Learning Forum

[周世恒]

Both of them are NdFeB magnets, and the magnetic poles of both magnets are beyond the X axis.

[DELI]

how does the mesh look like?

[周世恒]

hi´╝îTwo magnets set to 3000 elements each.

region set 100000 elements

[Paul Larsen]

If you just need static snapshots of the force at different positions, this should be very straight-forward in the magnetostatic solver by creating a parameter to physically rotate the magnet for each snapshot solution. The reason this will be so effective is because the adaptive meshing in the static solver will help refine the mesh specifically where its needed between the 2 magnets. You might consider trying to compare a couple example points from magnetostatic to the transient to show the overall trend and consistent results.
For the transient solver, we have to consider the mesh over the whole range of motions. You might consider focusing the mesh around the magnet objects. You can use nested objects with the same material as the background to assign additional mesh refinement only in the region of interest. (Note that you don't need to subtract nested objects as long as there is no partial overlaps.) For the motion, you should also assign mesh operations within the Band object, and consider using a container object nested outside the band optionally.
The one trick I can suggest is to add a "container object" around each magnet (can be nested within other objects used for mesh refinement), where this object is the same material as the background, but the faces are offset, slightly outside from the cylinder, and can be some simple shape like a polyhedron with 16 segments (no holes needed). You can use this container object for mesh control close to the magnets, but the real trick is to include both the magnet and the container objects when assigning the force and torque parameters. This improves the force calculation sensitivity very much (in both static and transient solvers) especially for magnet objects.
I'm confident with the combination of these suggestions you can improve the noise in your results and see a better trend whether using magnetostatic or transient.