CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM Bugs (http://www.cfd-online.com/Forums/openfoam-bugs/)
-   -   Prossible wrong in the LimitedScheme.C ? (http://www.cfd-online.com/Forums/openfoam-bugs/67685-prossible-wrong-limitedscheme-c.html)

liuhuafei August 24, 2009 01:38

Prossible wrong in the LimitedScheme.C ?
 
Prossible wrong in the LimitedScheme.C
In LimitedScheme.C
template<class Type, class Limiter, template<class> class LimitFunc>
tmp<surfaceScalarField> LimitedScheme<Type, Limiter, LimitFunc>::limiter
(
const GeometricField<Type, fvPatchField, volMesh>& phi
) const
{
..... for internal faces .................
forAll(pLim, face)
{
label own = owner[face];
label nei = neighbour[face];
pLim[face] = Limiter::limiter
(
CDweights[face],
this->faceFlux_[face],
lPhi[own],
lPhi[nei],
gradc[own],
gradc[nei],
C[nei] - C[own]
);
}
question 1: why the distance vector is C[nei]-C[own],
no correction for the nonorthgonal grid?

-----------for the boundary----------------------
GeometricField<typename Limiter::gradPhiType, fvPatchField, volMesh>
gradc(fvc::grad(lPhi));
forAll(bLim, patchi)
{
scalarField& pLim = bLim[patchi];
if (bLim[patchi].coupled())
{
Field<typename Limiter::phiType> plPhiN =
lPhi.boundaryField()[patchi].patchNeighbourField();
Field<typename Limiter::gradPhiType> pGradcN =
gradc.boundaryField()[patchi].patchNeighbourField();
// Build the d-vectors
vectorField pd =
mesh.Sf().boundaryField()[patchi]
/(
mesh.magSf().boundaryField()[patchi]
*mesh.deltaCoeffs().boundaryField()[patchi]
);
if (!mesh.orthogonal())
{
pd -= mesh.correctionVectors().boundaryField()[patchi]
/mesh.deltaCoeffs().boundaryField()[patchi];
}
forAll(pLim, face)
{
pLim[face] = Limiter::limiter
(
pCDweights[face],
pFaceFlux[face],
plPhiP[face],
plPhiN[face],
pGradcP[face],
pGradcN[face],
pd[face]
);
}
}
else
{
pLim = 1.0;
}
}
question 2:
The pGradcN[face] is neighbour cell's gradient. For the coupled boundary, for examples,the neighbour cell is virtual N',not the real cell?
The virtual N' is got by rotated by N? How can we get the gradient of N'? The gradient of N' is equal to that of N?

liuhuafei August 24, 2009 19:53

For example

For the cycle boundary,We have a pair of cell, P and N. We need rotate the N to N' by transformation. On calculation of the limiter function,we need the gradient of N' cell. How can we get it. The gradient function of openFoam doesn't treat the coupled boundary.

But the gradient of N' is equal to that of N?
We need the transformation of the gradient of N to get the gradient of N' like other vector variable?



template<class Type> void gaussGrad<Type>::correctBoundaryConditions
( const GeometricField<Type, fvPatchField, volMesh>& vsf,
GeometricField <typename outerProduct<vector, Type>::type, fvPatchField, volMesh >& gGrad){
forAll(vsf.boundaryField(), patchi)
{
if (!vsf.boundaryField()[patchi].coupled())
{
vectorField n =
vsf.mesh().Sf().boundaryField()[patchi]
/vsf.mesh().magSf().boundaryField()[patchi];
gGrad.boundaryField()[patchi] += n *
(
vsf.boundaryField()[patchi].snGrad()
- (n & gGrad.boundaryField()[patchi])
);
}
}
}

liuhuafei August 24, 2009 19:54

http://www.cfd-online.com/Forums/c:\Clip_12.jpg

liuhuafei August 24, 2009 19:56

Quote:

[IMG]file:///c:/Clip_12.jpg[/IMG]
the figure of cycle boundary

ybapat July 22, 2014 06:41

Hello Liu,

This is very old post from you, but if you can tell if gradients near cyclic AMI boundaries are calculated correctly?

