Discussion Forum

I haven't checked your model, but your description immediately leads me to ask you this: Are there any energy losses in your model? If not, then motions can last forever.

Hi
I have also looked at it, well actually you gave us the volume, so I made a brand new model from a cut in 2D in the midle of the magnet and solved it with (2D emqua) just to see what Maxwell stress tensor forces you get, the wire are then just 2 points, and the depth is by default "per meter".

And even without current (I=0) I get quite large (mN but still !?) forces in X and Y, these are rather mesh dependent, therefore I have tried to split the volumes through their symmetry planes, recenter the air gap on the magnet, refine the mesh, added infinite elements on the border, still the residual forces (magnet alone no current) are surprisingly high (?).

When you couple this into a structure (you can make it by extruding your 2 structural parts into a 3D geometry and add a structral 3D smsld application mode thereon) you get a constant deformation even without current, which indeed does not make sens to me.

Now you havn't told us the current directions in the wires, neither the magnetic Br direction, numerical values are not really relevant, but specific BC's yes (bondary conditions)

It is indeed puzzling, I will have to find some more time (that's what's lacking, for me ;) to try again

Goo luck
Ivar

I haven't checked your model, but your description immediately leads me to ask you this: Are there any energy losses in your model? If not, then motions can last forever.

Hi Robert,
I am not setting any energy loss. Since it is DC, it shouldn't vibrate, it should just stay up or down depends on the direction of current!

Thanks for help!

Ivar,
Thanks for help again. That is the part I am also confused about. I set the current both in +X direction (the magnetic Br is in +Y) . But when I set the current value, the force on it almost doesn't depends on the current value and mostly depends on the magnetic field, which absolutely doesn't make sense to me.

By the way, do you know do I have to put the device in a box full of air. Can I set it in vaccum?

Hi
air or vacumm in AC/DC is normally the same, it depends what you are usign as ur, er sigma.

But I know by experience that the Maxwell stress tensor calculations rely on having the object for which we calculate the force fully surrounded in an air/vacuum media, if you put a magnet directly against some iron you get a wrong force calculations because of the iron-magnet iron interface which is violating the hypothese used for the maxwell stress calculation.

That one get a little dependece on the current and more on the magnet, is rather expected, the magnetic field you generate with a simple wire is much less than what ou normally can store in a magnet, just look at the relative fields generated (that is why I always start with a simple static model, if possible in 2D, and youre case can be started in a 2D section, with only the rectangular magnet section and two points for the 2 wires, with normal (out of place currents), and a box of air/vacuum around). I would suggest to use infinite elements around the box, otherwise you must make it very big one and use rather fine mesh.

Nevertheless, what are we doing wrong to get such large residual force values for a static magnet surrounded by air ? (in fact a few mN is not that large, but because our partes are only a few microns across, your mass is very small and the resultng accelerations become large).

I need to study this further, but I have only time for that over the week-ends, need some sleep during the nights ;)

Good luck
Ivar

Hi Ivar,
I am studying the Virtual Work module see if this one works. I will try 2D model, but I am not sure how to show the bending of the bridge in 2D. My boss wanna see the real move of the bridge....ummmmm. Anyway, I agree that it is better to start with 2D.
Thanks for help and let me know if you get any new idea on that!

Shu

Hi Ivar, a quick question plz,
if I have this model need magnetostatics analysis and structure analysis. I solve it using the magnetostatics then I click on the stored solution in solver manager. (I am using the global expressions for surface tensor) But then I use structure analysis, the value of the global expressions are missing...
why is that?
I deleted the magnetostatics module when I use structure analysis
or I just need to deactivate all the subdomains under it?

I am using static btw!


Thanks for help!

Hi
Back again, I have found one thing about Maxwell Stress Tensor calculations, you need a very symmtric mesh, otherwise you have severe roundoff errors.
As to now, I was always working on 1/4 symmetry for my Maxwell Stress Tensor calculation cases, so I have never really noticed it, as I symmetryse my results by "hand" or separate equations.

