[sharonyue]

Hi guys,

In thies UEqn we have:

solve
        (
            UEqn
         ==
            fvc::reconstruct
            (
                (
                    fvc::interpolate(interface.sigmaK())*fvc::snGrad(alpha1)
                  - ghf*fvc::snGrad(rho)
                  - fvc::snGrad(p_rgh)
                ) * mesh.magSf()
            )
        );

Can I code it like this:

solve
        (
            UEqn
         ==
            
                    interface.sigmaK()*fvc::grad(alpha1)
                  - ghf*fvc::grad(rho)
                  - fvc::grad(p_rgh)
        );

Actually Im confused about the difference with the grad() and sngrad(). Any ideas?

[eysteinn]

Originally Posted by sharonyue

Hi guys,

In thies UEqn we have:

solve
        (
            UEqn
         ==
            fvc::reconstruct
            (
                (
                    fvc::interpolate(interface.sigmaK())*fvc::snGrad(alpha1)
                  - ghf*fvc::snGrad(rho)
                  - fvc::snGrad(p_rgh)
                ) * mesh.magSf()
            )
        );

Can I code it like this:

solve
        (
            UEqn
         ==
            
                    interface.sigmaK()*fvc::grad(alpha1)
                  - ghf*fvc::grad(rho)
                  - fvc::grad(p_rgh)
        );

Actually Im confused about the difference with the grad() and sngrad(). Any ideas?

grad() is the "normal" gradient calculated in the cell center while sngrad()
is the surface normal gradient that is defined on the face of the cell.
(Which explains why interface.sigmaK is interpolated from the cell center
to the face before performing the multiplication.)

fvc::reconstruct reconstructs the cell centered values from the face values.

Hope this helps

/Eysteinn

[sharonyue]

Originally Posted by eysteinn

grad() is the "normal" gradient calculated in the cell center while sngrad()
is the surface normal gradient that is defined on the face of the cell.
(Which explains why interface.sigmaK is interpolated from the cell center
to the face before performing the multiplication.)

fvc::reconstruct reconstructs the cell centered values from the face values.

Hope this helps

/Eysteinn

Hi,

I tried to revise it into this:

if (pimple.momentumPredictor())
    {
        solve
        (
            UEqn
         ==
                    interface.sigmaK()*fvc::grad(alpha1)
                  - gh*fvc::grad(rho)
                  - fvc::grad(p_rgh)               
        );
    }

Then wmake successfully, Then I made a test on dambreak, I get the totally same result(see the log). So it works, but I dont know why Weller make it interpolating U then reconstruct it, it would be easy in my way.

[eysteinn]

I have not used this solver, but my (wild) guess is that this is used to resolve some
convergence problems with the momentum equation.

When you ran the cavity case was the momentumPredictor set to yes
under pimple in the system/fvSolution?

/Eysteinn

[sharonyue]

Originally Posted by eysteinn

I have not used this solver, but my (wild) guess is that this is used to resolve some
convergence problems with the momentum equation.

When you ran the cavity case was the momentumPredictor set to yes
under pimple in the system/fvSolution?

/Eysteinn

Sorry, I did not, so the log is absolutely the same. While I tried to set it to yes, but the result is almost the same. not differ much.

[hchen]

Hi

I am also curious about it. Anybody has a comment on this issue?

[akidess]

Using cell centered values is not consistent and can lead to problems. I think you can find details in S. S. Deshpande, L. Anumolu, and M. F. Trujillo, Computational Science & Discovery 5, 014016+ (2012), ISSN 1749-4699, URL http://dx.doi.org/10.1088/1749-4699/5/1/014016. (which anyone working with interFoam should read).