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.

for NdFeB magnet, which constitutive relation is closer to reality

Please login with a confirmed email address before reporting spam

I'm modeling two NdFeB magnets placed very close to each other in a repulsive manner. I need to compute the repelling force between them.

There are two options for the constitutive relation. One is

B = mu0*(H+M)

where I specify the magnetization M vector based on the demagnetization curve of the magnet.

The other is

B = mu0*mur*H+Br

where Br is the remanent flux density of the magnet and mur can be calculated from the demagnetization curve.

The problem is that two methods seem to give different force results, with that of the first one bigger than the second. Which constitutive relation is closer to reality?


5 Replies Last Post 2013年1月21日 GMT-5 03:11
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 2010年8月1日 GMT-4 04:56
Hi

in V4 or 3.5 ? use the new tag options on the Forum, then it's easier for all others to understand you

Have you also checked the fluxes for your two cases, do they also differ ?

--
Good luck
Ivar
Hi in V4 or 3.5 ? use the new tag options on the Forum, then it's easier for all others to understand you Have you also checked the fluxes for your two cases, do they also differ ? -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2010年8月1日 GMT-4 16:48

Hi

in V4 or 3.5 ? use the new tag options on the Forum, then it's easier for all others to understand you

Have you also checked the fluxes for your two cases, do they also differ ?

--
Good luck
Ivar


Hi Ivar,

I'm using 3.3 (3.3.0.405) (yes it's quite old now) AC/DC Module, Azimuthal Induction Currents, Vector Potential (emqa) with 2D axi-symmetric mode.

I checked the fluxes for two cases. They are different mainly at the corners of the magnets. Also, I plotted the streamline of magnetization vector inside the magnets in two cases. They look different, too. After experimenting with the constitutive relation a little more, I think I found why these two cases are different. Basically it is not due to the software. Below is a brief explanation.

As a reminder, I'm try to compute the repulsive force between two magnet rings magnetized in axial direction and placed vertically one on top the other along z-axis.

Correct me if I am wrong, since in this application mode the vector potential is used, all the constitutive relation does is to let COMSOL know how to get H (vector) after knowing B (vector).

In Case 1, I chose the constitutive relation

B = mu0*H+mu0*M (here B, H, M are vectors). (1)

Since the magnets are assumed to be only magnetized in z-direction (they are also stacked in z-direction in a repulsive manner), I specified the magnetization vector as

M = [0, Mz], (2)

where Mr = 0 along r-direction (meaning no magnetization in r-direction) and Mz is a function of Bz which reads

Mz = Bz/mu0-Hz(Bz), (3)

where Hz(Bz) is the demagnetization curve of the magnet in 2nd quadrant. Note that Eq (2) and (3) are just valid for the bottom magnet where the positive directions agree with the positive z-direction. For the top magnet, different signs are used for Eq. (2) and (3).

In Case 2, I chose the constitutive relation

B = mu0*mur*H+Brem (here B, H, Brem are vectors ) (4)

I'm using Brem to denote remanence, just to differentiate it from the r-component of B which is Br. I specified the remanence flux density as

Brem = [0, Bremz] (5)

meaning Bremr = 0 along r-direction (no remanence flux in r-direction) and Bremz is constant from the magnet characteristics. Now mur in (4) becomes a function of Bz which reads

mur = (Bz-Bremz)/(mu0*Hz(Bz)) (6)

Again, Hz(Bz) is the demagnetization curve of the magnet in 2nd quadrant. Same as in the first case, sign considerations are shown for the bottom magnet.

The source of the problem is that I was using the ISOTROPIC option for mur in the second case (Eq. (4)). As a result, whatever the mur along z-direction ends up being will be copied to the r-direction to give

Br = mu0*mur*Hr. (7)

However, in Case 1, along r-direction I actually had

Br = mu0*Hr. (8)

Because mur calculated from the magnet characteristic is slightly greater than 1 (e.g. 1.05), Eq (7) & (8) are different in general. This small difference is the source of the difference in force results because Case 1 and Case 2 give exactly the same result if I use ANISOTROPIC option in Case 2 and specify

mur_anisotropic = [1, 0;
0, mur_zz].

with mur_zz = mur given by Eq (6).

In other words, along z-direction (direction of magnetization), these two constitutive relations are equivalent as long as things are defined correctly. However, along r-direction (perpendicular to direction of magnetization), some care needs to be taken to account for the r-axis relative permeability, which may or may not be unity 1.0 depending on the actual characteristics of the magnet material.

What still surprised me is that the force computation results are quite sensitive to the r-axis relative permeability. For mur_rr = 1.0, I'm getting 431 lbs, and for mur_rr = 1.07, I'm getting 412 lbs. Any suggestions?