Comsol does not have any way to copy and symmetrise a full mesh and the physics,
(COMSOL developpers: a pity it's wery useful in many cases, and easy to implement no? ;)
You can only copy by symmetry/mirror the geometry, it's already a good point.

The best way to get symmetric mesh is to use the mapped mesh, to ease the meshing you can divide your volume in a set of rectangles, you get some extra internal boundaries but that is not really important.

Try it out and report back

Good Luck
Ivar

PS I have posted a simple maxwell tensor calculation example on the Model Exchange, see

www.comsol.com/community/exchange/71/

This Info is really helpful! Thanks all!

Ivar,
Did you tried th Virtual work approach to get the force? I am working on the similar problem now and trying to do so. But it seems that moving mesh part is complicated. I am modeling a metal particle inside a magnetic field.
The magnetic force is to move the particle that I think I need the moving mesh.

Hi

I havnt tried out with something equivalent to "particles" *floating*" around, but I have played with the example in the ACDC model reference guide "Electromagnetic Forces on Parallel Current-Carrying Wires" and made variants there around. There migh be some new examples in the "Model Exchange" as a large set of new ones on ACDC have just arrived, I havnt had time to look more carefully into these, there might be also an example therin.

The maxwell stress tensor is correct so long you manage to get a regular and fine enough mesh around your particle, but I agree it can be difficilut to tell when its really OK, the virtual work is anyhow, I beleive, better but somewhat more complex to set up, as you point out.

Good luck
Ivar

Hi Ivar,
I read your "Maxwell Stress Tensor "pitfall" example". But I don't understand "the result is coherent to > 10 digits !" . How do you know the results of the Geom1 is wrong. I am not sure I am looking what you are. I go to postprocessing>>Global Variable Plot>>Magnet2_forcex(y)2_emqa2 to see what the value of the total force is. Why does it change during time....? The magnetic field is static..I guess.
Thanks!

Hi

I base my main conclusions for "geom1" on the comparison (and low match) of the left versus right or upper/lower values of the integration of Magnet_nTx_emqa (respectively "y") along the boundaries.

The model is perfctly symmetric so the results should be the same (or of opposite sign) in the 4 quadrants.

But we get differences that are quite large and can easily be of several mN, or more. If you use these values for a tiny particle of a few mgs will be largely accelerated, while we are in a static problem, there should be no net forces around.

Now if you look into geom2 and calculate the same variable over the same boundary the result matches far better, to many digits.

Now, I agree, are they correct in absolute ?
that I cannot state, but at least the result is a magnet / particle at rest under its own static field in free space, which is not the case/conclusions for geom1

Do we agree now ?

Thanks for your comment, indeed I made the model very quickly and didnt put enough effort to document it with a full slide presentation.

Ivar

Hi Ivar,
I guess I really need your help here. I am doing a magnetic force simulation. Basically, I am using a horse-shoe coil(carry 100mA DC) to actuate the magnet on top of a bridge. One side of the bridge is fixed. I need to show the transient simulation of how this square magnet move under the DC current. Attached is my model.
Do you have any idea on that? Do I need to use quad meshing? It seems hard to mesh it in 3D using mapped mesh...

Hi

indeed quad mesh is not easy, to try that I would subdivide the air volume to make everything as an assembly of rectangles. For maxwell tensor force calculations I would not try without a very regular mesh.

yuor model is also so symmetric that yo can cut it in two to start lighter.

On the other hand you get better results with the virtual work approach, but it's slightly more complex to set up the model.

There is a nice example in the documentation, acdc user guide chapter 2 Force and Torque Computations, V3.5a

Good luck
Ivar

Hi everyone,
I'm just starting with comsol and trying to solve very simple systems like two spherical bodies with different tensions between them. I have coupled the Electrostatics and the Solid Mechanics modules defining the electromagnetic force acting on one of the two bodies, and for the time being it is working.
Well, the problem appears when I try to calculate the force on the two bodies at the same time. I hoped that the two bodies would approach, however they don't.
Does someone know what I'm doing wrong?
Thanks a lot.
Miquel