Tushar@cfd July 23, 2014 05:05

Quote:

Originally Posted by liuhuafei (Post 227232)
Prossible wrong in the LimitedScheme.C
In LimitedScheme.C
template<class Type, class Limiter, template<class> class LimitFunc>
tmp<surfaceScalarField> LimitedScheme<Type, Limiter, LimitFunc>::limiter
(
const GeometricField<Type, fvPatchField, volMesh>& phi
) const
{
..... for internal faces .................
forAll(pLim, face)
{
label own = owner[face];
label nei = neighbour[face];
pLim[face] = Limiter::limiter
(
CDweights[face],
this->faceFlux_[face],
lPhi[own],
lPhi[nei],
gradc[own],
gradc[nei],
C[nei] - C[own]
);
}
question 1: why the distance vector is C[nei]-C[own],
no correction for the nonorthgonal grid?

-----------for the boundary----------------------
GeometricField<typename Limiter::gradPhiType, fvPatchField, volMesh>
gradc(fvc::grad(lPhi));
forAll(bLim, patchi)
{
scalarField& pLim = bLim[patchi];
if (bLim[patchi].coupled())
{
Field<typename Limiter::phiType> plPhiN =
lPhi.boundaryField()[patchi].patchNeighbourField();
Field<typename Limiter::gradPhiType> pGradcN =
gradc.boundaryField()[patchi].patchNeighbourField();
// Build the d-vectors
vectorField pd =
mesh.Sf().boundaryField()[patchi]
/(
mesh.magSf().boundaryField()[patchi]
*mesh.deltaCoeffs().boundaryField()[patchi]
);
if (!mesh.orthogonal())
{
pd -= mesh.correctionVectors().boundaryField()[patchi]
/mesh.deltaCoeffs().boundaryField()[patchi];
}
forAll(pLim, face)
{
pLim[face] = Limiter::limiter
(
pCDweights[face],
pFaceFlux[face],
plPhiP[face],
plPhiN[face],
pGradcP[face],
pGradcN[face],
pd[face]
);
}
}
else
{
pLim = 1.0;
}
}
question 2:
The pGradcN[face] is neighbour cell's gradient. For the coupled boundary, for examples,the neighbour cell is virtual N',not the real cell?
The virtual N' is got by rotated by N? How can we get the gradient of N'? The gradient of N' is equal to that of N?


......

LimitedScheme is actually correct, Prof. H. Jasak used different technique.

Have you seen Prof. H. Jasak PhD thesis?

It is recommended to check his explanation for these face variable calculation. You will get answer to all your queries.

In your case may be this boundary condition could be problematic but not Limitedscheme.

Anyways..
-
Best Luck!

ybapat July 23, 2014 05:46

Hello Tushar,

Can you please elaborate which section in thesis you are talking about? Are you talking about interpolation schemes?

To give you more background on my question, I am trying pressure jump BC(This is derived from cyclicAMI). I am not able to converge simulation if I do not use faceLimited gradients. Moment I switch on cellLimited gradient simulation diverges,. Even with faceLimited if I change form "faceLimited Gauss linear 1.0" to "faceLimited Gauss linear 0.99" simulation diverges.

Regards,
-Yogesh

Tushar@cfd July 23, 2014 06:55

Hello Yogesh,

I don't understand why you are writing your problem case to me.

I answered the previous post of "liuhuafei", as he specifically pointed "Possible wrong in the LimitedScheme.C ?". I clarified that post as there isn't any mistake with the code. (reference for same can be understood by reading the PhD thesis of Prof. H. Jasak)

About your case, you didn't mention any detail so it's hard to guess how come your case failed.

-
Best Luck!

ybapat July 23, 2014 07:08

Hello Tushar,

I thought you are replying to my post, as liuhuafei's post is very old.

-Yogesh

Tushar@cfd July 23, 2014 07:22

Hello Yogesh,

If you wish to share your problem case then you can post here. I will try to see if I can help you.

-
Best Regards!


All times are GMT -4. The time now is 17:49.