Menu

Discussion Forum

New Discussion
Filter

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.

who can tell me why the magnetic force different?

Posted Mar 15, 2011, 5:17 p.m. EDT Low-Frequency Electromagnetics 19 Replies

Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

I did a simple simulation in 2D or 3D, which include a block of soft magnetic material and a block of permanent magnet. I want to verify the force on soft magnetic material equals the force on permanent magnet. However, the simulation results indicate that they are huge different. I defined the names of the magnet and soft magnetic material from subdomain setting in comsol v3.5a, and then check the results from postprocessing regarding global data. I wonder what is the reason causing this different. I also check the tutorial model in which I just define the name for different material, and the conclusion is the same.
19 Replies Last Post Nov 25, 2013, 11:07 p.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: 1 decade ago Mar 15, 2011, 6:57 p.m. EDT
Hi

if you are using the Maxwell stress tensors, there are many threads on the Forum about this.

This method is very sensitive to the meshing, I have also deposited a few examples in the model exchange (but these were for 3.5a)

This point is worth to study in detail, and to understand, as its rather essential for correct force derivation

--
Good luck
Ivar
Hi if you are using the Maxwell stress tensors, there are many threads on the Forum about this. This method is very sensitive to the meshing, I have also deposited a few examples in the model exchange (but these were for 3.5a) This point is worth to study in detail, and to understand, as its rather essential for correct force derivation -- Good luck Ivar
Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 15, 2011, 9:57 p.m. EDT
Hi, Ivar,

Thank you very much for your response and suggestions. You are the most helpful person in this community. I will verify them for your suggestions. The question is that I checked the tutorial model such as the example of magnetic force calculation on U-shape magnet and rod for 3D model and the example of 2D linear actuator.The first model only calculates the force on the rod in the tutorial, and the second model only calculates the force on the coil. What I did is to compare the force on U-shape magnet with the force on the rod in the first model, and compare the force on the core and magnet with the force on the coil. The results are not satisfying. I do not completely agree with you for your suggestion in which you mentioned the mesh is not fine enough. I thought the software may exists some problems since when I check the force on the surrounding air, it is also a remarkable number. It is impossible in the actual application.

Could you give me some simple model about the interacting force calculation which you verified?

Appreciate a lot.

Peter
Hi, Ivar, Thank you very much for your response and suggestions. You are the most helpful person in this community. I will verify them for your suggestions. The question is that I checked the tutorial model such as the example of magnetic force calculation on U-shape magnet and rod for 3D model and the example of 2D linear actuator.The first model only calculates the force on the rod in the tutorial, and the second model only calculates the force on the coil. What I did is to compare the force on U-shape magnet with the force on the rod in the first model, and compare the force on the core and magnet with the force on the coil. The results are not satisfying. I do not completely agree with you for your suggestion in which you mentioned the mesh is not fine enough. I thought the software may exists some problems since when I check the force on the surrounding air, it is also a remarkable number. It is impossible in the actual application. Could you give me some simple model about the interacting force calculation which you verified? Appreciate a lot. Peter
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: 1 decade ago Mar 16, 2011, 2:13 a.m. EDT
Hi

then we agree, because it is NOT ONLY a question of fine enough mesh, but more about singularities and "sharp edges" in the model. Maxwell Tensor derivation is based on gradient evaluation, and strong gradients, when you take differences of large , but similar numbers, the result tend to have a poor resolution. That is my understanding.

take a square/rectangular magnet, in 2D view. How do you define the gradient on the point (intersection of two edges ?), depending on the numerics it will tend a but more to the left or the right, that is, in my view the essential of the culprit. The rest in binary numbers and time to wait. In fact even the finer you go with the mesh, you will get a higher peak at the singularity !

