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.

Modeling Electromagnet With Different Materials (AC/DC Module)

Please login with a confirmed email address before reporting spam

I am trying to simulate a DC electromagnet using the AC/DC module. It is a C-shaped core with a multiturn coil around one segment of the core. I am trying to achieve a 0.2T+-20% field in a cylindrical region between the ends of the core.

I was under the impression that I was doing everything correctly until today. I am using the HB curve from the core material for the magnetic field calculation. I got results using the built in Low Carbon Steel 1018. I then changed the material to Silicon Steel GO M-6 Rolling, and the magnetic field values barely changed. I was under the impression that I should get a larger magnetic field using the second material. I tried the simulation with a variety of different coil currents to ensure I wasn't just saturating both the materials, and still did not notice much of a difference between the materials.

These results don't make sense to me, which makes me wonder if I can trust any of my solutions. Do I have something wrong in my simulation set up? Am I misunderstanding the problem and have incorrect expectations?

Additionally, how can I achieve a view where I see the coil turns that COMSOL generates like so: cdn.comsol.com/release/51/elec...rical/acdc/newCoilGeometry.png

This is my first task using COMSOL, so there is a possibility that I'm missing something fundamental. Please let me know if I can provide more information to clarify my problem.


8 Replies Last Post 7 févr. 2017, 14:48 UTC−5

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 1 févr. 2017, 12:08 UTC−5
Hi Kyle,

With some designs, you will not notice much difference. Remember that the BH curves are derived from samples in a closed magnetic circuit. Your design has a fairly large airgap and the flux density is not homogenous through your steel cross section, so while your expectations are OK in theory, in practice with your design you may not see the benefits. You could test this theory out by designing a simple electromagnet with an almost closed magnetic circuit (small airgap) and look at the field in the gap at varying currents. The deltas between the results you get with the 2 materials should be more apparent.

As an aside, I have attached a model of an electromagnet that uses 1/4 symmetry (which you could use too in order to refine your mesh for the same solve time as you currently have), and also Infinite Element domains (these are used to better simulate the far field condition and will prevent enforcement of tangential flux density at your model boundary, which in your model may be impacting on your desired result). There may be something of use in there for you...

p.s. the attached model is v5.2a, not sure if you can open it?
Hi Kyle, With some designs, you will not notice much difference. Remember that the BH curves are derived from samples in a closed magnetic circuit. Your design has a fairly large airgap and the flux density is not homogenous through your steel cross section, so while your expectations are OK in theory, in practice with your design you may not see the benefits. You could test this theory out by designing a simple electromagnet with an almost closed magnetic circuit (small airgap) and look at the field in the gap at varying currents. The deltas between the results you get with the 2 materials should be more apparent. As an aside, I have attached a model of an electromagnet that uses 1/4 symmetry (which you could use too in order to refine your mesh for the same solve time as you currently have), and also Infinite Element domains (these are used to better simulate the far field condition and will prevent enforcement of tangential flux density at your model boundary, which in your model may be impacting on your desired result). There may be something of use in there for you... p.s. the attached model is v5.2a, not sure if you can open it?


Please login with a confirmed email address before reporting spam

Posted: 8 years ago 1 févr. 2017, 12:56 UTC−5
Thank you very much for the response. Makes sense to me. I will try a simplified geometry and small gap and see what happens.

I'm not 100% what you are describing in your second paragraph, but it sounds useful! However I cannot open the 5.2a file unfortunately. I'll do some googling and try and figure out what you mean.
Thank you very much for the response. Makes sense to me. I will try a simplified geometry and small gap and see what happens. I'm not 100% what you are describing in your second paragraph, but it sounds useful! However I cannot open the 5.2a file unfortunately. I'll do some googling and try and figure out what you mean.

Please login with a confirmed email address before reporting spam

Posted: 8 years ago 2 févr. 2017, 12:32 UTC−5
I've attempted to create a simplified geometry incorporating symmetry and IE domains. However, I ran
my 1/4 symmetry and a full symmetry case to ensure I am getting the same results for both cases, and found that the results are substantially different. I can't figure out where my problem lies. If you have an opportunity, it would be much appreciated if you would take a look at my 1/4 symmetry set up to see if I have made any obvious errors. I've attached the new geometry for both the full and 1/4 symmetry models.



