Re: [FEMM] Accuracy error

[David Meeker <dmeeker@xxxxxxxx>]

Dewald Pieterse wrote:

> Femm gives the following error when I want to mesh my problem, 
> attached is the fem document. What am I doing wrong or is there a 
> setting that I need to change??
>
> Error:  Ran out of precision at (145.840386091, 120.550876183).
> I attempted to split a segment to a smaller size than can
>   be accommodated by the finite precision of floating point
>   arithmetic.
> Try increasing the area criterion and/or reducing the minimum
>   allowable angle so that tiny triangles are not created. 

Triangle, the mesh generator, doesn't like the small angles at the pole 
tips in your geometry.  The best way to get around this error is 
probably to use v4 of the program, where the minimum allowable angle is 
one of the items specified in the problem definition.  The default is 30 
degrees, which normally produces a nice mesh, but can sometimes give you 
the above error.  If you change the minimum angle to 20 degrees, things 
will probably be ok.

Also to note: 
Some of the mesh sizes specified in your model are very small--it looks 
like the mesh would be roughly a 2,000,000 element mesh if mesh 
generation succeeeds.  I'd guess that this is unnecessarily fine.
The thin sections on the pole tips will be very highly saturated.  You 
need to define the BH curve for the iron out farther so that the curve 
is always interpolated, rather than extrapolated.  Otherwise, you will 
get unrealistically high flux densities in those sections.
This is a good candidate for a Kelvin Transformation type of "open" 
boundary condition.

Dave.