Thanks!
[QUOTE] Hi in V4 or 3.5 ? use the new tag options on the Forum, then it's easier for all others to understand you Have you also checked the fluxes for your two cases, do they also differ ? -- Good luck Ivar [/QUOTE] Hi Ivar, I'm using 3.3 (3.3.0.405) (yes it's quite old now) AC/DC Module, Azimuthal Induction Currents, Vector Potential (emqa) with 2D axi-symmetric mode. I checked the fluxes for two cases. They are different mainly at the corners of the magnets. Also, I plotted the streamline of magnetization vector inside the magnets in two cases. They look different, too. After experimenting with the constitutive relation a little more, I think I found why these two cases are different. Basically it is not due to the software. Below is a brief explanation. As a reminder, I'm try to compute the repulsive force between two magnet rings magnetized in axial direction and placed vertically one on top the other along z-axis. Correct me if I am wrong, since in this application mode the vector potential is used, all the constitutive relation does is to let COMSOL know how to get H (vector) after knowing B (vector). In Case 1, I chose the constitutive relation B = mu0*H+mu0*M (here B, H, M are vectors). (1) Since the magnets are assumed to be only magnetized in z-direction (they are also stacked in z-direction in a repulsive manner), I specified the magnetization vector as M = [0, Mz], (2) where Mr = 0 along r-direction (meaning no magnetization in r-direction) and Mz is a function of Bz which reads Mz = Bz/mu0-Hz(Bz), (3) where Hz(Bz) is the demagnetization curve of the magnet in 2nd quadrant. Note that Eq (2) and (3) are just valid for the bottom magnet where the positive directions agree with the positive z-direction. For the top magnet, different signs are used for Eq. (2) and (3). In Case 2, I chose the constitutive relation B = mu0*mur*H+Brem (here B, H, Brem are vectors ) (4) I'm using Brem to denote remanence, just to differentiate it from the r-component of B which is Br. I specified the remanence flux density as Brem = [0, Bremz] (5) meaning Bremr = 0 along r-direction (no remanence flux in r-direction) and Bremz is constant from the magnet characteristics. Now mur in (4) becomes a function of Bz which reads mur = (Bz-Bremz)/(mu0*Hz(Bz)) (6) Again, Hz(Bz) is the demagnetization curve of the magnet in 2nd quadrant. Same as in the first case, sign considerations are shown for the bottom magnet. The source of the problem is that I was using the ISOTROPIC option for mur in the second case (Eq. (4)). As a result, whatever the mur along z-direction ends up being will be copied to the r-direction to give Br = mu0*mur*Hr. (7) However, in Case 1, along r-direction I actually had Br = mu0*Hr. (8) Because mur calculated from the magnet characteristic is slightly greater than 1 (e.g. 1.05), Eq (7) & (8) are different in general. This small difference is the source of the difference in force results because Case 1 and Case 2 give exactly the same result if I use ANISOTROPIC option in Case 2 and specify mur_anisotropic = [1, 0; 0, mur_zz]. with mur_zz = mur given by Eq (6). In other words, along z-direction (direction of magnetization), these two constitutive relations are equivalent as long as things are defined correctly. However, along r-direction (perpendicular to direction of magnetization), some care needs to be taken to account for the r-axis relative permeability, which may or may not be unity 1.0 depending on the actual characteristics of the magnet material. What still surprised me is that the force computation results are quite sensitive to the r-axis relative permeability. For mur_rr = 1.0, I'm getting 431 lbs, and for mur_rr = 1.07, I'm getting 412 lbs. Any suggestions? Thanks!

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 2010年8月2日 GMT-4 04:39
Hi

Thanks for the detailed explanation, one should have once a good example of a magnet and the different cases for simulating it in the doc, it would be easier for everyone to understand the differences; and perhaps one on the model exchange, it's finally basic magnetism good exercice for a student (but I do not have any ;)

One thing though the Maxwell tensor approach is very sensitive to the gradients and even small differences, especially with a rather coarse or not fully symmetric mesh, gives rapidly large differences in force. As these forces are calculated as sums along the edges, and these are often rather symmetric (which means differences) you get substractions of large numbers almost equal, with the classical numerical error building up. You see this is fou integrate along the edges, one by one.
Therefore I prefer the virtual work approach, even if longer to set up.

--
Good luck
Ivar
Hi Thanks for the detailed explanation, one should have once a good example of a magnet and the different cases for simulating it in the doc, it would be easier for everyone to understand the differences; and perhaps one on the model exchange, it's finally basic magnetism good exercice for a student (but I do not have any ;) One thing though the Maxwell tensor approach is very sensitive to the gradients and even small differences, especially with a rather coarse or not fully symmetric mesh, gives rapidly large differences in force. As these forces are calculated as sums along the edges, and these are often rather symmetric (which means differences) you get substractions of large numbers almost equal, with the classical numerical error building up. You see this is fou integrate along the edges, one by one. Therefore I prefer the virtual work approach, even if longer to set up. -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2013年1月21日 GMT-5 02:24
I want to model NdFeB magnets in comsol? Can you help me? How to find the Ndfeb material?
I want to model NdFeB magnets in comsol? Can you help me? How to find the Ndfeb material?

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 2013年1月21日 GMT-5 03:11
Hi

check your suppliers material property values, get the Br and Hc values.
Then I mostly use a Material Magnetisation value (Ampere law node magnetic field tab) of Hc, and sometimes I add the mu_r to Br/Hc

This has so far given me useful results (but never trust a model, nor material properties to something better then some 10% ;)

--
Good luck
Ivar
Hi check your suppliers material property values, get the Br and Hc values. Then I mostly use a Material Magnetisation value (Ampere law node magnetic field tab) of Hc, and sometimes I add the mu_r to Br/Hc This has so far given me useful results (but never trust a model, nor material properties to something better then some 10% ;) -- Good luck Ivar

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.