Hi Kyle,

With some designs, you will not notice much difference. Remember that the BH curves are derived from samples in a closed magnetic circuit. Your design has a fairly large airgap and the flux density is not homogenous through your steel cross section, so while your expectations are OK in theory, in practice with your design you may not see the benefits. You could test this theory out by designing a simple electromagnet with an almost closed magnetic circuit (small airgap) and look at the field in the gap at varying currents. The deltas between the results you get with the 2 materials should be more apparent.

As an aside, I have attached a model of an electromagnet that uses 1/4 symmetry (which you could use too in order to refine your mesh for the same solve time as you currently have), and also Infinite Element domains (these are used to better simulate the far field condition and will prevent enforcement of tangential flux density at your model boundary, which in your model may be impacting on your desired result). There may be something of use in there for you...

p.s. the attached model is v5.2a, not sure if you can open it?


I've attempted to create a simplified geometry incorporating symmetry and IE domains. However, I ran my 1/4 symmetry and a full symmetry case to ensure I am getting the same results for both cases, and found that the results are substantially different. I can't figure out where my problem lies. If you have an opportunity, it would be much appreciated if you would take a look at my 1/4 symmetry set up to see if I have made any obvious errors. I've attached the new geometry for both the full and 1/4 symmetry models. [QUOTE] Hi Kyle, With some designs, you will not notice much difference. Remember that the BH curves are derived from samples in a closed magnetic circuit. Your design has a fairly large airgap and the flux density is not homogenous through your steel cross section, so while your expectations are OK in theory, in practice with your design you may not see the benefits. You could test this theory out by designing a simple electromagnet with an almost closed magnetic circuit (small airgap) and look at the field in the gap at varying currents. The deltas between the results you get with the 2 materials should be more apparent. As an aside, I have attached a model of an electromagnet that uses 1/4 symmetry (which you could use too in order to refine your mesh for the same solve time as you currently have), and also Infinite Element domains (these are used to better simulate the far field condition and will prevent enforcement of tangential flux density at your model boundary, which in your model may be impacting on your desired result). There may be something of use in there for you... p.s. the attached model is v5.2a, not sure if you can open it? [/QUOTE]


Please login with a confirmed email address before reporting spam

Posted: 8 years ago 6 févr. 2017, 07:23 UTC−5
Kyle,

Your geometry sequence seems overcomplicated. I have run a 1/4 symmetry model where I have disabled a large number of your geometry nodes to simplify matters. Also, your mesh in the airgap was too coarse to reliably resolve accurate field values.

For the 1/4 symmetry model I modelled the system using your default current and obtain a field at the centre of the airgap of about 0.76 mT. For the full geometry, I obtained the same.

I think you were getting different results because you have not divided the current by 2 in the symmetric version, (only half of the turns are modelled). The symmetry condition closes the coil electrically, but in terms of the coils axial length, with the 1/4 symmetry, you only have 1/2 of the coil, so the total current should be stated as half of the "true" total current.

I have attached a screenshot of the simplified geometry sequence. The highlighted region is used as a mesh control domain, (I limited the mesh size in this mesh control domain to 0.01[m].

Rather than having all of those features (the ones I have disabled), just create the raw geometry, then use a couple of "block/Boolean difference" features at the end to split the entire model into 4, much cleaner.

To analyse results, I much prefer to use cut planes, cut points etc... as a change in those does not require the model to be re-meshed. You can just update the solution, much quicker.

Another thing that can help with these types of models (in terms of solution speed) is the option to use linear discretization. You can choose this, rather than the default quadratic discretization, in the settings for the main mf node. Test with both options, if no significant difference in result, you can use linear for much faster solution times.

Hope this helps...

