[randolph]

Originally Posted by bigphil

Hi Rdf,

It should be possible to do this by looking at the "solveSegregated" function in the fvMatrix class (see fvMatrixSolve.C).

The "solveSegregated" function forms the linear system for each direction before solving it.

fvMatrix stores the matrix without the boundary condition contributions because the boundary condition contribution may differ for each direction.
Before solving each direction, fvMatrix makes a copy of the matrix diagonal and modifies it with the boundary conditions for that direction before solving it; you could use similar code to prepare the matrix for each direction. For example, see how the addBoundaryDiag function is used inside the solveSegregated function:

   for (direction cmpt=0; cmpt<Type::nComponents; cmpt++)
    {
        if (validComponents[cmpt] == -1) continue;

	// copy field and source                                                                                                                                

        scalarField psiCmpt(psi.primitiveField().component(cmpt));
        addBoundaryDiag(diag(), cmpt);

        scalarField sourceCmpt(source.component(cmpt));

        // ...

Philip

Philip,

Just a step back, would there any high-level controls that allow the fvVectorMatrix to solve for specific components? For example, I want to solve Ux, Uy, but not Uz.

I look into the code and it looks like I have to write a new class for this purpose. What is your opinion?

Thanks,
Rdf

[bigphil]

Hi Rdf,

One way to control which components are solved by fvVectorMatrix is to manipulate the "empty" directions. For example, by default, fvVectorMatrix will not solve in an empty direction (like for a 2-D case).

In the code, you could hack the empty directions, e.g.

Vector<label>& solD = const_cast<Vector<label>&>(mesh.solutionD());

// Enable solution in the X direction
solD[vector::X] = 1;

// Enable solution in the Y direction
solD[vector::Y] = 1;

// Disable solution in the Z direction
solD[vector::Z] = -1;

This is a hack: it may work for you, but it may also have unintended consequences.