[ywang]

Hi, everyone.

I want to calculate the gradients of a volScalarField parameter F in the boundary cells. Now, I am using fvc::grad(F,gradF) to calculate the gradients in all cells, and then look up the values in the boundary cells.

Is there any other more efficient method? I just want to reduce the time used for calculating.

Thanks.

Yong

[stevenvanharen]

Maybe take a look at the wallGradU utility?

Is that helpful for you?

[ywang]

Originally Posted by stevenvanharen

Maybe take a look at the wallGradU utility?

Is that helpful for you?

Thanks.

wallGradU is using
wallGradU.boundaryField()[patchi] =-U.boundaryField()[patchi].snGrad();.

Based on the Programmers Guide, snGrad still calculate all cells in the computational region. Right?

I am also reading the code about grad and check if it's helpful.

[stevenvanharen]

because u use:

U.boundaryField()[patchi]

only the faces on this patch are calculated

[eugene]

snGrad in this context only calculates the wall normal gradient. It does so in a completely local and efficient way though. If you want to calculate the cell centred gradient for a subset of cells, I suggest you look at the code in finiteVolume/lnInclude/gaussGrad.C. This will show you the steps to calculate a gradient on a cell-by-cell basis.

[stevenvanharen]

Originally Posted by eugene

snGrad in this context only calculates the wall normal gradient. It does so in a completely local and efficient way though. If you want to calculate the cell centred gradient for a subset of cells, I suggest you look at the code in finiteVolume/lnInclude/gaussGrad.C. This will show you the steps to calculate a gradient on a cell-by-cell basis.

Ah ok, I missed the fact that cell-center gradients were required.

[ywang]

Originally Posted by eugene

snGrad in this context only calculates the wall normal gradient. It does so in a completely local and efficient way though. If you want to calculate the cell centred gradient for a subset of cells, I suggest you look at the code in finiteVolume/lnInclude/gaussGrad.C. This will show you the steps to calculate a gradient on a cell-by-cell basis.

Yes, I have read the code of gaussGrad and found there was the very thing I wanted. Thanks.

[ywang]

Originally Posted by stevenvanharen

Ah ok, I missed the fact that cell-center gradients were required.

It's ok. I noticed it. Thanks very much .

[ywang]

Hi, everyone. As I want to calculate the gradient at the cells close to the boundaries. I checked the code in the gaussGrad.C:

PHP Code:

    const unallocLabelList& owner = mesh.owner();
    const unallocLabelList& neighbour = mesh.neighbour();
    const vectorField& Sf = mesh.Sf();

    Field<GradType>& igGrad = gGrad;
    const Field<Type>& issf = ssf;

   
    forAll(owner, facei)
    {
        GradType Sfssf = Sf[facei]*issf[facei];

        igGrad[owner[facei]] += Sfssf;
        igGrad[neighbour[facei]] -= Sfssf;
    }
    
    forAll(mesh.boundary(), patchi)
    {
        const unallocLabelList& pFaceCells = mesh.boundary()[patchi].faceCells();
        const vectorField& pSf = mesh.Sf().boundaryField()[patchi];
        const fvsPatchField<Type>& pssf = ssf.boundaryField()[patchi];

        forAll(mesh.boundary()[patchi], facei)
        {     
            igGrad[pFaceCells[facei]] += pSf[facei]*pssf[facei];      
        }      
    } 


However, in the first loop, all internal faces are calculated. So, I changed the code as:

PHP Code:

    const unallocLabelList& owner = mesh.owner();
    const unallocLabelList& neighbour = mesh.neighbour();
    const vectorField& Sf = mesh.Sf();

    Field<GradType>& igGrad = gGrad;
    const Field<Type>& issf = ssf;


    forAll(mesh.boundary(), patchi)
    {
        const unallocLabelList& pFaceCells = mesh.boundary()[patchi].faceCells();
        const vectorField& pSf = mesh.Sf().boundaryField()[patchi];
        const fvsPatchField<Type>& pssf = ssf.boundaryField()[patchi];

        forAll(pFaceCells, facei)
        {
            if(!mesh.isInternalFace(facei))
            {   
                igGrad[pFaceCells[facei]] += pSf[facei]*pssf[facei];      
            }
            else
            {
                GradType Sfssf = Sf[facei]*issf[facei];
                igGrad[owner[facei]] += Sfssf;
                igGrad[neighbour[facei]] -= Sfssf;     
            }
        }      
    } 


There is no error when I compile it. But the solver will lead to unphysical results. I know there must be something wrong with pFaceCells, but have no idea to revise it. Any message from you are welcome.

Thanks in advance.

Yong