Try it out ;) With COMSOL it's "easy" and very instructive

--
Good luck
Ivar
Hi then we agree, because it is NOT ONLY a question of fine enough mesh, but more about singularities and "sharp edges" in the model. Maxwell Tensor derivation is based on gradient evaluation, and strong gradients, when you take differences of large , but similar numbers, the result tend to have a poor resolution. That is my understanding. take a square/rectangular magnet, in 2D view. How do you define the gradient on the point (intersection of two edges ?), depending on the numerics it will tend a but more to the left or the right, that is, in my view the essential of the culprit. The rest in binary numbers and time to wait. In fact even the finer you go with the mesh, you will get a higher peak at the singularity ! Try it out ;) With COMSOL it's "easy" and very instructive -- Good luck Ivar
Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 16, 2011, 1:23 p.m. EDT
Hi, Ivar,

I feel this topic becoming more interesting. To avoid the singularities between edges, I try a new 2D model in which there are only two circular objects. One is setting the physics as permanent magnet, and another is setting the physics as soft iron. These two objects is surrounding with a circular air area in which I also defined an infinite element shell. The result about the streamline and surface plot of magnetic flux density looks reasonable. But the magnetic force from the magnet and from the soft iron which I designated in Physics-Subdomain Setting is still indicating huge different.

Could you try the simple model in your convenient time and tell me your feeling?

Thanks again.
Peter
Hi, Ivar, I feel this topic becoming more interesting. To avoid the singularities between edges, I try a new 2D model in which there are only two circular objects. One is setting the physics as permanent magnet, and another is setting the physics as soft iron. These two objects is surrounding with a circular air area in which I also defined an infinite element shell. The result about the streamline and surface plot of magnetic flux density looks reasonable. But the magnetic force from the magnet and from the soft iron which I designated in Physics-Subdomain Setting is still indicating huge different. Could you try the simple model in your convenient time and tell me your feeling? Thanks again. Peter
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: 1 decade ago Mar 16, 2011, 5:05 p.m. EDT
Hi

here are 10 quick examples in normal and extremely fine mesh for square and round edges, tri and mapped mesh, and square/round surroundings and with or without infinite elements, take a look

--
Good luck
Ivar
Hi here are 10 quick examples in normal and extremely fine mesh for square and round edges, tri and mapped mesh, and square/round surroundings and with or without infinite elements, take a look -- Good luck Ivar

Attachments:

Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 16, 2011, 10:50 p.m. EDT
Hi, Ivar,

Thanks a lot. I will check them tomorrow, and tell you my feeling.

Peter
Hi, Ivar, Thanks a lot. I will check them tomorrow, and tell you my feeling. Peter
Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 17, 2011, 1:17 p.m. EDT
Hi, Ivar,

I am sorry that I could not open the files you attached. My version is 3.5a and if there are some other models available, it will be very helpful to me.

Many thanks.
Peter
Hi, Ivar, I am sorry that I could not open the files you attached. My version is 3.5a and if there are some other models available, it will be very helpful to me. Many thanks. Peter
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: 1 decade ago Mar 17, 2011, 4:52 p.m. EDT
Hi

here are the PNG views of the "normal mesh" cases, the 5 others are the same, but meshed extremely fine.

I then compare the integrated force on the magnet (upper square) versus the iron (lower square).
The outer ring/square (if present) are the infinite elements (7 & 9)

Its easily run in 3.5.

Surrounding "air mur=1", iron mur=1000, magnet M 1E6Am

--
Good luck
Ivar

--
Good luck
Ivar
Hi here are the PNG views of the "normal mesh" cases, the 5 others are the same, but meshed extremely fine. I then compare the integrated force on the magnet (upper square) versus the iron (lower square). The outer ring/square (if present) are the infinite elements (7 & 9) Its easily run in 3.5. Surrounding "air mur=1", iron mur=1000, magnet M 1E6Am -- Good luck Ivar -- Good luck Ivar
Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 17, 2011, 5:17 p.m. EDT
Hi, Ivar,

Thanks for your intensive care about my question and providing a lot of models.

The problem is that I ran your models and got the similar images. However, when I check the force on the magnet and the force on the iron, it shows significant difference. I want to make sure or get some inspiration from you about what is the exact reason causing this difference. According to Newton's third law, these two forces are supposed to be equal or minor difference.

