[Getdp] Issue with constraint from pre-resolution

Ruth Vazquez Sabariego ruth.sabariego at kuleuven.be
Fri Feb 9 14:17:32 CET 2018

Dear Guillaume,

The answer is no, the formulation that computes the boundary condition is not called per time step, only once. AssignedFromResolution and TransferSolution where thought for doing just one computation (often expensive) to be reused further.

I think about another way of doing what you want…
Maybe the simplest way to go is just imposing it as an additional term in your formulation:

Galerkin { [ -ExtField[] , {h} ] ;
                  In Infinity ;
                  Jacobian SurfaceJacobian ;
                  Integration SimpleIntegration ; }

and then you remove the constraint. That’s what we do in high frequency for the boundary conditions at infinity.

Have a look, e.g. at the dipole case with Microwave.pro.
Galerkin { DtDof [ Sqrt[epsilon[]*nu[]] * ( Normal[] /\ Dof{e} ) /\ Normal[] , {e} ];
          In SigmaInf; Integration I1; Jacobian JSur;  }

If that does not work, then try to use a formulation that you call very time in your loop with a projection in between.

Best regards,

Prof. Ruth V. Sabariego
KU Leuven
Dept. Electrical Engineering ESAT/Electa, EnergyVille

Free software: http://gmsh.info | http://getdp.info | http://onelab.info

On 8 Feb 2018, at 16:32, DILASSER Guillaume <Guillaume.DILASSER at cea.fr<mailto:Guillaume.DILASSER at cea.fr>> wrote:

Dear Ruth,

Thank you for the quick answer, I completely missed the Group definition… To me it was obvious that, Phi being defined on the whole Domain and Infinity being a subset of it, Phi was also automatically defined on Infinity… I will now know that it is not the case.

Can I ask you another question ? I have noticed that dipole.pro and mstrip.pro are static problems, mine on the other hand is a dynamic one solved in the time domain. Therefore, I was wondering if the pre-resolution giving the boundary conditions will be called at each time step automatically or if I need to call it myself from the Resolution block. I do not think that there is an example for that in the demos are benchmarks…

Thank you again for your help,

Yours faithfully,

Guillaume Dilasser

De : Ruth Vazquez Sabariego [mailto:ruth.sabariego at kuleuven.be]
Envoyé : jeudi 8 février 2018 15:49
À : DILASSER Guillaume <Guillaume.DILASSER at cea.fr<mailto:Guillaume.DILASSER at cea.fr>>
Cc : getdp at onelab.info<mailto:getdp at onelab.info>
Objet : Re: [Getdp] Issue with constraint from pre-resolution

Dear Guillaume,

You need to include the ‘Infinity’ surface in the support of your basis function:

{ Name Hrot_Hphi_3D ; Type Form1 ;
    BasisFunction {
      { Name sn ; NameOfCoef phi ; Function BF_GradNode ;
        Support Region[{Domain, Infinity}] ; Entity NodesOf[ DomainCC ] ; }
      { Name se ; NameOfCoef he ; Function BF_Edge ;
        Support DomainC ; Entity EdgesOf[ All , Not BndDomainC ]; }
    Constraint {
      { NameOfCoef phi ;
        EntityType NodesOf ;
        NameOfConstraint ExtField ; }

Otherwise, it does not find the degrees of freedom as the research is done starting with that Support region.

If I do that I get
Info    : System 1/1: 60 Dofs
Info    : Generate[SystemBoundaryConditions]
Info    : Solve[SystemBoundaryConditions]
Info    : N: 60 - preonly lu mumps
Info    : TransferSolution[SystemBoundaryConditions]
E n d   P r e - R e s o l u t i o n  (1/1)
Info    : (Wall = 1.29631s, CPU = 0.027512s, Mem = 9.85547Mb)
E n d   P r e - P r o c e s s i n g
P r o c e s s i n g . . .
Info    : CreateDir[Results]
Info    : InitSolution[SystemMagnetics]
Info    : SaveSolution[SystemMagnetics]
Info    : TimeLoopTheta ...
Info    : Theta Time = 1e-06 s (TimeStep 1, DTime 1e-06)
Info    : GenerateJac[SystemMagnetics]
Info    : SolveJac[SystemMagnetics]
Info    : N: 1348 - preonly lu mumps
Info    :   0 Nonlinear Residual norm 0.000000000000e+00
Info    : GenerateJac[SystemMagnetics]
Info    : GetResidual[SystemMagnetics]

that does not converge, but at least the BC are there…


Prof. Ruth V. Sabariego
KU Leuven
Dept. Electrical Engineering ESAT/Electa, EnergyVille

Free software: http://gmsh.info<http://gmsh.info/> | http://getdp.info<http://getdp.info/> | http://onelab.info<http://onelab.info/>

On 8 Feb 2018, at 15:23, DILASSER Guillaume <Guillaume.DILASSER at cea.fr<mailto:Guillaume.DILASSER at cea.fr>> wrote:

Dear GetDP user,

I am writing to call for some help concerning the constraints of type AssignFromResolution.

I am using an H-Phi magnetodynamic model and I would like to control the field on the outer boundaries of the modelling domain to match a specific time-dependant profile. Because the external field I want to apply is too complex for the usual Assign type of constraint (it is the sum of two functions with different time dependences), I have been trying to use AssignFromResolution. My current model (not working) is attached with this email. The issue is, during pre-processing, I get the following warning about the system for the pre-resolution, with is apparently empty :
Info    : Generate[SystemBoundaryConditions]
Warning : Generated system is of dimension zero
Info    : Solve[SystemBoundaryConditions]
Warning : Zero-size system: skipping solve!
Unfortunately, I do not understand where those errors come from... I have tried to identify the problem by comparing my code with the benchmark examples antennas/dipole.pro and antennas/mstrip.pro that also use AssignFromResolution constraints but I can’t seem to find it… Thank you in advance for your help !

Yours faithfully,

Guillaume DILASSER
Ingénieur de recherche / Research engineer

CEA Saclay
Bât.123 – P. 213c
Point courrier n°11
91191 Gif-sur-Yvette Cedex

guillaume.dilasser at cea.fr<mailto:guillaume.dilasser at cea.fr>

getdp mailing list
getdp at onelab.info<mailto:getdp at onelab.info>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://onelab.info/pipermail/getdp/attachments/20180209/f94ee8b1/attachment-0001.html>

More information about the getdp mailing list