Calculation of approximate distance to nearest patch for all cells and boundary by solving the Eikonal equation in advection form with diffusion smoothing. More...

Inheritance diagram for advectionDiffusion:

[legend]

Collaboration diagram for advectionDiffusion:

[legend]

Public Member Functions
	TypeName ("advectionDiffusion")

	advectionDiffusion (const dictionary &dict, const fvMesh &mesh, const labelHashSet &patchIDs)

virtual bool	correct (volScalarField &y)

virtual bool	correct (volScalarField &y, volVectorField &n)

Public Member Functions inherited from patchDistMethod
	TypeName ("patchDistMethod")

	declareRunTimeSelectionTable (autoPtr, patchDistMethod, dictionary,(const dictionary &dict, const fvMesh &mesh, const labelHashSet &patchIDs),(dict, mesh, patchIDs))

	patchDistMethod (const fvMesh &mesh, const labelHashSet &patchIDs)

virtual	~patchDistMethod ()

const labelHashSet &	patchIDs () const

virtual bool	movePoints ()

virtual void	updateMesh (const mapPolyMesh &)

template<class Type >
Foam::wordList	patchTypes (const fvMesh &mesh, const labelHashSet &patchIDs)

Additional Inherited Members
Static Public Member Functions inherited from patchDistMethod
static autoPtr< patchDistMethod >	New (const dictionary &dict, const fvMesh &mesh, const labelHashSet &patchIDs, const word &defaultPatchDistMethod=word::null)

template<class Type >
static wordList	patchTypes (const fvMesh &mesh, const labelHashSet &patchIDs)

Protected Member Functions inherited from patchDistMethod
	patchDistMethod (const patchDistMethod &)=delete

void	operator= (const patchDistMethod &)=delete

Protected Attributes inherited from patchDistMethod
const fvMesh &	mesh_

const labelHashSet	patchIDs_

Detailed Description

Calculation of approximate distance to nearest patch for all cells and boundary by solving the Eikonal equation in advection form with diffusion smoothing.

If the diffusion coefficient is set to 0 this method is exact in principle but the numerical schemes used are not rendering the scheme approximate, but more accurate than the Poisson method. Also many models relying on the distance to the wall benefit from this field being smooth and monotonic so the addition of diffusion smoothing improves both the convergence and stability of the solution of the Eikonal equation and the behavior of the models using the distance field generated. However, it is not clear what the optimum value for the diffusion coefficient epsilon should be; a default value of 0.1 is provided but higher values may be preferable under some circumstances.

Convergence is accelerated by first generating an approximate solution using one of the simpler methods, e.g. Poisson. The method used for this prediction step is specified in the 'advectionDiffusionCoeffs' sub-dictionary, see below.

References:

    P.G. Tucker, C.L. Rumsey, R.E. Bartels, R.T. Biedron,
    "Transport equation based wall distance computations aimed at flows
     with time-dependent geometry",
    NASA/TM-2003-212680, December 2003.

Example of the wallDist specification in fvSchemes:

    laplacianSchemes
    {
        .
        .
        laplacian(yPsi) Gauss linear corrected;
        laplacian(yWall) Gauss linear corrected;
        .
        .
    }

    wallDist
    {
        method advectionDiffusion;

        // Optional entry enabling the calculation
        // of the normal-to-wall field
        nRequired false;

        advectionDiffusionCoeffs
        {
            method    Poisson;
            epsilon   0.1;
            tolerance 1e-3;
            maxIter   10;
        }
    }

Also the solver specification for yWall is required in fvSolution, e.g. for simple cases:

    yPsi
    {
        solver          PCG;
        preconditioner  DIC;
        tolerance       1e-4;
        relTol          0;
    }

    yWall
    {
        solver          PBiCG;
        preconditioner  DILU;
        tolerance       1e-4;
        relTol          0;
    }

    relaxationFactors
    {
        equations
        {
            .
            .
            yWall           1;
            .
            .
        }
    }

or for more complex cases:

    yPsi
    {
        solver          GAMG;
        smoother        GaussSeidel;
        cacheAgglomeration true;
        nCellsInCoarsestLevel 10;
        agglomerator    faceAreaPair;
        mergeLevels     1;
        tolerance       1e-4;
        relTol          0;
    }

    yWall
    {
        solver          GAMG;
        smoother        symGaussSeidel;
        cacheAgglomeration true;
        nCellsInCoarsestLevel 10;
        agglomerator    faceAreaPair;
        mergeLevels     1;
        tolerance       1e-4;
        relTol          0;
    }

    relaxationFactors
    {
        equations
        {
            .
            .
            yWall           0.7;
            .
            .
        }
    }

See also: Foam::patchDistMethod::Poisson Foam::patchDistMethod::meshWave Foam::wallDist

Source files

Definition at line 178 of file advectionDiffusionPatchDistMethod.H.

Constructor & Destructor Documentation

◆ advectionDiffusion()

advectionDiffusion	(	const dictionary &	dict,
		const fvMesh &	mesh,
		const labelHashSet &	patchIDs
	)

Definition at line 48 of file advectionDiffusionPatchDistMethod.C.

Member Function Documentation

◆ TypeName()

TypeName ( "advectionDiffusion" )

◆ correct() [1/2]

bool correct ( volScalarField & y )

virtual

Implements patchDistMethod.

Definition at line 77 of file advectionDiffusionPatchDistMethod.C.

References GeometricField::null(), and y.

Here is the call graph for this function:

◆ correct() [2/2]

bool correct	(	volScalarField &	y,
		volVectorField &	n
	)

virtual

Implements patchDistMethod.

Definition at line 84 of file advectionDiffusionPatchDistMethod.C.

References GeometricField::boundaryFieldRef(), Foam::dimless, Foam::fvm::div(), Foam::fvc::div(), Foam::fvc::grad(), Foam::fvc::interpolate(), Foam::fvm::laplacian(), Foam::mag(), n, IOobject::NO_READ, IOobject::NO_WRITE, Foam::notNull(), patches, fvMatrix::relax(), fvMatrix::solve(), Foam::fvm::Sp(), y, and Foam::Zero.