Thanks a lot.
Peter
Hi, Ivar, Thanks for your intensive care about my question and providing a lot of models. The problem is that I ran your models and got the similar images. However, when I check the force on the magnet and the force on the iron, it shows significant difference. I want to make sure or get some inspiration from you about what is the exact reason causing this difference. According to Newton's third law, these two forces are supposed to be equal or minor difference. Thanks a lot. Peter
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: 1 decade ago Mar 17, 2011, 5:32 p.m. EDT
Hi

then try to look at the maxwell stress tensor component side by side (for each of the 4 sides for the "easy cases") of the iron and of the magnet, then add them up to the total force (manually).

If you notice: a coarse square mesh gives almost good results (because the model is square), while the others not so good, but anyhow what is <10% off ? Rounded off corners improves things, so does infinite elements (or larger air region ...)

This is the modeling world, nothing is perfect, most probably your material data is not going to be better than 10% anyhow: homogeneity, shape, precision, Temperature, etc ... all adds up

Your magnets will never have truly such strong singularities in the "corners", at a certain moment the material magnetic properties becomes also non linear ...

--
Good luck
Ivar
Hi then try to look at the maxwell stress tensor component side by side (for each of the 4 sides for the "easy cases") of the iron and of the magnet, then add them up to the total force (manually). If you notice: a coarse square mesh gives almost good results (because the model is square), while the others not so good, but anyhow what is
Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 17, 2011, 11:09 p.m. EDT
Hi, Ivar,

I verified a very simple model which is similar to the tutorial model for two parallel wires, but I changed the current carrying wires with circular magnet and iron. This model is supposed to have no singlarity, and the material which I used for iron is from material library. The result shows the force on the magnet is about -90N, and the force on the iron is about 260N. This is what I mentioned huge difference. I want to figure out why it happened like this.

If you have time, please run that model and compare the force on left circle (assigned it as iron) and right circle (assigned as magnet). The result will shows the huge difference. It is not 10% or 20% difference.

Thanks a lot.
Peter
Hi, Ivar, I verified a very simple model which is similar to the tutorial model for two parallel wires, but I changed the current carrying wires with circular magnet and iron. This model is supposed to have no singlarity, and the material which I used for iron is from material library. The result shows the force on the magnet is about -90N, and the force on the iron is about 260N. This is what I mentioned huge difference. I want to figure out why it happened like this. If you have time, please run that model and compare the force on left circle (assigned it as iron) and right circle (assigned as magnet). The result will shows the huge difference. It is not 10% or 20% difference. Thanks a lot. Peter
Peter Wen
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Mar 18, 2011, 12:17 a.m. EDT
Hi, Ivar,

You are absolutely correct. The difference of action and reaction forces become smaller and smaller with the finer and finer of the meshing. I worked on another computer with large memory and 64 bit system. The value of the two forces become closer and closer.

All in all, appreciate a lot for your help.
Peter
Hi, Ivar, You are absolutely correct. The difference of action and reaction forces become smaller and smaller with the finer and finer of the meshing. I worked on another computer with large memory and 64 bit system. The value of the two forces become closer and closer. All in all, appreciate a lot for your help. Peter
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: 1 decade ago Mar 18, 2011, 2:24 a.m. EDT
Hi

Indeed discretisation of a model is not a trial thing, in some sens its more an art than pure theoretical science, at least to get it quickly work in such a huge field of application we have with COMSOL

--
Good luck
Ivar
Hi Indeed discretisation of a model is not a trial thing, in some sens its more an art than pure theoretical science, at least to get it quickly work in such a huge field of application we have with COMSOL -- Good luck Ivar
Mahdi Mohammadi
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Sep 13, 2013, 12:31 p.m. EDT
Hi Folks,

