[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [femm] Torque integration
ito ito wrote:
Hello! My apologies if this is a stupid question, for I haven't been
on this list or in the magnetics business for very long.
I am modeling a three-phase permanent magnet alternator for use in a
small wind turbine. Of immediate interest is the cogging (detent)
torque, which must be overcome for the machine to begin turning from a
stop.
I have read the manual section on the torque integral, set up my air
gap with at least four nodes between magnet and teeth, and the field
solution that is generated looks qualitatively reasonable. But the
values given by the torque integral around the bore are inconsistent.
First of all, the sign of the value seems to switch randomly, perhaps
because of some subtlety of how I highlight the curve.
Secondly, my method is to repetitively solve the geometry with the
rotor rotated incrementally some increasing fraction of a tooth pitch.
(If anyone knows a better method for cogging I would be thrilled to
know it; could this be scripted with the "lua" business? I've been
re-exporting the cad file and redefining everything for each
increment.) I was expecting to see a smooth, roughly sinusoidal
result of torque vs angle, but so far the data appear random, though
the order of magnitude (~3Nm/m) is in good agreement with experimental
results for peak cogging torque.
If anyone has tackled this problem I would be very thankful for advice.
Thanks,
Ben Polito
It can often take a pretty fine mesh to do a good job with cogging
torque. Cogging torques are a usually small fraction of the total
design torque, and you want to determine this small fraction
accurately. This means determining the torque to an accuracy level of
fractions of a percent of the total torque.
Anyhow, I've put an example geometry and a Lua script that analyzes the
cogging torque automatically in the Yahoo groups file section at:
http://groups.yahoo.com/group/femm/files/scripting/scog.zip
This uses the "weighted stress tensor" integration, so you need to be
running the latest-n-greatest 3.3a3 version to run it (or you can
comment out the block integral part). I also compute force using a
"regular" stress tensor integration that runs down the center of the air
gap. Both yield roughly the same answers, but it appears that the error
in the "weighted stress tensor" is about a third of that in the
"regular" stress tensor for this particular case.
Dave.
--
David Meeker
dmeeker@xxxxxxxx
http://femm.berlios.de/dmeeker