[santiagomarquezd]

Hi FOAMers, I would like to do a question about the A() method in fvMatrix class. From the theory we know it represents the values of matrix diagonal coefficients, which, I suppose, have the contribution of BC's. Usually, before linear systems are solved BC contributions are added to diagonal coeffs. via addBoundaryDiag method so you can recover the same matrix as you can assemble by hand. Nevertheless is you take this coeffs. they will be different of those you can obtain by the A() method. Checking the code in fvMatrix.C we have, for A()

00638 template<class Type>
00639 Foam::tmp<Foam::volScalarField> Foam::fvMatrix<Type>::A() const
00640 {
00641     tmp<volScalarField> tAphi
00642     (
00643         new volScalarField
00644         (
00645             IOobject
00646             (
00647                 "A("+psi_.name()+')',
00648                 psi_.instance(),
00649                 psi_.mesh(),
00650                 IOobject::NO_READ,
00651                 IOobject::NO_WRITE
00652             ),
00653             psi_.mesh(),
00654             dimensions_/psi_.dimensions()/dimVol,
00655             zeroGradientFvPatchScalarField::typeName
00656         )
00657     );
00658 
00659     tAphi().internalField() = D()/psi_.mesh().V();
00660     tAphi().correctBoundaryConditions();
00661 
00662     return tAphi;
00663 }

D() method is called, which is:

00605 template<class Type>
00606 Foam::tmp<Foam::scalarField> Foam::fvMatrix<Type>::D() const
00607 {
00608     tmp<scalarField> tdiag(new scalarField(diag()));
00609     addCmptAvBoundaryDiag(tdiag());
00610     return tdiag;
00611 }

BC's contribution is done via addCmptAvBoundaryDiag:

00134 template<class Type>
00135 void Foam::fvMatrix<Type>::addCmptAvBoundaryDiag(scalarField& diag) const
00136 {
00137     forAll(internalCoeffs_, patchI)
00138     {
00139         addToInternalField
00140         (
00141             lduAddr().patchAddr(patchI),
00142             cmptAv(internalCoeffs_[patchI]),
00143             diag
00144         );
00145     }
00146 }

which calls cmptAv. This method assigns the average of vector contribution components as the number that is added to diagonal coefficients.

My questions are:

1. Why "true" BC influenced diagonal coeffs aren't used?
2. Why the contributions are calculated this way instead?

All help will be appreciated. Regards.

[santiagomarquezd]

Any clues??

[hua1015]

maybe because the internalCoeff are (1,1,1) or (0,0,0)，and the Ap is the same for U,V,W equation,
so using cmptAv is reasonable.

[simrego]

Hi!


Facing the same problem (same in the H() operator).
Anyone knows it?
Thanks!


Regő