Re: [femm] Problem in force calculations

[Dave Squires <djsquires@xxxxxxxx>]

David Meeker wrote:

> Dave Squires wrote:
> \Anyhow, there are several things that you might look at that could be
> influencing things. A lot of these issues are summarized in section 4.10 of
> the manual. The first is the line over which you perform the integration.
> Generally, it's a lot more accurate to calculate forces along a contour
> through the air that surrounds a body, rather than on a line at the surface
> of the body. The reason is that there can be discontinuities in B and H at
> the surface of a ferromagnetic or PM body. These sudden changes aren't
> resolved that well by a first-order triangle formulation. In theory, the
> integration path doesn't matter so long as it encloses of interest, so more
> accurate results are obtained by drawing a path through the air, where
> things are better behaved. The second tip is to always use a relatively
> fine mesh in the area where perform the integration.

OK, I did do both of these things, fine mesh in air.

> Lastly, be careful
> that you have located your boundaries adequately far away from the magnets.
> For example, the "default" boundary condition (normal derivative of A=0)
> functionally behaves like the surface of an iron object.

This is one thing I wondered about. I did not know how the boundary proximity
affected things. I suspected it might. If it behaves like iron then I need to
keep
it quite long way away from a neodymium magnet. Problem is that the simulation
time goes up substantially if the mesh needs to be fine all the way out.
Can I keep the fine mesh in close proximity, say 10 to 20 elements away and
go to a coarse mesh farther away. Will this affect the force results much?
I would think it might introduce some small error.

> Permanent magnets
> are attracted to the boundary in just the same way as if you'd put a big
> metal plate there. As long as the boundaries are "far enough" away, there's
> no problem--the forces from the boundary as small enough that they don't
> muck things up too much. If the boundaries are too close, they influence
> the forces of the object of interest in a significant way. (See appendix A3
> for more stuff about approximating "open" domains).

BTW, I was using simple rectangular cross sections of equal thickness for
the magnets and the iron section. I can send you the file if you like.
I would like to know the conditions that get the right answer so I can
have a good "feel" for what is the right way to do it when doing other
simulations.

Dave Squires