Discussion Forum

Hi

check the examples in the model library, you can either enter the field via the Br or the M magnetization

--
Good luck
Ivar

I have. The problem I'm having is that every time I create the magnet cylinder, the program interprets I'm creating a coil and shows me the magnetic field to the left and right of the cylinder when I should see something to the top and bottom, if you get my meaning.

Hi no not really ;)

first I'm using v4.1 there might be some differences (it's far more stable than the early v4.0 pls consider to update)

I start COMSOL:
3D, launch MF, + Select Stationary solver + Finish-flag
Geometry right click cylinder "radius = 4[mm], Length = 20[mm]) Position x=0, y=0, z= -10[mm]+ Update" (I get a cylinder , centered, with an axis along Z by default)
Geometry - right click sphere "radius = 100[mm]" ( you add a sphere, of "air around centered at 0,0,0)
Materials right click Add + sigma 1[S/m], epsilonr=1 mur=1 (this is the air and its define everywhere, you might rename it or add "air to the name)
Materials right click Add + select the central cylinder Domain "2" (use the Selection list, and but the view in "Transparency mode". I use almost the same properties sigma=1, epsilonr=1, mur=2 (small difference to clearly identify them, this is (will be) the magnet)

Note: in MF sigma might be =0, but in MEF you should always have at least a low conductivity to help the solver (1 might be too low, depending on the other materials, air conductivity up to 1000 might be OK and improve convergence, check the results for coherence)

Go to Physics Magnetic Fields MF, right click + Add "Amperes Law 2", select the magnet (domain "2") and change the "Relative Permeability" to Magnetization and change the Mz from default "zero" to Mz = 6E5

Note: you might define a "Global Definition Parameter M0 = 6E5[A/m] and refer to M0, at will. This value used here is arbitrary and not really representative of any particular magnet, check your magnet doc for more appropriate values

Note: you need this second "Ampere law 2" for the magnet as we are changing part of its material properties but not for the air. Notice that air was defined for all domains, then by adding this 2nd "Ampere Law 2 node" the domain "2" appear with an "overridden" in the main "Ampere law 1" node, only one Ampere law can be active for each domain, by default it's the last one in the model physics tree.

Leave the mesh to default and hit Solve right click Compute it should run with less than 1Gb ram

To see the results better right click Results 3D plot Group 1 + add Streamline Set Streamline Positioning to uniform density and 0.1 as separation distance

Then Results 3D lot Group 2 - Slice 1 set Plane Data Planes = 1

To better see the flux lines in a "cut plane do the following:

Results Data Sets + right Click Add "3D cut plane"
Right click Results + add 2D plot Group
Right click Results - 2D Plot group 1 + add Surface plot + Update plot

Note: because of the rather coarse mesh chosen the filed lines might be somewhat staggered.
You might want to zoom in to better see the results

One thing you see clearly the sharp edges of your cylinder are creating singularities (observe how hey worsen with finer mesh) ideally in ACDC your should use rounded corners wherever possible to avoid such singularities, in particular when using the maxwell stress tensor for force calculations

To go further you might replace the sphere by a square cube and see how the field lines are influenced (or not depending on size) from the corners of the air box, just as you might make a smaller sphere and add a thin r)1/10) of air) "infinite elements "spherical" node

--
Good luck
Ivar

Hi

one more thing:

In the example above, you might (to make the view nicer) Select Model 1 Definitions View 1 + right click+ add Hide geometric entities + select the BOUNDARY and the hit the external boundary (1/8 of a sphere) that is between your eye and the magnet + then do an "update" (the "eye" icon on the top border of the GUI)

Then you might turn off the transparency and look inside the air sphere.

If you are interested only what is happening inside the magnet you might remove the full entity/domain" 1 of the air


Note: in 3D geometry COMSOL does not have any 3D fillets (yet, hopefully to come). For that you need to create your cylinder as a revolved 2D workplane with a geometry made from an axial rectangle and you fillet with a given radius the two external corner (not on the axis of symmetry) I let that to you to exercise. The effect on the edge magnetic field is important

I have added my simple model hereby with all overlapping geometries, you should deselected some, with the chosen finer mesh you need some 2Gb of ram to run it

a 4.1.0.154 model is attached


--
Good luck
Ivar

Hi, Ivar. I followed your instructions from your second post but I still think there's something wrong somewhere, or maybe my electromagnetism is a little rusty. I'm attaching the file of my magnet. You'll see the field is stronger at the edges of the magnet when I think it should be stronger everywhere inside the magnet, with the rainbow gradually turning to blue on the outside.

Hi

Indeed, one could believe so.

BUT I suspect that it's your mesh density that is also playing you a game there.

If you select the magnet itself and add a "tet" extremely fine mesh and then solve, you will see that the red region concentrates on the upper and lower edge (as these are singularities) and the external boundary.

