CFD Online Discussion Forums - fvc::interpolate -> harmonic interpolation?

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)

- OpenFOAM Running, Solving & CFD (https://www.cfd-online.com/Forums/openfoam-solving/)

- - fvc::interpolate -> harmonic interpolation? (https://www.cfd-online.com/Forums/openfoam-solving/77862-fvc-interpolate-harmonic-interpolation.html)

fvc::interpolate -> harmonic interpolation?

Dear all,

I would like to know how OF interpolates physical properties onto faces. I know there is linear interpolation and I have seen some threads (Interpolation in OpenFOAM and About interFoam solver) on related issues, but not quite what I am looking for.
My questions are the following:

How does OF interpolate physical properties (e.g. viscosity in interFoam) onto faces and where is it defined?
Is there a harmonic interpolation scheme for physical properties (can fvc::interpolate be set to harmonic somehow)?

Many thanks in advance!

Hi Gijsbert,

@1: I think it is always done using fvc::interpolate() (e.g. for viscosity in interFoam you can find it in src/transportModels/incompressible/incompressibleTwoPhaseMixture/twoPhaseMixture.C; member funtion mu())

@2: fvc::interpolate should read its schemes from fvSchemes-Dictionary. Therefore you should be able to apply what you think to be the best for your purpose.

Regards,
Stefan

Hi Stefan,

Thanks for your reply!

Quote:

@1: I think it is always done using fvc::interpolate() (e.g. for viscosity in interFoam you can find it in src/transportModels/incompressible/incompressibleTwoPhaseMixture/twoPhaseMixture.C; member funtion mu())

In the member functions (for e.g. muf) in src/transportModels/incompressible/incompressibleTwoPhaseMixture/twoPhaseMixture.C fvc::interpolate is also used, but I can' find anywhere how things are interpolated.

As you write, I also thought of changing schemes in fvSchemes:

Quote:

@2: fvc::interpolate should read its schemes from fvSchemes-Dictionary. Therefore you should be able to apply what you think to be the best for your purpose.

But when I write

Code:

interpolationSchemes

{

    default         banana;//linear;

}

I get a list of (what seem to be) advection schemes (not face interpolation schemes (?)):

Code:

Interpolation dsicretization schemes:



Unknown discretisation scheme banana



Valid schemes are :



56

(

linear

quadraticUpwindFit

limitedCubic

limitedLinear

downwind

OSPREV

Gamma

localMax

quadraticLinearUpwindFit

outletStabilised

quadraticFit

MUSCLV

reverseLinear

GammaV

blended

UMISTV

limitWith

linearPureUpwindFit

vanLeer

vanLeerV

linearUpwind

Minmod

filteredLinear2V

MinmodV

Phi

cubic

SuperBee

filteredLinear3

biLinearFit

filteredLinear3V

quadraticLinearFit

QUICK

UMIST

limitedLinearV

SuperBeeV

weighted

skewCorrected

midPoint

vanAlbadaV

linearFit

linearUpwindV

localMin

fixedBlended

SFCDV

vanAlbada

OSPRE

MUSCL

clippedLinear

filteredLinear2

upwind

limitedCubicV

localBlended

SFCD

cubicUpwindFit

filteredLinear

QUICKV

)

In the mean time I did find a harmonic interpolation scheme in src/finiteVolume/interpolation/surfaceInterpolation/schemes/harmonic/harmonic.H. But when I plug it into the UEqn as a test (in interFoam), like below, the compiler complains that harmonic is a virtual function so it needs an object.

Code:

    

surfaceScalarField muEff

(

    "muEff",

    twoPhaseProperties.muf()

    + harmonic::interpolate(rho*turbulence->nut())

//       + fvc::interpolate(rho*turbulence->nut())

);

Would anyone have an idea about this? Am I doing something nonsensible syntax-wise? Thanks in advance!

