[Getdp] Re-2: Magnetic force
Colignon David
David.Colignon at ulg.ac.be
Mon Aug 1 10:51:08 CEST 2005
Hi,
in the file solver.par, try to increase the Nb_Fill value from 20 to 40 or
60 or ... until the solver converge to a solution.
Dave
--
David Colignon, Ph.D.
ELAP - Service d'Electricité Appliquée
Institut Montefiore B28
Université de Liège
4000 LIEGE - BELGIQUE
Tél: +32 (0)4 366 37 32
Fax: +32 (0)4 366 29 10
mailto:David.Colignon at ulg.ac.be
http://elap.montefiore.ulg.ac.be
> Dear Christophe,
>
> thank you for the hints. I used an older version of gmsh which had no
> gradient plugin. Could I ask you again for an example for the Maxwell's
> stress tensor or virtual works approach?
>
> I am trying to solve the magnetic force problem in a split coil geometry
> (see the attached files). Unfortunately the body in the gap is not
> axisymmetric. So I have to go 3D.
>
> Even the basic problem without a body is not solved, using default
> parameters in solver.par. When I reduece the characteristic length (not
> touching anything else) I get different error messages from getDP:
>
> lc = 6*mm
>
> P r o c e s s i n g . . .
> Operation : Generate[A]
> Solver : Loading parameter file 'solver.par'
> Info : Setting System {A,b} to zero
> Resources : cpu 7.801000 s / mem 20180 kb
> Operation : Solve[A]
> Solver : No scaling of system of equations
> Solver : RCMK algebraic renumbering
> Solver : N: 18322, NZ: 236746, BW max/avg: 38/12, SW max: 18247
> Resources : cpu 8.71000 s / mem 22056 kb
> Solver : ILUTP (Float, fill-in = 20)
> Solver : N: 18322, NZ: 709271, BW max/avg: 40/38, SW max: 2938
> Resources : cpu 22.993000 s / mem 27624 kb
> Solver : Generalized Minimum RESidual (GMRES)
> 1 Inf NaN
> Solver : 1 Iterations / Residual: NaN
> Resources : cpu 23.43000 s / mem 20264 kb
> Operation : SaveSolution[A]
> Resources : cpu 23.543000 s / mem 20272 kb
> E n d P r o c e s s i n g
>
>
> lc = 4*mm
>
> P r o c e s s i n g . . .
> Operation : Generate[A]
> Solver : Loading parameter file 'solver.par'
> Info : Setting System {A,b} to zero
> Resources : cpu 38.305000 s / mem 66580 kb
> Operation : Solve[A]
> Solver : No scaling of system of equations
> Solver : RCMK algebraic renumbering
> Solver : N: 76863, NZ: 988519, BW max/avg: 48/12, SW max: 76756
> Resources : cpu 39.616000 s / mem 78216 kb
> Solver : ILUTP (Float, fill-in = 20)
> Error : Zero row encountered in ILU
> Stop Fri Jul 29 11:00:26 2005
> Runtime: 227 Seconds
>
>
> Do you have an idea where to search for a remedy?
>
> Best regards
>
> Matthias
>
>
> ---------------------------------------------------------------------
> Matthias A. Fenner
>
> email: m.fenner at gmx.net Hugo-Eckener-Str. 94
> fon: +49 6131 622 10 60 D-55122 Mainz
>
> ---------------------------------------------------------------------
>
>
>
>
>
>
>
>
> -------- Original Message --------
> Subject: Re: [Getdp] Magnetic force (08-Jul-2005 22:32)
> From: geuzaine at gmail.com
> To: m.fenner at omicron.de
>
>> Matthias A. Fenner wrote:
>> > Am Thu, 07 Jul 2005 22:31:30 +0200 hat Christophe Geuzaine
>> > <geuzaine at gmail.com> geschrieben:
>> >
>> >> Matthias A. Fenner wrote:
>> >>
>> >>> Dear Christophe,
>> >>>
>> >>> I am trying to calculate the force on a paramagnetic body by f =
>> >>> chi/mu_0 (B * nabla) B. Is it possible to evaluate something like
>> >>> Grad[CompX[{d a}]] in the post processing of a magnetostatic
>> problem?
>> >>> My first straight forward attempts to do so failed. Is there a hack
>> >>> to get it?
>> >>
>> >>
>> >> Not directly in getdp: there is no "general" (numerical)
>> implementation
>> >> of differential operators, only support for those which are
>> implemented
>> >> as basis functions.
>> >>
>> >>
>> >>> If not: is there an alternative approach to calculate the force
>> >>> (-density)?
>> >>
>> >>
>> >> Yes, there are nice ways to compute the force (using virtual works or
>> >> Maxwell's stress tensor). I'm not sure if there are examples on the
>> list.
>>
>> >
>> >
>> > I fear there are not, please correct me if I am wrong.
>>
>> I did a quick search and did not find any hits either. I'll try to post
>> an example on the wiki if Patrick of Francois don't.
>>
>> >
>> >> If not, we should post an example on the wiki. Francois and Patrick
>> >> have probably ready-to-use .pro files for this.
>> >>
>> >> If you want to use the formula you mentioned above, you could also
>> use
>> >> some external tools (like Gmsh) to directly evaluate the expression
>> >> numerically.
>> >>
>> > I generally use gmsh for pre and post processing. I also tried to use
>> > the Extract and Evaluate plugins to calculate this expression, but
>> > failed to implement the derivative. Could you give a more detailed
>> hint
>> > on how to do this?
>>
>> You can use Plugin(Extract) to extract the x-component of B, then apply
>> Plugin(Gradient) to compute the gradient.
>>
>> Christophe
>>
>> >
>> >> Christophe
>> >>
>> >>
>> >>
>> >>>
>> > Thanks in advance and best regards
>> >
>> > Matthias
>> >
>> > ---------------------------------------------------------------------
>> > Matthias A. Fenner
>> > email: m.fenner at gmx.net Hugo-Eckener-Str.
>> 94
>> > fon: +49 6131 622 10 60 D-55122 Mainz
>> > ---------------------------------------------------------------------
>> >
>> >
>>
>>
>> --
>> Christophe Geuzaine
>> Applied and Computational Mathematics, Caltech
>> geuzaine at acm.caltech.edu - http://geuz.org
>>
>
> _______________________________________________
> getdp mailing list
> getdp at geuz.org
> http://www.geuz.org/mailman/listinfo/getdp
>