Its more apparent if you turn off the smoothing and use a raw No Refinement Resolution for the graph. One should always check the shapes in raw mode, smoothing is nice but you can quickly be fooled by artefact's

If you round the edges off with a fillet radius this effect will reduce somewhat but the .

--
Good luck
Ivar

This is what I'm looking at. I followed your instructions and created an "Extremely fine" tetrahedral mesh for the magnet and left Mesh1-->Size at "Fine."

I'm facing a different challenge this time. I've added both the coil and the magnet and a corresponding Ampère's Law 2 and External Current Density 1. If I compute this solution, I only get to see the magnetic field of the permanent magnet while the coil retains the colour of the normal plane. If, however, I disable Ampère's Law 2, then I see the magnetic field of the coil and, obviously, the permanent magnet is now the blue, inactive, element. I'm guessing there's some kind of conflict between the two, maybe one overrides the other? Is there a way to see the magnetic fields of both elements in the same slice? Thanks.

Hi

indeed in v4 some of the nodes are additive, some are exclusive, you distinguis them by the green and red trangle/arrows on the icones, check the doc and the help files (v4.1 doc is far better than the early 4.0).

Normally you should be able to see both coil and magnet (if the model is set up correctly) but be aware that you need a high current to generate the field corresponding to a modern high power magnet, and its easy to input an absolute current, while COMSOL is mostly set up, by default, for current densities, always chekc the units ;)

--
Good luck
Ivar

Does that mean that, since the External Current Density node is not additive, I cannot see the magnetic fields generated by both objects? Only one at a time? I'm giving the coil a current density corresponding to a total current of 3 amps. Furthermore, I need to then calculate the force the coil exerts on the magnet and, hopefully, run a simulation of the movement of the magnet given that force. Is any of this possible?

Hi

no normally, when the BC are correctly defined you solve for permanent magnets and currents "all in one" :)

You must have missed something in your model build up

--
Good luck
Ivar

Here's the model. I don't think it's missing anything but I must be wrong or otherwise it would have shown the magnetic fields of both elements.

Hi

there are a few things with your model:

1) your coil is of "air" not Copper. Now you have set the conductivity of the air to 1[S/m] (good idea) so the current will pass, but you will need probably a MV to get the current trough, (but that does not scare COMSOL, at most you get some numerical imprecisions)
2) if you select the work plane boundary section of your coil, and integrate the current density mf.normJ over your coil section you get about 3[A] so COMSOL is considering the current,
But further, if you integrate the value 1 to get the section I see its 0.0011 m^2 so your average current density is hardly more than 2.8[mA/mm^2], while I would expect some 1'000 to 2'000 times MORE (3-6[A/mm^2])
3) you have added a selection list on your results Data Set. This limits your data to the external surfaces, so you will not visualise the fields and currents correctly in your cut plane.

Then to visualise the current, use the mf.Jex,mf.Jey arrow plot this will show yo clearly that the current is there

You can compare the almost 1MA/m magnetisation with the 3kA/m^2 coil current for a coil that is somewhat larger than the magnet

So once I correct the elements above, I reset your model in v4.1 and resolv it, I do see some current and induced B field, albeit weak. To really see it, try a "log(mf.normB)" in the X-Y cut plane view

Finally I can only suggest that you update to the latest v4 version as the early v4.0 has many issues like this that are greatly improved in the new release

--
Good luck
Ivar

I think my problem is with my trying to oversimplify my model. See, the coil in my model is supposed to have 800 turns divided into 20 layers. The copper wire has a diameter of 0.8 mm and can withstand up to 3 amps. I think I made a mistake when I decided to create a single block and force 3 amps to run through it, since, electromagnetically, it can't be the same to have a single copper wire some 10.60 cm^2 in area and run 3 amps through it, right? Making some calculations with the real coil, I estimate I have something in the order of 5.968 A/mm^2 (since I'm running 3 amps through a wire that's .8 mm in diameter). Naturally, if I want the same current over a larger area, the number of A/mm^2 falls dramatically (2.83*10^-3 A/mm^2). I could convert my real coil into a new one with only 1 layer but that would pretty much destroy my hopes of simulating my prototype. Perhaps, it's as simple as inputting the 5.968 A/mm^2 over the larger 10.60 cm^2 area? This would result in a larger current running through my model coil but it would probably describe the magnetic field being generated by my real coil more accurately, right? Of course, the alternative would be to model an 800-turn-20-layer coil but that's pretty much out of my reach. Any thoughts?

EDIT: I tried imposing the 5.968 A/mm^2 and can now see a very faint magnetic field being generated by the coil. I'm guessing this is to be expected compared to the powerful magnetic field being generated by the neodymium magnet that dominates the color table. By adjusting with log() I can see that both magnetic fields are indeed present. Is there a way to move the magnet at will and see how the field lines change? Thanks.