Hi Gijsbert,

the problem is, that the interpolationScheme defined in fvSchemes in used for every call of fvc::interpolate (and this are 12 in interFoam!). Some of this interpolations might not fit to the harmonic scheme. This should cause the errors and should also be the reason, why it is not listed in your banana-Test.

Nevertheless you won't have to change any piece of code I think. Instead you can specify the harmonic scheme to single operations e.g. by the following entry in fvSchemes:

Code:

interpolationSchemes

(

     default                      linear;

     interpolate(nu1)          harmonic;

     interpolate(nu2)          harmonic;

     interpolate((rho*nut))  harmonic;

)

You can test if these entries are used, by replacing harmonic by banana again.

Have a lot of fun,
Stefan

Hello Gjis!!!

hope you're fine.

If you don't want this to be accessed from anywhere else you can hardcode that one as well!

tmp< GeometricField< Type, fvPatchField, volMesh > > interpolate ( const GeometricField< Type, fvPatchField, volMesh > &

const surfaceScalarField & tvf,

const word & name
)
Name than would be the name of your interpolation scheme.

Best!

Kathrin

@ Stefan:

Quote:

Some of this interpolations might not fit to the harmonic scheme

Aha, good point!
Thanks for the help, I implemented it and it works.

@ Kathrin:
Hi Kathrin, I'm fine, thanks :). Hope you're well too!

Many thanks for the help. Hardcoding would perhaps be good at some point. But I am not quite clear on the snippet, so I have some (possibly dumb) questions ... For a harmonic scheme myHarmonicScheme I suppose the code goes like this:

Code:

tmp<  GeometricField<  Type,  fvPatchField,  volMesh  > > interpolate                      

(                      

const GeometricField< Type,  fvPatchField, volMesh >& const surfaceScalarField &tvf,

          const word &myHarmonicScheme

)

Isn't "tvf" a tensorVolumeField (just a guess)? Is that ok for viscosity? Also, should I then put this code into $FOAM_SRC//finiteVolume/finiteVolume/fvc/? And shouldn't there be an actual implementation of the interpolation algorithm?

Gijsbert, the snippet Kathrin showed you is already part of fvc (see openfoam documentation). You should be able to hard call a harmonic interpolation by using:

Code:

surfaceScalarField some_ssf = fvc::interpolate(some_vsf, "harmonic");

- Anton

EDIT: I just tested this, unfortunately it doesn't work. This snippet tries to find a "harmonic" entry in fvSchemes instead, which has little advantage over directly specifying the term in fvSchemes.

Hello to all. Perhaps someone can quickly enlighten on this topic, I need to interpolate cell centered values to cell faces. The cell centered faces needed to interpolate to cell faces are near the boundary, and not really quite sure how to code it. How is this:

const fvPatchScalarField& Tw = thermo.T().boundaryField()[patchI];
const scalarField Tadj = Tw.patchInternalField();//gives T for cell adjacent to wall
surfaceScalarField Tsurf=fvc::interpolate(Tadj);

This code makes sense, but I'm not sure if it is correct. Furthermore, does the interpolation computes face values for the east, west, north, and southern faces? I need the east, west, and northern face values, so how do get those from the interpolation operation?:confused:

Cheers,
Deji

face value of a parameter

Hi FOAMERS
I am trying to solve a set of equations in the following form
--------
(Density)*fvm::ddt(U)
+ (Density/gL)*fvm::div(phi, U)
- (Visc)*(fvm::laplacian(U))
- GravityVector*g*Density*((BetaT*(T-TNot))+BetaC*(C-CNot))
+ fvm::Sp(gL*Visc/Perm,U)
--------
Perm is the permeability which is calculated by the following relation

Perm[celli] = (pow(SDAS.value(),2))*pow((gL[celli]),3)/(180*(pow((1-gL[celli]),2)));

Is there any way to make sure that the permeability at the cell face is calculated by mean harmonic interpolation?