[femm] Re: FEMM files' format and other

[Dcm3c@xxxxxxx]

In a message dated 1/10/00 10:44:40 AM Eastern Standard Time, 
tecnico@xxxxxxxxx writes:

> My company makes electrical machinery.
> I have written my own c++ subroutines to read and write femm's input and 
> output files. This is useful for interfacing directly my calculation 
programs 
> with femm kernel. My library works for 2.1 version files. Shall I expect 
file 
> format changes in version 2.1a or next versions?

I've been working on a version 3. There are a lot of new features, and that 
has necessitated a new file format. I've tried to iron out the file format 
so that it isn't as hare-brained as the 2 file format. The new file format 
is philosophically like the dxf format, in the sense that it contains a 
series of keys and values for these keys. The advantage of this type of 
format, as opposed to the old one, is that it is more robust. If I add a new 
feature, that's just a new key that gets added to the file. If the file is 
read in with an older version of the program that doesn't understand the key, 
it can just ignore it (rather than choking...). In the other direction, if a 
newer version is reading an older file, it just reads in the keys in the file 
and assumes default behavior for everything else.

> If so, may I have also a description of the new formats? (I have 
understood 
> them one time, but it is not nice work)

I will write up a description of the file format when I release the version 3 
program.

> I would like to run triangle.exe and fkern.exe from my programs, what 
> command line must I invoke?

Well, the problem is that triangle doesn't understand arc segments, some the 
femme preprocessor has to do some sort of "premeshing" to break down arc 
segments into a bunch of lines, write an input file that triangle 
understands, and call triangle. I've written a little command line program 
that does the "premeshing" part and writes out the datafile that triangle 
needs. I've put it up at:

http://www.egroups.com/docvault/femm/

The particular file is premesh.zip. You should copy the premesh.exe 
executable to the femm\bin directory. The included file, solve.bat, shows 
how one calls it to analyze a *.fem file without entering the preprocessor 
explicitly. If you check out the premeshing program, let me know if you get 
it to work ok.

> Is it possible (or will it be in a future version) to run femmview.exe 
from 
> command line asking for some operation and having the result on a file? (I 
> think of somthing like"femmview.exe -AJintegral inputfile outputfile", 
where 
> inputfile contains the line and outputfiles the result; at the moment I do 
it 
> in my subroutines).

I have been thinking about the best way to do this sort of thing, and I will 
probably put some sort of more automated capabilities in. I don't yet know 
if this will make it into the next release, or one farther down the line.

> I have seen that currents defined as point properties are not taken into 
> account in AJ integral (and perhaps in some other calculation); must I 
always 
> use block properties in order to have a correct result in AJ integral?

Yes--the A dot J integral only gives valid results for current densities. 
The reason is that point currents have an infinite inductance in theory, 
since the value of A goes to infinity at the point current's location for an 
analytical solution. In a finite element solution, A doesn't really go to 
infinity at the point current location, because the effects of the current 
are distributed to the elements that contain the point.

An interesting demonstration is to consider a cylinder of air 2 cm high and 2 
cm in radius, with a 1 cm radius current ring of 100 A centered in the 
cylinder. A=0 is applied to the boundary of the cylinder. One can apply an 
increasingly fine mesh density and then evaluate the stored energy (which is 
closely related to inductance). One gets the following results

Sidelength Stored Energy
0.25 cm 0.0177 J
0.125 cm 0.0221 J
0.050 cm 0.0265 J
0.025 cm 0.0309 J
0.0125 cm 0.0352 J
0.005 cm 0.0418 J

The stored energy keeps increasing as the mesh density gets finer (and the A 
dot J integral is basically a different form of stored energy). In contrast, 
a coil with a finite cross-section area converges on one specific value of 
stored energy as the mesh density gets finer. In short, point currents can 
be a convenient fiction in some cases, but you can't really use them to 
compute inductance.

Dave.
--
http://members.aol.com/dcm3c