mirror of
https://develop.openfoam.com/Development/openfoam.git
synced 2025-11-28 03:28:01 +00:00
To be used instead of zeroGradientFvPatchField for temporary fields for which zero-gradient extrapolation is use to evaluate the boundary field but avoiding fields derived from temporary field using field algebra inheriting the zeroGradient boundary condition by the reuse of the temporary field storage. zeroGradientFvPatchField should not be used as the default patch field for any temporary fields and should be avoided for non-temporary fields except where it is clearly appropriate; extrapolatedCalculatedFvPatchField and calculatedFvPatchField are generally more suitable defaults depending on the manner in which the boundary values are specified or evaluated. The entire OpenFOAM-dev code-base has been updated following the above recommendations. Henry G. Weller CFD Direct
380 lines
11 KiB
C
380 lines
11 KiB
C
/*---------------------------------------------------------------------------*\
|
|
========= |
|
|
\\ / F ield | OpenFOAM: The Open Source CFD Toolbox
|
|
\\ / O peration |
|
|
\\ / A nd | Copyright (C) 2013-2016 OpenFOAM Foundation
|
|
\\/ M anipulation |
|
|
-------------------------------------------------------------------------------
|
|
License
|
|
This file is part of OpenFOAM.
|
|
|
|
OpenFOAM is free software: you can redistribute it and/or modify it
|
|
under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation, either version 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
OpenFOAM is distributed in the hope that it will be useful, but WITHOUT
|
|
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
|
|
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
|
|
for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with OpenFOAM. If not, see <http://www.gnu.org/licenses/>.
|
|
|
|
\*---------------------------------------------------------------------------*/
|
|
|
|
#include "Implicit.H"
|
|
#include "fixedValueFvsPatchField.H"
|
|
#include "fvmDdt.H"
|
|
#include "fvmDiv.H"
|
|
#include "fvmLaplacian.H"
|
|
#include "fvcReconstruct.H"
|
|
#include "volPointInterpolation.H"
|
|
#include "zeroGradientFvPatchFields.H"
|
|
|
|
// * * * * * * * * * * * * * * * * Constructors * * * * * * * * * * * * * * //
|
|
|
|
template<class CloudType>
|
|
Foam::PackingModels::Implicit<CloudType>::Implicit
|
|
(
|
|
const dictionary& dict,
|
|
CloudType& owner
|
|
)
|
|
:
|
|
PackingModel<CloudType>(dict, owner, typeName),
|
|
alpha_
|
|
(
|
|
IOobject
|
|
(
|
|
this->owner().name() + ":alpha",
|
|
this->owner().db().time().timeName(),
|
|
this->owner().mesh(),
|
|
IOobject::NO_READ,
|
|
IOobject::NO_WRITE
|
|
),
|
|
this->owner().mesh(),
|
|
dimensionedScalar("zero", dimless, 0.0),
|
|
zeroGradientFvPatchScalarField::typeName
|
|
),
|
|
phiCorrect_(NULL),
|
|
uCorrect_(NULL),
|
|
applyLimiting_(this->coeffDict().lookup("applyLimiting")),
|
|
applyGravity_(this->coeffDict().lookup("applyGravity")),
|
|
alphaMin_(readScalar(this->coeffDict().lookup("alphaMin"))),
|
|
rhoMin_(readScalar(this->coeffDict().lookup("rhoMin")))
|
|
{
|
|
alpha_ = this->owner().theta();
|
|
alpha_.oldTime();
|
|
}
|
|
|
|
|
|
template<class CloudType>
|
|
Foam::PackingModels::Implicit<CloudType>::Implicit
|
|
(
|
|
const Implicit<CloudType>& cm
|
|
)
|
|
:
|
|
PackingModel<CloudType>(cm),
|
|
alpha_(cm.alpha_),
|
|
phiCorrect_(cm.phiCorrect_()),
|
|
uCorrect_(cm.uCorrect_()),
|
|
applyLimiting_(cm.applyLimiting_),
|
|
applyGravity_(cm.applyGravity_),
|
|
alphaMin_(cm.alphaMin_),
|
|
rhoMin_(cm.rhoMin_)
|
|
{
|
|
alpha_.oldTime();
|
|
}
|
|
|
|
|
|
// * * * * * * * * * * * * * * * * Destructor * * * * * * * * * * * * * * * //
|
|
|
|
template<class CloudType>
|
|
Foam::PackingModels::Implicit<CloudType>::~Implicit()
|
|
{}
|
|
|
|
|
|
// * * * * * * * * * * * * * * * Member Functions * * * * * * * * * * * * * //
|
|
|
|
template<class CloudType>
|
|
void Foam::PackingModels::Implicit<CloudType>::cacheFields(const bool store)
|
|
{
|
|
