[Getdp] Nonlinear BH

Christophe Geuzaine geuzaine at acm.caltech.edu
Sat May 4 05:12:11 CEST 2002


janvrot at infolink.com.br wrote:
> 
> Ok, All
> 
> Substituting "good" scaled data for InterpolationLinear the ILU message
> is gone, convergence is within 27 calculations, but results are really
> bad. Here is the data:
> mu_steel_couples = {
>     0.000, 0.000,
>   159.155, 0.0012566,
>   238.733, 0.0021782,
>   557.043, 0.0017952,
>  1591.550, 0.0008922,
>  3183.100, 0.0005026,
>  7077.800, 0.0002549,
> 29076.400, 0.0000735304,
> 79577.500, 0.0000286513
> };
> 
> Substituting 0.0012 for all the lines in the second column things go
> right, but then we are going linear, not the case for my application
> 
> So, still don't know how to treat nonlinear. I supose Resolution has to
> be interactive, but I did not understand the reference in the manual.
> Perhaps I'll have to review some theory or the foundations of getdp. Is
> there available any material on that?

...

> > Resolution {
> >   { Name MagSta_phi;
> >     System {
> >       { Name A; NameOfFormulation MagSta_phi; }
> >     }
> >     Operation {
> >       Generate[A]; Solve[A]; SaveSolution[A];
> >     }
> >   }
> > }

You need to specify an iterative loop in the resolution. The simplest is
the Picard scheme, i.e.

NL_NbrMaxIter = 30;
NL_Eps = 1.e-4;
NL_Relax = 1;

Operation {
  IterativeLoop[NL_NbrMaxIter, NL_Eps, NL_Relax] {
    GenerateJac[A]; SolveJac[A]; 
 }
 SaveSolution[A];
}

Christophe

-- 
Christophe Geuzaine

Tel: (626) 395-4552    http://www.geuz.org
Fax: (626) 578-0124    mailto:geuzaine at acm.caltech.edu