CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   OpenFOAM Programming & Development (https://www.cfd-online.com/Forums/openfoam-programming-development/)
-   -   anisotropic porous media in OpenFOAM (https://www.cfd-online.com/Forums/openfoam-programming-development/108441-anisotropic-porous-media-openfoam.html)

 hawkeye321 October 23, 2012 18:47

anisotropic porous media in OpenFOAM

Does anyone have any experience in simulating flow in an-isotropic porous media in OF? I mean simulations in which the permeability in x direction is different from the permeability in y direction.

My main problem is that when I am adding the darcy soource term as
+ ((1/Perm)*(fvm::Sp(Visc,U)))

I do not know how to tell the OF that Perm will be different based on the direction!

 nimasam October 24, 2012 04:48

hi mahdi
is in this condition permeability ("Prem") a vectorField?

if yes, you can define a volVectorField for Prem, then
assign your directional variation with setFields in 0 folder

 hawkeye321 October 24, 2012 09:37

Permeability Vector

Hi nimasam
Thanks for the comment.
But, if I define the permeability as volVectorField then, when I want to multiply it by velocity what kind of vector multiplication do I have to use; definitely not inner or cross product. Right?

 Hisham October 24, 2012 09:52

Hi Mahdi

In the case of non-isotropic permeability, the permeability is a symmetric tensor. For anisotropy, you should assign zero off-diagonal values and be left with kxx, kyy and kzz.

Hope this helps!

Regards
Hisham

 hawkeye321 October 24, 2012 09:57

multiplying permeability tensor and velocity vector

Dear Hisham
Thanks for the comment. I know that true, but, as I mentioned in my previous post, I am not sure how to multiply permeability tensor and velocity vector in the momentum equation. What kind of multiplication operator do I have to use?

 nimasam October 24, 2012 10:26

dear mahdi,
if you want to divide Ux/Kx , Uy/Ky , Uz/Kz
then i guess you may want to use some thing like this :
forAll(U,celli)
{
S[celli].x() = U[celli].x()/Kx[celli].x();
........
}

 Hisham October 24, 2012 10:35

I think you first would have to define the tensor as k (kxx 0 0 kyy 0 kzz), then get the inverse of the tensor as inv(k). The product should be an inner product: + (inv(k) & (fvm::Sp(Visc,U))

I would appreciate it if you report what happened!

Best regards,
Hisham

 hawkeye321 October 24, 2012 10:55

Dear Hisham
I have define a tensor named PermTenInv which has diagonal terms of 1/kxx,1/kyy and 1/kzz. Now when i am doing explicit calculations in form of

+ ((PermTen) & (fvc::Sp(Visc,U)))

there is no error. However, for implicit calculations in form of

((PermTen) & (fvm::Sp(Visc,U)))

I get the following error
--------------------------------------------------------------------
BR1IM003.C:84:40: error: no match for ‘operator&’ in ‘PermTen & Foam::fvm::Sp(const Foam::dimensionedScalar&, Foam::GeometricField<Type, Foam::fvPatchField, Foam::volMesh>&) [with Type = Foam::Vector<double>, Foam::dimensionedScalar = Foam::dimensioned<double>](((Foam::GeometricField<Foam::Vector<double>, Foam::fvPatchField, Foam::volMesh>&)(& U)))’
/opt/openfoam171/src/OpenFOAM/lnInclude/wordI.H:167:19: note: candidates are: Foam::word Foam::operator&(const Foam::word&, const Foam::word&)
/opt/openfoam171/src/OpenFOAM/lnInclude/dimensionSet.H:273:29: note: Foam::dimensionSet Foam::operator&(const Foam::dimensionSet&, const Foam::dimensionSet&)
make: *** [Make/linux64GccDPOpt/BR1IM003.o] Error 1
--------------------------------------------------------------------------
Unfortunately, I have to discretise this source term implicitly. Any ideas how to fix this error?

 Hisham October 24, 2012 11:23

Dear Mahdi,

It is really sad that this did not work directly. You can report this as a bug (assuming a similar feature should be available). If you are highly motivated for your problem you can try to add a new fvm::Sp function to fvmSup.C/.H that can take a "dimensionedVector & sp" argument. You then have to dig deep into OpenFOAM's number crunching mechanisms and figure out which elements of fvm.diag() should you multiply by which component. Here ends my ideas. I hope you can find an easier solution!
Code:

```template<class Type> 00153 Foam::tmp<Foam::fvMatrix<Type> > 00154 Foam::fvm::Sp 00155 ( 00156    const dimensionedScalar& sp, 00157    const GeometricField<Type, fvPatchField, volMesh>& vf 00158 ) 00159 { 00160    const fvMesh& mesh = vf.mesh(); 00161 00162    tmp<fvMatrix<Type> > tfvm 00163    ( 00164        new fvMatrix<Type> 00165        ( 00166            vf, 00167            dimVol*sp.dimensions()*vf.dimensions() 00168        ) 00169    ); 00170    fvMatrix<Type>& fvm = tfvm(); 00171 00172    fvm.diag() += mesh.V()*sp.value(); 00173 00174    return tfvm; 00175 }```
Regards,
Hisham

 hawkeye321 October 24, 2012 12:44

THnaks Hisham. I am going to discuss it in a new post to be sure that this is a bug in OF.

 AnjaMiehe September 16, 2013 09:05

is there any progress in the subject? I am coding something similar for solidification and encounter the same problem.

Regards, Anja

 vitor.geraldes@ist.utl.pt April 7, 2016 03:34

:::::::::::::::::::::::::::::::::::::::::

 mateusps July 25, 2016 16:20

Any solution?

Hi y'all,

I also need to implement an implicit anisotropic source term. Did anyone find a way to do that?

fvm::Sp seems to not accept a tensor as coefficient...

thanks

 vitor.geraldes@ist.utl.pt July 25, 2016 17:14

Try to create an homogeneous hydrodynamic resistance and then subtract a given fraction (fAniso) in the direction nT (unitary vector):

+ fvm::Sp(Rmush,U)
- fAniso*Rmush*(U&nT)*nT

if fAniso is 1, the resistance in the direction nT is zero, if fAniso is 0 then you have no anisotropy.

 mateusps July 25, 2016 17:29

Didn't get it

Thanks for the reply, Vitor, but I didn't get it.

In your suggestion, what is Rmush? Because if it is a tensor, it does not work...

Maybe what you're suggesting is something I've already tried: to implement that in 2 terms, one implicit and isotropic, and one explicit and anisotric. But in my case it does not work, because the largest part of the resistance is specifically the anisotropic part, and any explicit treatment of that refuses to converge...

 vitor.geraldes@ist.utl.pt July 25, 2016 17:59

Rmush is a volScalarField which may be computed for instance by the Kozeny Carman correlation.

 mateusps July 25, 2016 23:14

How to code the anisotropic heterogeneous source term as implicit?

Ok, but this way the anisotropic part is still explicit, right? Any way to code the anisotropic heterogeneous source term as implicit?

 All times are GMT -4. The time now is 19:03.