Mark

edit: use the second attachment (better), the first image did not capture the entire sequence.
Kyle, Your geometry sequence seems overcomplicated. I have run a 1/4 symmetry model where I have disabled a large number of your geometry nodes to simplify matters. Also, your mesh in the airgap was too coarse to reliably resolve accurate field values. For the 1/4 symmetry model I modelled the system using your default current and obtain a field at the centre of the airgap of about 0.76 mT. For the full geometry, I obtained the same. I think you were getting different results because you have not divided the current by 2 in the symmetric version, (only half of the turns are modelled). The symmetry condition closes the coil electrically, but in terms of the coils axial length, with the 1/4 symmetry, you only have 1/2 of the coil, so the total current should be stated as half of the "true" total current. I have attached a screenshot of the simplified geometry sequence. The highlighted region is used as a mesh control domain, (I limited the mesh size in this mesh control domain to 0.01[m]. Rather than having all of those features (the ones I have disabled), just create the raw geometry, then use a couple of "block/Boolean difference" features at the end to split the entire model into 4, much cleaner. To analyse results, I much prefer to use cut planes, cut points etc... as a change in those does not require the model to be re-meshed. You can just update the solution, much quicker. Another thing that can help with these types of models (in terms of solution speed) is the option to use linear discretization. You can choose this, rather than the default quadratic discretization, in the settings for the main mf node. Test with both options, if no significant difference in result, you can use linear for much faster solution times. Hope this helps... Mark edit: use the second attachment (better), the first image did not capture the entire sequence.


Please login with a confirmed email address before reporting spam

Posted: 8 years ago 6 févr. 2017, 18:36 UTC−5
Hi Mark,

Thanks for your continued, detailed help.

I revised my geometry to be simpler, and also utilized a mesh control domain (same discretization as you) in the gap.

I created 3 different models (all attached), one of the full magnet (emag_V4.3_full.mph), one with half symmetry (emag_V4.3_hsym.mph, no PMC boundary) and one with quarter symmetry (emag_V4.3_qsym.mph). Maybe I'm just not using a fine enough discretization, but the results do not convince me that the symmetry scenarios (especially the quarter symmetry) are sufficiently equivalent and trustworthy.

I've attached plots for 2 different cut lines in the gap (horizontal and vertical) for the 3 geometries, and while I feel the full and 1/2 symmetry are in reasonable agreement, the 1/4 is not. This is with respect to both amplitude of the field and the distribution/field pattern. If I was simply off by a factor of 2 in the current/# of turns, I would expect my magnitude to be off but the pattern to be relatively similar.

Now regarding the coil and the current. I don't understand why you said that the current should be halved. For the full geometry, I am using 700 turns and 50A for the most recent simulation. When I go down to 1/2 symmetry, I am cutting the coil perpendicular to the direction of the current. From my understanding, I would assume that I should leave the 700 turns and 50A for this 1/2 symmetry. Further, when I reduce to 1/4 symmetry, now the input surface of the coil is 1/2 its original size, and I assume that I should therefore reduce the number of turns to 350, but leave the coil current at 50A. Please explain if this assessment is incorrect.

Additionally, I have not explored linear vs. quadratic discretization. I will once I resolve this symmetry issue.

Thanks again,
Kyle


Kyle,

Your geometry sequence seems overcomplicated. I have run a 1/4 symmetry model where I have disabled a large number of your geometry nodes to simplify matters. Also, your mesh in the airgap was too coarse to reliably resolve accurate field values.

For the 1/4 symmetry model I modelled the system using your default current and obtain a field at the centre of the airgap of about 0.76 mT. For the full geometry, I obtained the same.

I think you were getting different results because you have not divided the current by 2 in the symmetric version, (only half of the turns are modelled). The symmetry condition closes the coil electrically, but in terms of the coils axial length, with the 1/4 symmetry, you only have 1/2 of the coil, so the total current should be stated as half of the "true" total current.

I have attached a screenshot of the simplified geometry sequence. The highlighted region is used as a mesh control domain, (I limited the mesh size in this mesh control domain to 0.01[m].

Rather than having all of those features (the ones I have disabled), just create the raw geometry, then use a couple of "block/Boolean difference" features at the end to split the entire model into 4, much cleaner.

To analyse results, I much prefer to use cut planes, cut points etc... as a change in those does not require the model to be re-meshed. You can just update the solution, much quicker.

Another thing that can help with these types of models (in terms of solution speed) is the option to use linear discretization. You can choose this, rather than the default quadratic discretization, in the settings for the main mf node. Test with both options, if no significant difference in result, you can use linear for much faster solution times.

Hope this helps...

