[Getdp] Periodic Boundary Conditions

Lars Rindorf lhr at com.dtu.dk
Wed Feb 6 16:46:18 CET 2008

Dear Raymond
A simple model is always the best. I don't understand your physics well. That was I was trying to say. 
Do you know about the coefficient in the constraint?  It specifies a real number, say c,  such that u(left) = c*u(right), where 'u' is your function and 'left' and 'right' indicate the left and right boundaries. In the model I sent you I changed it to -0.5 (for no apparent reason). I think you want it to be 1 such that the solutions are symmetric around the y-axis. Is this correctly understood?
Kind regards


From: Raymond Michaux [mailto:raymond.michaux at freesbee.fr]
Sent: Wed 2/6/2008 3:06 PM
To: Lars Rindorf
Subject: Re: [Getdp] Periodic Boundary Conditions

Dear Lars,

Thank you for your prompt answer. I am sorry, I oversimplified the model,
just to make
simple and highlight my problem. I understand that the proposed geometry is
not very realistic,
this is a simple step to check the periodic boundary conditions.
Unfortunatly, the results are just slightly better with your corrections,
but not enough to go on.
You will find attached three ".jpg" pictures of the results.
    ULg1.jpg is the field map for case n°1 (a=0 on 3 sides, and a symmetry
condition on the
        fourth side y=0) : this is the reference result
    ULg2.jpg and ULg3.jpg are the results of case n°2 before and after
correction. I expected to
        find a=0 on the left and right sides. Instead of that, the field map
in ULg3 is unsymmetrical
        and vector potential a is max positive on the right side
Thank you for your opinion

----- Original Message -----
From: "Lars Rindorf" <lhr at com.dtu.dk>
To: "Raymond Michaux" <raymond.michaux at freesbee.fr>; <getdp at geuz.org>
Sent: Wednesday, February 06, 2008 9:36 AM
Subject: RE: [Getdp] Periodic Boundary Conditions

Dear Raymond

I don't know what you exactly are trying to model, but it seems to me that
your Constraint{ ...}'s were wrong. I have changed your constraint and
functionspace, and now the script seems to work according to what you have

I have attached the .pro file

Hope this helps you.
Lars Rindorf


From: getdp-bounces at geuz.org on behalf of Raymond Michaux
Sent: Tue 2/5/2008 9:39 PM
To: getdp at geuz.org
Subject: [Getdp] Periodic Boundary Conditions

Dear all,

I'am trying to use the "Periodic Boundary Conditions" in a simplified model
of permanent magnet
linear motor, according to an information given by Christophe Geuzaine on
the same topic (see mailing
list June 2005).
In order to check the method, I used a model with magnet centered on the
teeth, I used the vector
potential formulation and I ran the same problem with two sets of  boundary
conditions :
    1. A Dirichlet a = 0 on three sides
    2. A Dirichlet a = 0 on upper border and periodic conditions between
left and right side
For both cases, there was no error reported.
As the geometry is a symmetrical one, both methods should be equivalent but
the calculation results are
obviously different.

Could you help me. The files are attached for tests. The file ".don"
contains the main dimensions
and is included in both ".geo" and ".pro". For the second case, please
modify ".don" and ".pro" as
indicated inside the files.