[gaza]

Hi
I have a volScalarField defined in standard manner:

rho
    (
        IOobject
        (
            "rho",
            U.mesh().time().timeName(),
            U.mesh(),
            IOobject::NO_READ,
            IOobject::NO_WRITE
        ),
        U.mesh(),
        dimensionedScalar("rho", dimDensity, 0.0)
    )

Later in the code I calculate the values of rho for each cell using forAll loop.
However, in the program I get divide error because the forAll loop sets values only
in cells and at the boundaries the 0 is left. I solved it by applying the second forAll
loop over boundary patches but it slows down the calculations remarkably.
Is there any method to avoid the second loop over patches?

[frantov]

This reminded me a case where div by zero is fixed.

To prevent that code blows up due to a division by zero in a viscosity calculation, it is advised to add SMALL.

SMALL is a static constant scalar included in

https://www.openfoam.com/documentati...8H_source.html


for example

return nu0_/scalar(1) + pow(k_* strainRate() + SMALL, (1-n));


perhaps you can try a similar approach?


...................

(also, I just edited my response)

You can also write Zero

this is done in BC initialization

new Field<Type>(this->size(), Zero)


https://develop.openfoam.com/Develop...FvPatchField.C

[gaza]

Originally Posted by frantov

This reminded me a case where div by zero is fixed.

To prevent that code blows up due to a division by zero in a viscosity calculation, it is advised to add SMALL.

SMALL is a static constant scalar included in

https://www.openfoam.com/documentati...8H_source.html


for example

return nu0_/scalar(1) + pow(k_* strainRate() + SMALL, (1-n));


perhaps you can try a similar approach?


...................

(also, I just edited my response)

You can also write Zero

this is done in BC initialization

new Field<Type>(this->size(), Zero)


https://develop.openfoam.com/Develop...FvPatchField.C

Hi Frantov,
Thank you very much for your hint.
It is good solution however it does not solve my problem.
I need to set the values at the boundaries in each iteration the same
as in the cells that adhere to the boundaries.


A while ago I found the solution:
1) added the line at the end of rho constructor

"zeroGradient"

2) instead of the second forAll (over boundaries) I wrote

rho.correctBoundaryConditions();

This solution is slightly faster than that with two forAll loops.
I hope it is correct and gives the same results