Questions about GetDP - Valparaíso - Chile

Christophe Geuzaine Christophe.Geuzaine at ulg.ac.be
Thu Jul 27 11:17:34 CEST 2000


Alejandro Angulo wrote:
> 
> I find the information thts you sent me extremely
> useful (in my formulation, the only thing that it
> lacked was to add the terms <T> and <a> // which is
> the function of the symbol <> in GetDP ????), 

Hi Alejandro,

The <>[] operator is a "cast" (a bit like in the C programming
language). It forces the type of its argument. It is needed here because
the the "a-v" formulation is written in complex arithmetic, and the
"T"one in real arithmetic. So, I force the type of {T} with <T>[{T}],
so that "T" is not converted to a complex number. (Well, it is a bit
tricky: we'll try to suppress that in future versions...)

> but do I
> need to include in my formulation convection and
> radiation.
> Apparently, my problem here has relationship with the
> definition of the integral on the boundary of the
> studied domain; for example, at the moment, I have
> defined in the following way the integral associated
> to the convection condition:
> 
> Galerkin { [ hc * (Dof{T}-Ta) , {T} ]; In 
> ;
> Jacobian Jac ; Integration Int ; }
> 
> Please, could you send me information in this respect
> and to indicate me the problems of my current
> formulation??. The parameters 'Jacobian' and
> 'Integration' should be defined in a special way or
> are the same ones used in the rest of the
> formulation??.

Yes, I would introduce the term in the same way. Pay attention that you
have to specify a SURFACE jacobian (you integrate on a 1D domain, placed
in a 2D framework). So you should use something like:

Galerkin { [ hc[] * (Dof{T}-Ta) , {T} ] ; 
           In Frontera_2D ; Jacobian MyJacSurf ;
           Integration IntSurf ; }

with 

Jacobian {
  { Name MyJacVol ;
    Case { { Region All ; Jacobian Vol ; }
    }
  }
  { Name MyJacSurf ;
    Case { { Region All ; Jacobian Sur ; }
    }
  }
}

(If your problem is axisymmetric, replace Vol and Sur by VolAxi and
SurAxi.) And you should have integration methods defined for line
elements, too.

> On the other and, I also need to define connetions
> among conductors inisde my model. How I carry out this
> connections (series or parallel)?? , If my interest is
> to analyze transformers, will it be very complicated
> to define the problem due to the great quantity of
> involved conductors???

You have to consider global quantities to define the current/tension,
with a stranded inductor approach. When I solve such problems, I extend
the "a-v" formulation somehow like this. I define a special function
space for stranded coils:


  /* Current in stranded coil (2D) */
  { Name Hregion_i_Mag_2D ; Type Vector ;
    BasisFunction {
      { Name sr ; NameOfCoef ir ; Function BF_RegionZ ;
        Support DomainB_Mag ; Entity DomainB_Mag ; }
    }
    GlobalQuantity {
      { Name Ib ; Type AliasOf        ; NameOfCoef ir ; }
      { Name Ub ; Type AssociatedWith ; NameOfCoef ir ; }
    }
    Constraint {
      { NameOfCoef Ub ; EntityType Region; NameOfConstraint Voltage_2D ;
}
      { NameOfCoef Ib ; EntityType Region; NameOfConstraint Current_2D ;
}
    }
  }

And I introduce the global quantities Ib and Ub in the formulation like
this (Ns and Sc designate the number of turns and their cross section):

  { Name Magnetodynamics_av_2D ; Type FemEquation ;
    Quantity {
      { Name a  ; Type Local  ; NameOfSpace Hcurl_a_Mag_2D ; }
      { Name ur ; Type Local ; NameOfSpace Hregion_u_Mag_2D ; }
      { Name ir ; Type Local ; NameOfSpace Hregion_i_Mag_2D ; }
      { Name Ub ; Type Global ; NameOfSpace Hregion_i_Mag_2D [Ub] ; }
      { Name Ib ; Type Global ; NameOfSpace Hregion_i_Mag_2D [Ib] ; }
    }
    Equation {

      (... classic terms: cf. last e-mail +)

      Galerkin { [ - Ns[]/Sc[] * Dof{ir} , {a} ]  ; In DomainB_Mag ; ...
}
      Galerkin { DtDof [ Ns[]/Sc[] * Dof{a} , {ir} ] ; In DomainB_Mag ;
...}
      Galerkin { [ Ns[]*Ns[]/(Sc[]*Sc[])/sigma[] * Dof{ir} , {ir} ] ;
... }
      GlobalTerm { [ Dof{Ub} , {Ib} ] ; In DomainB_Mag ; }
    }
  }
}


-- 
Christophe Geuzaine

Tel: +32-(0)4-366.37.10    mailto:Christophe.Geuzaine at ulg.ac.be
Fax: +32-(0)4-366.29.10    http://www.geuz.org