[titio]

Hi Foamers,

I have a small question concerning relaxation. Is there any difference, when writting the code, between the
following forms of relaxation, when solving the Navier stokes equation for laminar flow of newtonian fluids.

Form 1

fvVector UEqn
(
fvm::ddt(U)
+ fvm::div(phi,U)
- fvm::laplacian(nu,U)
);

UEqn.relax();

solve (UEqn == -fvc:: grad(p));

Form 2

fvVector UEqn()
(
fvm::ddt(U)
+ fvm::div(phi,U)
- fvm::laplacian(nu,U)
);

UEqn().relax();

solve (UEqn() == -fvc:: grad(p));

Is there any significant difference between the two formulations.

Thanks,

Titio

[gschaider]

No. The second UEqn is probably a autoPtr of a fvMatrix.

[henrik]

Hi Titio,

Bernhard is absolutely right.

Form 2 (and Form 1) are quoted incorrectly. Form 2 should read:

tmp<fvVectorMatrix> UEqn
(
fvm::ddt(U)
+ fvm::div(phi,U)
- fvm::laplacian(nu,U)
);

(see simpleFoam)

By using tmp (or alternatively autoPtr) one is able throw away the U-Matrix before solving the p-equation. This way, you can save some peak-memory in steady-state solvers.

Regards, Henrik

[gschaider]

Originally Posted by henrik

Bernhard is absolutely right.

... and to short to be understood. I planned to add a second sentence "This is only an 'administrative' construct and doesn't change the physical meaning of the equation" and obviously forgot that.

But Henrik's answer is better anyway