trying to running this. Should I use "deformed geometry" instead of ALE? I found the settings are almost the same. 4. Is there any way to do one way coupling analysis manully. For example, 1st step: Couple magnetic force to Solid mechanics model. 2nd step: Couple deformation in Solid mechanics model with AC/DC model. THEN repeat the former steps as iteration. I mean if I can ignore the ALE and use something else to express the change of mesh. Thanks for your reply:)

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 8 mai 2013, 14:26 UTC−4
Hi

it must be something in the way you have assembled the model, not evident to say more without opening it up and studying it ;)

Certainly you can solve physics by physics, one way is to make sequential solver steps and restriang one physics at the time, and link the next solver to the results from the previous one, you can also then separate the initial values reading for variables solved for and those not solved for. This is basically what the segregated solver is doing, but then running in an internal loop

--
Good luck
Ivar
Hi it must be something in the way you have assembled the model, not evident to say more without opening it up and studying it ;) Certainly you can solve physics by physics, one way is to make sequential solver steps and restriang one physics at the time, and link the next solver to the results from the previous one, you can also then separate the initial values reading for variables solved for and those not solved for. This is basically what the segregated solver is doing, but then running in an internal loop -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 8 mai 2013, 14:36 UTC−4
Hi Ivar,

Thanks for your reply.:)

You may see the attached file, which is partial of my model. Do you have any examples on the solving physics by physics?

Thanks again.:)

Dan
Hi Ivar, Thanks for your reply.:) You may see the attached file, which is partial of my model. Do you have any examples on the solving physics by physics? Thanks again.:) Dan


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 9 mai 2013, 15:10 UTC−4
Add one moe thing, I tried the simulation with changing ALE(change mesh) to have air and the single bar. The result is still not good. Anyone has any suggestions?

Thanks.

Dani
Add one moe thing, I tried the simulation with changing ALE(change mesh) to have air and the single bar. The result is still not good. Anyone has any suggestions? Thanks. Dani

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 11 mai 2013, 23:32 UTC−4

Hi Ivar,

Thanks for your reply.:)

You may see the attached file, which is partial of my model. Do you have any examples on the solving physics by physics?

Thanks again.:)

Dan


Hi Ivar,

Would you mind taking a look at the model? Thanks.

Dani
[QUOTE] Hi Ivar, Thanks for your reply.:) You may see the attached file, which is partial of my model. Do you have any examples on the solving physics by physics? Thanks again.:) Dan [/QUOTE] Hi Ivar, Would you mind taking a look at the model? Thanks. Dani

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 13 mai 2013, 18:02 UTC−4
please see the attached model and do me a favor. Thanks so much.
please see the attached model and do me a favor. Thanks so much.

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 13 mai 2013, 18:30 UTC−4
Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation.

Thanks
Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation. Thanks

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 13 mai 2013, 18:30 UTC−4
Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation.

Thanks
Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation. Thanks

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 13 mai 2013, 18:30 UTC−4
Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation.

Thanks
Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation. Thanks

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 14 mai 2013, 13:39 UTC−4
Have you got any updated news?
Have you got any updated news?

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 29 mai 2013, 10:29 UTC−4

Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation.

Thanks


I think I worked out the problem by changing the ALE conditions. You may try to change the prescribed displacement from u to X+u, v to Y+v, w to Z+w.
[QUOTE] Im having the exact same problem... I also would be really thankful if someone can take a look at the simulation. Thanks [/QUOTE] I think I worked out the problem by changing the ALE conditions. You may try to change the prescribed displacement from u to X+u, v to Y+v, w to Z+w.

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 21 juin 2016, 19:31 UTC−4
Hello,

I have seen your question/model and the idea of the work you have for your thesis.
I am doing my master thesis as well and I have a problem with magneto structural coupling that probably you could help me in some way!
Would you be available? Or any way to contact you?

Best regards
Hello, I have seen your question/model and the idea of the work you have for your thesis. I am doing my master thesis as well and I have a problem with magneto structural coupling that probably you could help me in some way! Would you be available? Or any way to contact you? Best regards

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.