My problem is that my simulation results do not agree with the experiment results.
I have two cylindrical Neodymium magnets of sizes shown in the attached file, and an Iron part between them. I measured the magnetic field of the magnets using a Gauss meter and they perfectly match with the simulation, in various distances.
I need the magnetic force in two situations:
1- The magnetic attraction between the small magnet and washer: in this case the magnetization for the big magnet is zero. The force I get from simulation is 0.3897 N but from the experiment that force is more than 0.8 N.
2- Magnetic attraction between small magnet and big magnet when the gap between small magnet and washer is 10 mm (instead of 1 mm): in this case the magnetization for the big magnet is 1000 [kA/m]. The force I get from simulation is 0.2272 N but from the experiment that force is more than 0.6 N.

I also did a 2D axisymmetric simulation which gives the same results.

Thank you in advance for the help.
Mahdi
Hi Folks, My problem is that my simulation results do not agree with the experiment results. I have two cylindrical Neodymium magnets of sizes shown in the attached file, and an Iron part between them. I measured the magnetic field of the magnets using a Gauss meter and they perfectly match with the simulation, in various distances. I need the magnetic force in two situations: 1- The magnetic attraction between the small magnet and washer: in this case the magnetization for the big magnet is zero. The force I get from simulation is 0.3897 N but from the experiment that force is more than 0.8 N. 2- Magnetic attraction between small magnet and big magnet when the gap between small magnet and washer is 10 mm (instead of 1 mm): in this case the magnetization for the big magnet is 1000 [kA/m]. The force I get from simulation is 0.2272 N but from the experiment that force is more than 0.6 N. I also did a 2D axisymmetric simulation which gives the same results. Thank you in advance for the help. Mahdi

Attachments:

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

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Sep 14, 2013, 9:42 a.m. EDT
Mahdi,

the thread above gives really valuable hints for this kind of issue. Did you consider all aspects that were already mentioned, such as meshing and geometry aspects?
Your model geometry doesn't look like you did, e.g. you have sharp corners and edges in the geometry.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
Mahdi, the thread above gives really valuable hints for this kind of issue. Did you consider all aspects that were already mentioned, such as meshing and geometry aspects? Your model geometry doesn't look like you did, e.g. you have sharp corners and edges in the geometry. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology
Mahdi Mohammadi
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Sep 15, 2013, 3:47 p.m. EDT
Edgar,

Thank you for your response. I tried different mesh sizes and it changes the results very slightly. I also tried fillets of various sizes on my 2D model, the results gets even worse, perhaps because the total volume decreases. I upload my 2D model here, I appreciate your help. Mahdi
Edgar, Thank you for your response. I tried different mesh sizes and it changes the results very slightly. I also tried fillets of various sizes on my 2D model, the results gets even worse, perhaps because the total volume decreases. I upload my 2D model here, I appreciate your help. Mahdi
Edgar J. Kaiser Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Sep 15, 2013, 6:15 p.m. EDT

Mahdi,

I will have a look, but won't be able to get back to it before the end of the week due to travel.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
Mahdi, I will have a look, but won't be able to get back to it before the end of the week due to travel. Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology
Mahdi Mohammadi
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Sep 27, 2013, 3:01 p.m. EDT
Hi Edgar,

I am still waiting for response. Thank you, Mahdi
Hi Edgar, I am still waiting for response. Thank you, Mahdi
Janet Kissho
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago Nov 25, 2013, 11:07 p.m. EST

Hi

if you are using the Maxwell stress tensors, there are many threads on the Forum about this.

This method is very sensitive to the meshing, I have also deposited a few examples in the model exchange (but these were for 3.5a)

This point is worth to study in detail, and to understand, as its rather essential for correct force derivation

--
Good luck
Ivar

well answered
[QUOTE] Hi if you are using the Maxwell stress tensors, there are many threads on the Forum about this. This method is very sensitive to the meshing, I have also deposited a few examples in the model exchange (but these were for 3.5a) This point is worth to study in detail, and to understand, as its rather essential for correct force derivation -- Good luck Ivar [/QUOTE] well answered

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