Elmer workflow for electrical machine designers

General discussion about Elmer
mzenker
Posts: 1747
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Elmer workflow for electrical machine designers

Post by mzenker » 25 Jul 2018, 16:32

Hi,

I am not sure I understood correctly: Are you saying that someone should solve your problem for you, so that you don't have to consult the documentation...?

Matthias

ulrich
Posts: 8
Joined: 15 Jun 2018, 19:47
Antispam: Yes

Re: Elmer workflow for electrical machine designers

Post by ulrich » 27 Aug 2018, 19:36

Hello Matthias,

i was playing with Pavels wonderful workflows for electrical machine Designers.
But i would like to relize my own ideas with ELMER concerning induction machines,
as i did a couple of years before with FEMM. I made a program to compute a torque- and current over speed
characteristic and one to compute the performance data, which works very fine and quite accorate.
Both in SCILAB and in MATLAB/OCTAVE.
My job is to design induction machines and my self written software was very useful during a lot of years.


FEMM has interfaces for MATLAB/OCTAVE, SCILAB LUA and even some more.
For postprocessing for example there are numerous commands like:

mo blockintegral(type)

to calculate a block integral for the selected blocks
with 24 items like these:

Type Definition
0 A· J
1 A
2 Magnetic field energy
3 Hysteresis and/or lamination losses
4 Resistive losses
5 Block cross-section area
6 Total losses
7 Total current
8 Integral of Bx (or Br) over block
9 Integral of By (or Bz) over block
10 ...... 24.

I did not find something in the ELMER documentation, what is similar.
That's my Problem.

Best regards,
ulrich

mzenker
Posts: 1747
Joined: 07 Dec 2009, 11:49
Location: Germany

Re: Elmer workflow for electrical machine designers

Post by mzenker » 28 Aug 2018, 10:01

Hi Ulrich,

I did not study the induction machine case, so I cannot help you with the specific workflow.
However, in general here are the most common ways to get data out of Elmer simulations:

* Every solver calculates values for at least one variable, sometimes more. Some solvers have flags to activate calculation for additional variables. More information can be found in the models manual.

* There are the SaveData/SaveScalars solvers which can calculate additional variables. They are also described in the models manual.

* For postprocessing, ParaView is the recommended tool. It has countless filters to calculate almost everything from the variables in the Elmer output. You can also make your own filters with Python. More information can be found in the ParaView documentation and in their forum and mailing list.

If you already found all this and still did not get what you need, then you can try to extend Elmer's functionality. The code of the solvers is available and can be modified so that it does what you want it to do. You can even write your own solver. More information can be found in the solver and programmer manuals.

Hope this helps,

Matthias

ulrich
Posts: 8
Joined: 15 Jun 2018, 19:47
Antispam: Yes

Re: Elmer workflow for electrical machine designers

Post by ulrich » 29 Aug 2018, 16:52

Hi Matthias,
thank you very much for your advice.
Best regards,
Ulrich

raback
Site Admin
Posts: 3277
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Elmer workflow for electrical machine designers

Post by raback » 02 Sep 2018, 21:42

Hi

Matthias is right that for many of these there is no electrical machine -centric solution available but rather some generic routines in SaveScalars to integrate over stuff using different operators. With masks you can control where the integration takes place.

That being said there are also more specialized developments. There is a concept of "component" which consists of given bodies and it has some automated ways to integrate stuff over it. In test case mgdyn_airgap_force this is demonstrated. This is the recommended method for computing forces from the generalized nodal loads.

Also, there is a way to compute losses by making on-the-fly fourier series and computing the losses from different types of functional dependencies. Look at Ch 33 in current version of Elmer Models Manual.

To conclude I would think all of these could be computed even now but you're right in pointing out that it is not very streamlined. This would deserve some further effort.

-Peter

Post Reply