[KeiJun]

Dear Sir,

I'm trying to add a source term to the UEqn.H, which is named "ASm".

ASm = C * ∂q/∂t, which is a tensol and is calculated in advance (C is a constant coefficient).

Then, I want to add ASm * U as a source term, but it is a product of tensol and vector.

What should I describe this product to the UEqn.H source code?

I checked "DarcyForchheimer", which also calculate a product of a tensol and U, but I don't understand how to do.

Could you advise me, please.


KeiJun

[KeiJun]

Dear Sir,

I keep trying making my solver, but cannot compile.

I want to add a new source term to UEqn.H, which is ASm * U.

ASm is C * ∂q/∂t, in which C is a constant (scalar) and ∂q/∂t is calculate by another equation (in qEqn.H) in advance.
q is the reaction amount per catalyst weight (the unit is mol/kg), so ∂q/∂t is the reaction rate.

Then, is the type of calculated ∂q/∂t fvScalarMatrix?
If so, is the type of ASm also fvScalarMatrix?
And then, is the type of new source term I want to add, ASm * U, defined as volVectorField? (which is inner product of fvScalarMatrix and volVectorField, and described as ASm & U.)

When I compile this solver, the error message:
no match for 'operator&' (operand types are 'Foam::fvScalarMatrix {aka ~}' and 'Foam::volVecorField {aka ~}')
has come.

I have no idea why this error occurs.

Could you advise me how to descibe this new source term.
I am really looking forward to your reply.


Sincerely,

KeiJun

[mkraposhin]

Hi,

when you are adding source term to any equation, you must decid on how this term will be discretized: in explicit or implicit way?

If you want to discretize you source in explicit way, then you must create volVectorField for UEqn, with dimensions equal to density*acceleration:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    ==
    SourceU
);

Where SourceU - volVectorField

If you know, that your source field depends on velocity and you can write it like next expression:
SourceU = C*U,
then you can discretize source in implicit way:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    ==
    fvm::Sp(C,U)
);

U - is a volVectorField, C - volScalarField, dimensions of C is "density/time", so that C*U dimensions will be density*acceleration,
operator fvm::Sp(C,U) will return fvVectorMatrix. It is a good practice to be sure that C field is negative in implicit formulation

Next, if you know that field C can be positive, then it means that it's values will be subtracted from the diagonal, and this will lead to ill-conditioned matrix. Then, you can make semi-implicit formulation, which switches betweeb implicit/explicit depending on sign of C:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    +
    fvm::SuSp(-C,U)        
);

[Shibi]

Hi,

Sorry is this is too basic, but is the there any difference if the explicit/ or any of the other sources is made as:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    -
    SourceU
);

instead of

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    ==
    SourceU
);

They will both assemble a matrix and enter the solve function. The "==" is just to make it with closer resemblance to the mathematical equation, right?

[mkraposhin]

Yes '==' is just an alias to '-'

[openfoam_aero]

Originally Posted by mkraposhin

Hi,

when you are adding source term to any equation, you must decid on how this term will be discretized: in explicit or implicit way?

If you want to discretize you source in explicit way, then you must create volVectorField for UEqn, with dimensions equal to density*acceleration:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    ==
    SourceU
);

Where SourceU - volVectorField

If you know, that your source field depends on velocity and you can write it like next expression:
SourceU = C*U,
then you can discretize source in implicit way:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    ==
    fvm::Sp(C,U)
);

U - is a volVectorField, C - volScalarField, dimensions of C is "density/time", so that C*U dimensions will be density*acceleration,
operator fvm::Sp(C,U) will return fvVectorMatrix. It is a good practice to be sure that C field is negative in implicit formulation

Next, if you know that field C can be positive, then it means that it's values will be subtracted from the diagonal, and this will lead to ill-conditioned matrix. Then, you can make semi-implicit formulation, which switches betweeb implicit/explicit depending on sign of C:

solve
(
    fvm::ddt(rho,U)
    +
    fvm::div(phi,U)
    +
    fvm::SuSp(-C,U)        
);

what if C is a vector here?

[shouvik ghorui]

Hi,
How to hendle if the source term is (gammaPsi)*grad(C), where equation is
fvm::ddt(C)
+fvm::div(phiup,C)
-fvm::laplacian(gammaPsi,C)
Thanks