PackingModel<CloudType>::cacheFields(store);
|
|
|
|
if (store)
|
|
{
|
|
const fvMesh& mesh = this->owner().mesh();
|
|
const dimensionedScalar deltaT = this->owner().db().time().deltaT();
|
|
const word& cloudName = this->owner().name();
|
|
|
|
const dimensionedVector& g = this->owner().g();
|
|
const volScalarField& rhoc = this->owner().rho();
|
|
|
|
const AveragingMethod<scalar>& rhoAverage =
|
|
mesh.lookupObject<AveragingMethod<scalar>>
|
|
(
|
|
cloudName + ":rhoAverage"
|
|
);
|
|
const AveragingMethod<vector>& uAverage =
|
|
mesh.lookupObject<AveragingMethod<vector> >
|
|
(
|
|
cloudName + ":uAverage"
|
|
);
|
|
const AveragingMethod<scalar>& uSqrAverage =
|
|
mesh.lookupObject<AveragingMethod<scalar>>
|
|
(
|
|
cloudName + ":uSqrAverage"
|
|
);
|
|
|
|
mesh.setFluxRequired(alpha_.name());
|
|
|
|
// Property fields
|
|
// ~~~~~~~~~~~~~~~
|
|
|
|
// volume fraction field
|
|
alpha_ = max(this->owner().theta(), alphaMin_);
|
|
alpha_.correctBoundaryConditions();
|
|
|
|
// average density
|
|
volScalarField rho
|
|
(
|
|
IOobject
|
|
(
|
|
cloudName + ":rho",
|
|
this->owner().db().time().timeName(),
|
|
mesh,
|
|
IOobject::NO_READ,
|
|
IOobject::NO_WRITE
|
|
),
|
|
mesh,
|
|
dimensionedScalar("zero", dimDensity, 0),
|
|
zeroGradientFvPatchField<scalar>::typeName
|
|
);
|
|
rho.internalField() = max(rhoAverage.internalField(), rhoMin_);
|
|
rho.correctBoundaryConditions();
|
|
|
|
// Stress field
|
|
// ~~~~~~~~~~~~
|
|
|
|
// stress derivative wrt volume fraction
|
|
volScalarField tauPrime
|
|
(
|
|
IOobject
|
|
(
|
|
cloudName + ":tauPrime",
|
|
this->owner().db().time().timeName(),
|
|
mesh,
|
|
IOobject::NO_READ,
|
|
IOobject::NO_WRITE
|
|
),
|
|
mesh,
|
|
dimensionedScalar("zero", dimPressure, 0),
|
|
zeroGradientFvPatchField<scalar>::typeName
|
|
);
|
|
|
|
tauPrime.internalField() =
|
|
this->particleStressModel_->dTaudTheta
|
|
(
|
|
alpha_.internalField(),
|
|
rho.internalField(),
|
|
uSqrAverage.internalField()
|
|
)();
|
|
|
|
tauPrime.correctBoundaryConditions();
|
|
|
|
|
|
// Gravity flux
|
|
// ~~~~~~~~~~~~
|
|
|
|
tmp<surfaceScalarField> phiGByA;
|
|
|
|
if (applyGravity_)
|
|
(
|
|
phiGByA = tmp<surfaceScalarField>
|
|
(
|
|
new surfaceScalarField
|
|
(
|
|
"phiGByA",
|
|
deltaT*(g & mesh.Sf())*fvc::interpolate(1.0 - rhoc/rho)
|
|
)
|
|
)
|
|
);
|
|
|
|
|
|
// Implicit solution for the volume fraction
|
|
// ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
surfaceScalarField
|
|
tauPrimeByRhoAf
|
|
(
|
|
"tauPrimeByRhoAf",
|
|
fvc::interpolate(deltaT*tauPrime/rho)
|
|
);
|
|
|
|
fvScalarMatrix alphaEqn
|
|
(
|
|
fvm::ddt(alpha_)
|
|
- fvc::ddt(alpha_)
|
|
- fvm::laplacian(tauPrimeByRhoAf, alpha_)
|
|
);
|
|
|
|
if (applyGravity_)
|
|
{
|
|
alphaEqn += fvm::div(phiGByA(), alpha_);
|
|
}
|
|
|
|
alphaEqn.solve();
|
|
|
|
|
|
// Generate correction fields
|
|
// ~~~~~~~~~~~~~~~~~
|
|
|
|
// correction volumetric flux
|
|
phiCorrect_ = tmp<surfaceScalarField>
|
|
(
|
|
new surfaceScalarField
|
|
(
|
|
cloudName + ":phiCorrect",
|
|
alphaEqn.flux()/fvc::interpolate(alpha_)
|
|
)
|
|
);
|
|
|
|
// limit the correction flux
|
|
if (applyLimiting_)
|
|
{
|
|
volVectorField U
|
|
(
|
|
IOobject
|
|
(
|
|
cloudName + ":U",
|
|
this->owner().db().time().timeName(),
|
|
mesh,
|
|
IOobject::NO_READ,
|
|
IOobject::NO_WRITE
|
|
),
|
|
mesh,
|
|
dimensionedVector("zero", dimVelocity, vector::zero),
|
|
fixedValueFvPatchField<vector>::typeName
|
|
);
|
|
U.internalField() = uAverage.internalField();
|
|
U.correctBoundaryConditions();
|
|
|
|
surfaceScalarField phi
|
|
(
|
|
cloudName + ":phi",
|
|
linearInterpolate(U) & mesh.Sf()
|
|
);
|
|
|
|
if (applyGravity_)
|
|
{
|
|
phiCorrect_() -= phiGByA();
|
|
}
|
|
|
|
forAll(phiCorrect_(), faceI)
|
|
{
|
|
// Current and correction fluxes
|
|
const scalar phiCurr = phi[faceI];
|
|
scalar& phiCorr = phiCorrect_()[faceI];
|
|
|
|
// Don't limit if the correction is in the opposite direction to
|
|
// the flux. We need all the help we can get in this state.
