TR: [Getdp] Re: (no subject)

[Patrick Lefevre]

Hello Janvrot

Hello Patrick
 
> I'm also trying to solve non linear magnetostatic problems. Before asking
> questions, i have one remark. For the phi formulation, I think you don't
> have to impose a constraint on the symmetry line. .......

As of my understanding (note I'm a beginer) if you constraint a line as
a constant phi field will be normal to it; if you don't field will be
parallel. In my example a normal fiel will do it; in your case, with hc
in Y I guess the normal will deal with the symetry in the left of line
(7775). 

You are totally right. Sorry for this remark.

> Now the questions about solving a non linear problem with the phi
> formulation:
> 
> In my problem, magnet's are non linear (Alnico magnets). To give an
example
> of the method i use for solving see attached file "hlk.geo, hlk.pro,
> solver.par". I'm not sure this is the good way to solve the problem, but
the
> results are better when i compare to magnetic field measure.

I don't know how to deal with hc and non-linearity in the same region,
but it looks like you defined mu with H,B couples when it should be done
with H,mu couples. See recent discussion on InterpolationLinear Function
on the list. These same e-mails also show how to compute non linear with
the Picard scheme. Note mu[{d phi}] both in the Galerkin and in
PostProcessing, and make sure that Linear and NonLinear regions don't
overlap.

Ok, but the problem is that the information about materials are always
associated with B-H curve and never with mu-H curve. I was thinking that the
b-h curve slope gives a good approximation of the permittivity.

I'm now looking forward to additional help in the phi formulation with a
Newton scheme (Garlekin and PostProcessing), to run 3D with strong non
linearity. For 2D you may eventually consider the vector potential
formulation by Johan in this thread.

Greetings

I also try now to solve 3D magnetostatic problem but with little non
linearity.
Thanks for your help
Bye
-- 
Janvrot IVM
janvrot at infolink.com.br