[antoniofct]

Hello all,

I am trying to do an average of cells neighbors. For instance, if I want to get an average of Temperature in the internal field I can easily type

sum(T.internalField())/mesh.nCells()

.

What could be the function to do it just in cell neighbors? Could it be inside geometricField.C?

Thanks!

António

[duongquaphim]

I think you need to do a bit more for that:

const labelList& neighbour = mesh.cellCells()[celli]

scalar sum = scalar(0);

forAll(neighbour,i)
{
sum += T[neighbour[i]];
}

sum /= neighbour.size();

And loop over all the cells with that will give you an average of cell neighbours.

You can find more of functions relating to the mesh information in primitiveMesh class.

Duong

[antoniofct]

thanks a lot, I will give it a try!

Antonio

[vitor.geraldes@ist.utl.pt]

This a nice implementation, but, in parallel, it does not recognize neighbour cells if they are in subdomains different from the one of the central cell. Is there a simple implementation to address this problem?

[tom_flint2012]

I think if you use field.correctBoundaryConditions(); it will work OK. Where field is the new neighbour field you assign.

This works fine for me

All the best

[CFDanielGER]

Hello everybody, for a new hybrid turbulence model I also need to calculate the mean of the current and neighboring cells. I am curious if there is a more efficient way doing the average instead of using a forAll loop?

[olesen]

Originally Posted by CFDanielGER

Hello everybody, for a new hybrid turbulence model I also need to calculate the mean of the current and neighboring cells. I am curious if there is a more efficient way doing the average instead of using a forAll loop?

It definitely needs a loop (actually more than one) , but if you are doing all cells (not just a single cell) then it makes much more sense to loop across the faces. You obviously need your output field for storing the result and an additional array for counting the number of neighbours.
Walk each internal face, get owner/neighbour locations and use that to add to the output (in two locations) and the neighbour count (again in two locations). After that walk the boundary faces (owner only). Finally, iterate across the boundary patches, adding in the off-processor information.

Finalise your output field by preforming the averaging operation.

Depending on what you need, interpolate cell vales to the faces and vol reconstruct from them could also be an interesting thing to look at.

[CFDanielGER]

Originally Posted by olesen

It definitely needs a loop (actually more than one) , but if you are doing all cells (not just a single cell) then it makes much more sense to loop across the faces. You obviously need your output field for storing the result and an additional array for counting the number of neighbours.
Walk each internal face, get owner/neighbour locations and use that to add to the output (in two locations) and the neighbour count (again in two locations). After that walk the boundary faces (owner only). Finally, iterate across the boundary patches, adding in the off-processor information.

Finalise your output field by preforming the averaging operation.

Depending on what you need, interpolate cell vales to the faces and vol reconstruct from them could also be an interesting thing to look at.

I think the last thing you mentioned is how I currently implemented it since I only need to average over the current cell and it's closest neighbors

volScalarField VTMav = fvc::surfaceSum(this->mesh().magSf() * fvc::interpolate(VTM(gradU))) / fvc::surfaceSum(this->mesh().magSf());