Hi

then why do you have many turns ? If you have 5A in 1 wire you have 5*800 in 800 turns, but still, as the area increases the current density remains the same (you need to take into account the fill factor typically 60%)

so, indeed a 3-5A/mm^2 over the full coil section is more or less reasonable (at room temperature, non supraconducting), simulated as a bulk block, this remains valid at least in DC

in V4 you have multiturn coil BC's to better simulate a N turn coil without going into the details of each wire.

--
Good luck
Ivar

I'm afraid regarding the singularity problem, I have the similar question. Is it simply because of the corners that caused such nonuniform distribution of the field strength in the magnet? I attached a similar picture of the cylindrical magnet resulted from my simulation. Thank you very much!

Hi, Ivar,
I tried your method based on your model but still get more red on the edge. I don't understand and am totally confused. I'd really appreciate if you could give me some help on this. Thank you so much!

Best,
Chang

However, when the geometry's aspect radio gets close to unity or greater, it becomes more uniform. Now I'm confused if it's physics or it's artifacts due to the simulation boundary setting, etc.

Hi

I have been out of houise for some time ;) for me the field concentration ou nthe outer part of the coil is normal, in the sens that depending on the physics and BC you use you will have (or not at all) the effect on the current density by the field itself, that will push the current to the edges (skin effect). this is also very frequency dependent. Mostly with simple analytical means one forget these "2nd order" effect, even if they are often far from negligible.

But they can only easily be shown with tolls like COMSOL (and there are not many).

So sharp edges = singularities = partially wrong results for numerical reasons,

but current densities tending to increase at the outer border = physics applied correctly, and is a "feature" of a bulk coil

That is also why one tend to use multiple turn small wire diameter, or even coils made of thin films and not wires. Just as for higher frequencies, the current rtend to circulate at the border of the conductor and NOT in the middle, as the electroncs are repelled from centre by the generated magnetic field

--
Good luck
Ivar

Hi, Ivar,
You said "but current densities tending to increase at the outer border = physics applied correctly, and is a "feature" of a bulk coil ".
Does this apply to permanent as well? I was simulating a piece of permanent magnet without any coils. Could you please provide me more information on this? Or could you please guide me to some reading materials for better understanding of why concentrated B at border happens? I really appreciate your help!!! Thank you so much for your quick response!

Best,
Chang

By the way, I posted a question asking about a 2D solenoid problem. If I could get some help on that too, I'd really appreciate all help. Thank you!

www.comsol.com/community/forums/general/thread/22649/#p60523

--
Charles

Hi

I would suggest that you do the tutorial and verification models of the ACDC module and see how these behave, including reading carefully the pdf files of the description. In ACDC one takes easily shortcuts, but mostly one burn our fingers on those "short cuts" ;)

--
Good luck
Ivar

Could you please tell me which verification models do you refer to? I followed this model www.comsol.com/showroom/gallery/78/

Thank you!

--
Charles

Hi

that is a good one for mangets, you have other for the current in wires, do a "file open model library" in COMSOL, but be sure you have updated your model library to have all latest models, COMSOL adds regularly models each week ;)

And a suggestion use the latest version, v4.0 was an early bird, far from fully ironed out ;)
among others, if I remember right, their was a factor 2 missing in the formula for B from I of a coil in MEF

--
Good luck
Ivar

hi，
i want to simulate rectangular permanent magnet could you give me some hint on this modeling.

Hi

what if you study in details the model library ACDC permanent magnet model, this gives you a good introduction Do not forget to update your model library regularly, COMSOL adds new models or updates old ones on a weekly basis

--
Good luck
Ivar

Hey, sorry to post on an old forum, but my question closely follows this one.

I'm trying to model two permanent magnets in close proximity to one another. I'm on an education (through the university) license, and therefore only have access to 2D Magnetic Fields. I can't do Magnetic Fields with No Current on 2D or 3D, and I can't do Magnetic Fields in 3D... It's all symmetrical so it should be fine. My question is whether or not I can do my modelling of permanent magnets using the 2D axisymmettric MF module.

I was trying to follow www.comsol.com/showroom/documentation/model/78/, but obviously couldn't since I don't have access to that module.

I was able to semi-recreate what you had posted in post #2, and that seems to help.. I just want to make sure that this module is giving me accurate models.

Thanks!

Hi

I want to simulate a solenoid(which is not symmetric) both in 3D and 2D but I just don't know how to deal with MF part
Id be glad if you could help me with both 2D and 3D or either of them

Thank you in advance

3d option in add physics has no Magnetic Field option in Comsol4.2....
the 2d option has Magnetic Field in Add Physics section

hello
i am using Comsol 4.3. I want to create permanent magnet using comsol....i want not two infact upto 5 permanent magnets next to each other ..could you please tell me if you solved it.

i would want to put a superconductor on top of these magnets and look at their results
Thank you.