Torque between two cylindrical magnets

[nakaAnsys]

Hi,

I am new in this forum, I hope somebody can help me to answer my question.
It seems like a simple problem.

I have 2 cylindrical magnet which have opposite magnetic direction,one is placed in the top of the other (with small gap).
I have tried to use a FEA-based software (Maxwell) to simulate the problem, I got a reasonable result of repulsive force between these 2 magnets. But the problem is that I got a significant value of torque between these 2 magnet.
I was thinking that the torque produced should be 0.
Is there any idea about this result?

 

[BrianG]

I suspect the field distribution will not be equal around the full 360 degress of each magnet, therefore the stronger portions of each field will repel harder to produce a torque, so that the magnets settle with the weaker parts of the field aligned but still opposed. (Remember the school physics lab experiment with the bar magnets, one suspended on a string above the other on the bench. The suspended magnet rotates quickly to align N-S, S-N with the one on the bench below.)

 

[nakaAnsys]

Brian,
Thank you for your answer,
I agree with you about non-uniform magnetic field distribution.
But since I got this result from a simulation, I assumed that the magnetic field produced by each magnet would be identical (if not depend on the mesh).
Do you think we have analytical method to calculate torque/force for a simple cylindrical/bar magnet?
Thank you in advance.

 

[electricpete]

In the ideal case where the geometry is symmetrical and the magnets are uniformly magnetized axially, of course no torque (by symmetry... which direction should it turn when no direction is different). When the symmetry is disturbed slightly a torque may occur.

I would suggest to re-run the simulation with finer mesh density (especially in the gap and the highest-flux areas) to see if the computed torque decreases. If it decreases, then try it again with even smaller mesh density....keep reducing mesh density until computed torque no longer decreases... I suspect the computed torque would be very small at that point.

=====================================
(2B)+(2B)' ?

 

[electricpete]

Also I would try centering your geometry within the simulation domain (both axially and radially) and double-checking the boundary conditions used at the edge of the simulation domain.

=====================================
(2B)+(2B)' ?

 

[MagMike]

As electricpete mentioned, it is the size of you mesh. Any small non-uniformities in the distribution will come out as torque.

Since your geometry is rotationally symmetric, you'll have a simpler time if you model in 2D.

 

[electricpete]

Good point about 2D. In FEMM software, it would be the “axisymmetric option”. It exploits the symmetry about the center axis to fully represent the 3-d problem using a 2-D mesh representing a slice of the r-z plane (with appropriate changes for cylindrical form of the differential equations).

=====================================
(2B)+(2B)' ?

 

[nakaAnsys]

Thank you all for the suggestions and advices.

In the first model, I have already placed my model to the center of simulation domain.
Based on the suggestions, I tried to refine the mesh and re-run the simulation, -> the torque decreased up to micro Nm for a model size 80mm in D and 2mmm gap!!
Probably I will make a 2D model and see the differences.

Thank you.