Mark

edit: use the second attachment (better), the first image did not capture the entire sequence.


Hi Mark, Thanks for your continued, detailed help. I revised my geometry to be simpler, and also utilized a mesh control domain (same discretization as you) in the gap. I created 3 different models (all attached), one of the full magnet (emag_V4.3_full.mph), one with half symmetry (emag_V4.3_hsym.mph, no PMC boundary) and one with quarter symmetry (emag_V4.3_qsym.mph). Maybe I'm just not using a fine enough discretization, but the results do not convince me that the symmetry scenarios (especially the quarter symmetry) are sufficiently equivalent and trustworthy. I've attached plots for 2 different cut lines in the gap (horizontal and vertical) for the 3 geometries, and while I feel the full and 1/2 symmetry are in reasonable agreement, the 1/4 is not. This is with respect to both amplitude of the field and the distribution/field pattern. If I was simply off by a factor of 2 in the current/# of turns, I would expect my magnitude to be off but the pattern to be relatively similar. Now regarding the coil and the current. I don't understand why you said that the current should be halved. For the full geometry, I am using 700 turns and 50A for the most recent simulation. When I go down to 1/2 symmetry, I am cutting the coil perpendicular to the direction of the current. From my understanding, I would assume that I should leave the 700 turns and 50A for this 1/2 symmetry. Further, when I reduce to 1/4 symmetry, now the input surface of the coil is 1/2 its original size, and I assume that I should therefore reduce the number of turns to 350, but leave the coil current at 50A. Please explain if this assessment is incorrect. Additionally, I have not explored linear vs. quadratic discretization. I will once I resolve this symmetry issue. Thanks again, Kyle [QUOTE] Kyle, Your geometry sequence seems overcomplicated. I have run a 1/4 symmetry model where I have disabled a large number of your geometry nodes to simplify matters. Also, your mesh in the airgap was too coarse to reliably resolve accurate field values. For the 1/4 symmetry model I modelled the system using your default current and obtain a field at the centre of the airgap of about 0.76 mT. For the full geometry, I obtained the same. I think you were getting different results because you have not divided the current by 2 in the symmetric version, (only half of the turns are modelled). The symmetry condition closes the coil electrically, but in terms of the coils axial length, with the 1/4 symmetry, you only have 1/2 of the coil, so the total current should be stated as half of the "true" total current. I have attached a screenshot of the simplified geometry sequence. The highlighted region is used as a mesh control domain, (I limited the mesh size in this mesh control domain to 0.01[m]. Rather than having all of those features (the ones I have disabled), just create the raw geometry, then use a couple of "block/Boolean difference" features at the end to split the entire model into 4, much cleaner. To analyse results, I much prefer to use cut planes, cut points etc... as a change in those does not require the model to be re-meshed. You can just update the solution, much quicker. Another thing that can help with these types of models (in terms of solution speed) is the option to use linear discretization. You can choose this, rather than the default quadratic discretization, in the settings for the main mf node. Test with both options, if no significant difference in result, you can use linear for much faster solution times. Hope this helps... Mark edit: use the second attachment (better), the first image did not capture the entire sequence. [/QUOTE]


Please login with a confirmed email address before reporting spam

Posted: 8 years ago 7 févr. 2017, 06:01 UTC−5
Kyle,

I meant to say to half the turns, not the current.

For reference, see the attached Excel file (zipped). In this I have modelled a simple coil with 1200 turns, a current of 2.262 Amps with a wire diameter of 0.5[mm].

Such a coil would have a resistance of about 10.4 Ohms, and would produce a field of about 57 mT at it's centre.

There are ways we can use symmetry to simulate this coil, we could cut the coil in half in 2 ways (refer to attached images).

The excel table shows the values that should be used in your model. Also, the Excel files shows what you should use as coil inputs.

In your half model, you have not defined the coil input boundary correctly. When the coil is split for geometry reasons in this way, you need to define the coil input and output boundaries as the two "cut" boundaries of the coil in the model. Also, under the coil definition you should use N/2 as the number of turns.

When these corrections are made, the field at the centre of your airgap is around 409 mT, rather than 8.5 mT as in your original model.

