Re: [femm] periodic boundary condition problem

[David Meeker <dmeeker@xxxxxxxx>]

enp enp wrote:

> Hi,
>
> It depends on your problem. If you would like to
> nalyse no load. You may just model one pole or 1/2
> pole span of the PM machine.
> If you analyse the on load operation. you would need
> to model at least one pole ans set the anti-periodic
> conditions at the azimutal boundaries.
> You can also model two poles you must set in this case
> the periodic conditions on these boundaries.
> On the radial boundaries in the all cases set Dirchlet
> conditions, set the the vector potentiel to zero A=0.
>
> Riyad

Another issue that the original poster may have is how to apply these 
boundary conditions within the program. Unlike "regular" boundary 
conditions, which can be applied to as many different arcs or segments 
or combinations thereof, the periodic boundary conditions are special in 
that they can only be applied to either two line segments or two arc 
segments, and each component of the pair of segments or line segments 
must have the same length. For example, if the edge that you wanted to 
make (anti)periodic was composed of 5 line segments, you'd have to 
define 5 (anti)periodic boundary conditions and apply one to each of the 
segments on the edge and its mate on the adjoining boundary.

Perhaps all of this would be clearer with an example. I've attached 
pole's worth of a four pole PMSM (which is actually cribbed from 
Quickfield's Magn4 example). Antiperiodic boundary conditions are used 
so that only the one pole need be modeled. I've defined all the parts 
in the rotor to be in group number one, so that the rotor can be turned, 
if desired.

Dave.

--
David Meeker
dmeeker@xxxxxxxx
http://femm.berlios.de/dmeeker

Attachment: zip00042.zip
Description: Zip compressed data