|
|
if (phiCurr*phiCorr < 0)
|
|
{}
|
|
|
|
// If the correction and the flux are in the same direction then
|
|
// don't apply any more correction than is already present in
|
|
// the flux.
|
|
else if (phiCorr > 0)
|
|
{
|
|
phiCorr = max(phiCorr - phiCurr, 0);
|
|
}
|
|
else
|
|
{
|
|
phiCorr = min(phiCorr - phiCurr, 0);
|
|
}
|
|
}
|
|
|
|
if (applyGravity_)
|
|
{
|
|
phiCorrect_() += phiGByA();
|
|
}
|
|
}
|
|
|
|
// correction velocity
|
|
uCorrect_ = tmp<volVectorField>
|
|
(
|
|
new volVectorField
|
|
(
|
|
cloudName + ":uCorrect",
|
|
fvc::reconstruct(phiCorrect_())
|
|
)
|
|
|
|
);
|
|
uCorrect_->correctBoundaryConditions();
|
|
|
|
//Info << endl;
|
|
//Info << " alpha: " << alpha_.internalField() << endl;
|
|
//Info << "phiCorrect: " << phiCorrect_->internalField() << endl;
|
|
//Info << " uCorrect: " << uCorrect_->internalField() << endl;
|
|
//Info << endl;
|
|
}
|
|
else
|
|
{
|
|
alpha_.oldTime();
|
|
phiCorrect_.clear();
|
|
uCorrect_.clear();
|
|
}
|
|
}
|
|
|
|
|
|
template<class CloudType>
|
|
Foam::vector Foam::PackingModels::Implicit<CloudType>::velocityCorrection
|
|
(
|
|
typename CloudType::parcelType& p,
|
|
const scalar deltaT
|
|
) const
|
|
{
|
|
const fvMesh& mesh = this->owner().mesh();
|
|
|
|
// containing tetrahedron and parcel coordinates within
|
|
const label cellI = p.cell();
|
|
const label faceI = p.tetFace();
|
|
const tetIndices tetIs(cellI, faceI, p.tetPt(), mesh);
|
|
List<scalar> tetCoordinates(4);
|
|
tetIs.tet(mesh).barycentric(p.position(), tetCoordinates);
|
|
|
|
// cell velocity
|
|
const vector U = uCorrect_()[cellI];
|
|
|
|
// face geometry
|
|
vector nHat = mesh.faces()[faceI].normal(mesh.points());
|
|
const scalar nMag = mag(nHat);
|
|
nHat /= nMag;
|
|
|
|
// get face flux
|
|
scalar phi;
|
|
const label patchI = mesh.boundaryMesh().whichPatch(faceI);
|
|
if (patchI == -1)
|
|
{
|
|
phi = phiCorrect_()[faceI];
|
|
}
|
|
else
|
|
{
|
|
phi =
|
|
phiCorrect_().boundaryField()[patchI]
|
|
[
|
|
mesh.boundaryMesh()[patchI].whichFace(faceI)
|
|
];
|
|
}
|
|
|
|
// interpolant equal to 1 at the cell centre and 0 at the face
|
|
const scalar t = tetCoordinates[0];
|
|
|
|
// the normal component of the velocity correction is interpolated linearly
|
|
// the tangential component is equal to that at the cell centre
|
|
return U + (1.0 - t)*nHat*(phi/nMag - (U & nHat));
|
|
}
|
|
|
|
|
|
// ************************************************************************* //
|