Re: [femm] Re: Incorrect force direction

[dcm3c@xxxxxxx]

In a message dated 12/31/00 10:42:43 PM Eastern Standard Time, 
gowatson@xxxxxxxxxxxxxxx writes:


> Hi Rob,
>
> Removing one of the parallel bars eliminates the return flux path 
> between the bars.  This return flux direction between the bars is 
> such that the magnet's own return flux will be in the same direction 
> and thus in opposition.  This is where the negative force (to the 
> left) comes from.  This force is correct and is present but is not 
> the whole picture.
>
> The problem, as I see it, is that FEMM doesn't take into 
> consideration how the magnet's flux distribution will be effected by 
> the different working points encountered in the parallel bars.  I 
> think it assumes a symmetrical distribution where it should be skewed 
> to the right.  This should generate a positive (to the right) force, 
> which should be must greated than the much weaker negative (to the 
> left) repulsive force generated above.
>
> I'm sure David will work it out.


The program is doing the right thing--the force generated by your 
configuration is just really close to zero.  The force reported by your 
example problem is about -5.8 Newtons per Meter of length in the 
into-the-page direction.  For those of us who tend to have more of a feel for 
English units, that's about -0.033 Pounds/Inch, or very close to zero.  The 
seemingly odd sign is an artifact of the accuracy of this type of calculation 
not being that accurate to begin with (see discussion of manual on this 
point).  I'm pretty sure that I got the signs right and that I'm evaluating 
the integral correctly.  As an example, you can look at the qfld1.fem example 
that comes with the distribution.  There is an analogous problem, magn1, that 
comes with the quickfield student version (freely available).  The forces on 
the parts in that problem agree quickfield to on the order of 1%, with the 
same signs. &!
nbsp;This is a problem with nonlinear iron and permanent magnets, so 
it's a benchmark that's of a similar class of problem to this one.

Anyhow, the reason that the force is very nearly zero is that once the magnet 
in your problem is between the iron bars, there's not much incentive for the 
magnet to want to center further.  The reluctance of the flux path that the 
magnet sees has almost no dependence on position once the magnet is between 
the bars, which translates into a very small centering effect.  

A different and fun way of thinging about it is to consider that at the 
surface of highly permeable materials, the stress tensor reduces to 
B^2/*(2*mu_o) where mu_o is the permeability of free space and B^2 is the 
flux density passing normal to the surface of the iron.  This force acts in a 
direction that is perpendicular to the surface of the iron.  This means that 
the only part of the iron that is "contributing" to a lateral force is the 
very ends of the bars.  When the magnet is between the iron bars, very little 
flux passes out the end of the bars, and the force is close to zero. 

Also, as far as the magnet model, the program assumes that the magnet is 
uniformly _magnetized_ in the specified direction, but this doesn't imply 
anything at all about the levels of flux anywhere in the magnet.  The 
magnet's interaction with everything else (i.e. a finite element problem) has 
to be solved to determine the flux levels, "working point", etc.

Dave.