help with GetDP needed

Christophe Geuzaine Christophe.Geuzaine at ulg.ac.be
Wed Dec 1 14:56:40 CET 1999


Samuel Kvasnica wrote:
> 
> > Your mistake was that you associated the HBF basis functions to the
> > nodes, not to the edges...
> 
> Well, I'm really blind, I checked everything twice and didn't notice that. Now it works
> perfectly, very smooth results. Thanks a lot.  Is there some reasonable short .pdf or .ps
> document
> covering overview of FEM and explanations of basis functions, integration methods etc. ?

Hmm, I dont't have something in mind right now, but we were thinking
about adding some more 'technical' comments to the user's guide, just to
explain a bit more the fundamentals. Otherwise, I could recommend you a
very good book, by Alain Bossavit :

@book{mixt:bossavit:98,
   author = "A. Bossavit",
   title = "Computational Electromagnetism, Variational Formulations,
Edge
            Elements, Complementarity",
   publisher = "Academic Press", 
   year = 1998}


> 
> > PS : is your problem truly 2D? Or is it axisymmetric? In the latter,
> > your jacobian methods are not correct...
> 
> Yes, it's truly _axissymetric_. Seems like I forgot to replace Vol and VolSphShell by
> VolAxi and VolAxiSphShell in jacobian. So it seems like I was computing  field of
> 6 infinite inducotors with infinite core. Now I replaced it. There's almost no difference
> in
> the form of the field but the results don't quantitatively correspond to my measurements !
> The Br component should be 0.07T 17mm from axis at pole position (with 150 windings of
> inductor with
> 7A current at each segment), which was in a strange way
> true for Vol and VolSphShell. Now it is ~50 times less. Well, actually a factor of 16*pi.
> It there something else to be changed except jacobian ?


Oups, yes, there is an another small error in the PostProcessing field:
you should give each postprocessing quantity the type of jacobian to use
in the interpolation.

Before:

PostProcessing {
  { Name MagSta_a_2D ; NameOfFormulation Magnetostatics_a_2D ;
    PostQuantity {
      { Name a  ; Value { Term { [ {a} ]          ; In Domain_Mag ; } }
}
      { Name az ; Value { Term { [ CompZ[{a}] ]   ; In Domain_Mag ; } }
}
      { Name b  ; Value { Term { [ {d a} ]        ; In Domain_Mag ; } }
}
      { Name h  ; Value { Term { [ nu[] * {d a} ] ; In Domain_Mag ; } }
}
    }
  }
}

After:

PostProcessing {
  { Name MagSta_a_2D ; NameOfFormulation Magnetostatics_a_2D ;
    PostQuantity {
      { Name a  ; Value { Term { [ {a} ]          ; In Domain_Mag ;
Jacobian Vol ; } } }
      { Name az ; Value { Term { [ CompZ[{a}] ]   ; In Domain_Mag ;
Jacobian Vol ; } } }
      { Name b  ; Value { Term { [ {d a} ]        ; In Domain_Mag ;
Jacobian Vol ; } } }
      { Name h  ; Value { Term { [ nu[] * {d a} ] ; In Domain_Mag ;
Jacobian Vol ; } } }
    }
  }
}

It's a 'bug' in this version of the code: if no jacobian is given, getdp
assumes that the classical placement has to be applied (-> without the
axysymetric transformation...)



-- 
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.montefiore.ulg.ac.be/~geuzaine/