Your quarter symmetry model has no coil inputs defined. Adding those and again using N/2 as the number of turns gives a field at the airgap centre of 409 mT also (see attachments for reference).

You full geometry model gives a field at the centre of the airgap of 420 mT (although I did use linear discretization for the full model to speed up the solution). I would say these results are all equivalent.

general comments:
Don't change values in your parameters list and also your coil input values, else you will end up in a muddle. I would recommend keeping your parameter list as it should be for the full model (i.e. N=700 and III=50[A]). Just change the term in the coil definition (for example N/2) and if you are calculating the resistance or voltage drops as derived values, make sure you use an appropriate multiplier in the derived values formula (as shown in the excel table).

make sure your coil input/output boundaries are correctly defined.

Your line plots are based on cut lines in close proximity to the pole face edges. You will need a much finer mesh in these regions to accurately solve for the fields in these locations, (notice that your line plots are not very smooth).

The Cut Point 3D I defined was at location (0, 0.125, 0), for reference.

Good luck!

Mark
Kyle, I meant to say to half the turns, not the current. For reference, see the attached Excel file (zipped). In this I have modelled a simple coil with 1200 turns, a current of 2.262 Amps with a wire diameter of 0.5[mm]. Such a coil would have a resistance of about 10.4 Ohms, and would produce a field of about 57 mT at it's centre. There are ways we can use symmetry to simulate this coil, we could cut the coil in half in 2 ways (refer to attached images). The excel table shows the values that should be used in your model. Also, the Excel files shows what you should use as coil inputs. In your half model, you have not defined the coil input boundary correctly. When the coil is split for geometry reasons in this way, you need to define the coil input and output boundaries as the two "cut" boundaries of the coil in the model. Also, under the coil definition you should use N/2 as the number of turns. When these corrections are made, the field at the centre of your airgap is around 409 mT, rather than 8.5 mT as in your original model. Your quarter symmetry model has no coil inputs defined. Adding those and again using N/2 as the number of turns gives a field at the airgap centre of 409 mT also (see attachments for reference). You full geometry model gives a field at the centre of the airgap of 420 mT (although I did use linear discretization for the full model to speed up the solution). I would say these results are all equivalent. general comments: Don't change values in your parameters list and also your coil input values, else you will end up in a muddle. I would recommend keeping your parameter list as it should be for the full model (i.e. N=700 and III=50[A]). Just change the term in the coil definition (for example N/2) and if you are calculating the resistance or voltage drops as derived values, make sure you use an appropriate multiplier in the derived values formula (as shown in the excel table). make sure your coil input/output boundaries are correctly defined. Your line plots are based on cut lines in close proximity to the pole face edges. You will need a much finer mesh in these regions to accurately solve for the fields in these locations, (notice that your line plots are not very smooth). The Cut Point 3D I defined was at location (0, 0.125, 0), for reference. Good luck! Mark


Please login with a confirmed email address before reporting spam

Posted: 8 years ago 7 févr. 2017, 12:36 UTC−5
Thank you! That was a dumb oversight on my part with the input/output surfaces. It all makes sense now and I am getting consistent results with 1/4 symmetry and full symmetry. All the coil turn/current calculations that you posted are what I assumed.

I really appreciate your time, I've learned quite a bit from your replies. I'm much more confident in my results now.

The last question I have is this: how can I achieve a view where I see the coil turns that COMSOL generates like in the attached image?

Thanks again,
Kyle


Thank you! That was a dumb oversight on my part with the input/output surfaces. It all makes sense now and I am getting consistent results with 1/4 symmetry and full symmetry. All the coil turn/current calculations that you posted are what I assumed. I really appreciate your time, I've learned quite a bit from your replies. I'm much more confident in my results now. The last question I have is this: how can I achieve a view where I see the coil turns that COMSOL generates like in the attached image? Thanks again, Kyle


Please login with a confirmed email address before reporting spam

Posted: 8 years ago 7 févr. 2017, 14:48 UTC−5
Nevermind about the coil. I figured it out. For anyone else looking for it, it's a Streamline plot using Magnetic Fields > Coil Parameters > Coil Direction
Nevermind about the coil. I figured it out. For anyone else looking for it, it's a Streamline plot using Magnetic Fields > Coil Parameters > Coil Direction

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.