[martel]

Hi all,

I am trying to understand how openfoam generates the matrices for the differential operators.

I have programmed a simple example for the laplacian operator:

// Define a matrix that contains the coefficients of the Laplacian

    fvScalarMatrix TEqnLaplaciano 
	  ( 
	   fvm::laplacian(DT, T) 
	    ); 

// Extract source term, matrix diagonal, upper and lower nonzero elements and their indices

    scalarField& source_  = TEqnLaplaciano.source();
    scalarField& lower_   = TEqnLaplaciano.lower();
    scalarField& upper_   = TEqnLaplaciano.upper();
    scalarField& diag_    = TEqnLaplaciano.diag();

    const unallocLabelList& l_ = TEqnLaplaciano.lduAddr().lowerAddr();
    const unallocLabelList& u_ = TEqnLaplaciano.lduAddr().upperAddr();

    Info<< "\n diagvalues:  \n " << diag_  << endl;
    Info<< "\n lowervalues: \n " << lower_ << endl;
    Info<< "\n lowerAddr:   \n " << l_     << endl;
    Info<< "\n Uppervalues: \n " << upper_ << endl;
    Info<< "\n UpperAddr:   \n " << u_     << endl;

    Info<< "\n sourcevalues:  \n " << source_  << endl;

And it gives:
* Matrix Diagonal terms
* Matrix out of diagonal terms
* Indices of the out of diagonal terms
* source terms (right hand side of the linear system) (zero in my example)

But the effect of the boundary conditions is not yet included in the matrix and source term that I get.

My question is:

How can I get the contribution (coefficients and indices) of the Boundary Conditions to the matrix coefficients (diag and out-of-diag) and to the source term?


Thanks in advance for your help.

--CME

[gigilentini8]

same question here. any news on this?

[martel]

Originally Posted by gigilentini8

same question here. any news on this?

Hi gigilentin8i,

Not much on this unfortunaltely:

I have use these two extra lines to get the information from the boundary conditions:

 Info <<"\n  TEqnLaplaciano.boundaryCoeffs()     \n " <<       TEqnLaplaciano.boundaryCoeffs() << endl;
    Info <<"\n  TEqnLaplaciano.internalCoeffs()     \n " <<       TEqnLaplaciano.internalCoeffs() << endl;

But I have realized that OpenFoam does not construct the complete matrix of the laplacian because the non-orhogonal contributions to the operator are ALWAYS computed explicitely. So now I am trying to export the mesh coefficients to build the matrices myself in MATLAB.

--Carlos

[gigilentini8]

Thanks Carlos,

this explains why I am getting incorrect results for the eigenvalues/functions of the laplacian when the mesh is non cartesian.
does this mean that the additional non-orthogonal contribution are in TEqn.source()? If this is the case I could just try to put them back in the matrix.
or how about using an uncorrected or fourth scheme?

I was able to add the boundary condition contribution to the matrix with this code

    forAll(TEqn.internalCoeffs(), patchI)
       {
        const label*  fPtr = TEqn.lduAddr().patchAddr(patchI).begin();
        Field<double> intF(TEqn.internalCoeffs()[patchI].component(0));
        forAll(TEqn.lduAddr().patchAddr(patchI), faceI)
          {
              label fCell = fPtr[faceI];
              diag(fCell) -= intF[faceI];
          }
       }

Originally Posted by martel

Hi gigilentin8i,

Not much on this unfortunaltely:

I have use these two extra lines to get the information from the boundary conditions:

 Info <<"\n  TEqnLaplaciano.boundaryCoeffs()     \n " <<       TEqnLaplaciano.boundaryCoeffs() << endl;
    Info <<"\n  TEqnLaplaciano.internalCoeffs()     \n " <<       TEqnLaplaciano.internalCoeffs() << endl;

But I have realized that OpenFoam does not construct the complete matrix of the laplacian because the non-orhogonal contributions to the operator are ALWAYS computed explicitely. So now I am trying to export the mesh coefficients to build the matrices myself in MATLAB.

--Carlos