[mmany]

Hi,

I am developing an AI based solver for Openfoam and I am currrently trying to do the following thing:
1 - I have a rectangular mesh of dimension 256 * 128 * 1 cells ( adapted for a channel flow case) and some VolScalarFields.
2 - I have a neural network which gives a matrix of dimension 256 * 128.
3 - I want to assign the matrix values to the corresponding cell in OpenFoam
4 - I already have found a way to loop through the cells centers in a particular order to assign the matrix values and it works.

void CNN::updateFoamFieldChannelFlow(torch::Tensor &ai_output, Foam::volScalarField &field){
	int it = 0;
	forAll(field.mesh().C(), celli){
		int i = it / 256;
		int j = it % 256;
		field[celli] = ai_output[i][j];
		it++;
	}
}

5 - How can I also access the boundaries of the field[celli], to also change its value ? Is there a better way, more efficient to assign values to a VolScalarField ?

Thanks,
mmany

[kont87]

Hi,

I was looking for a similar answer, did you find a more efficient way to assign values back, instead of looping over?

Kind regards

Originally Posted by mmany

Hi,

I am developing an AI based solver for Openfoam and I am currrently trying to do the following thing:
1 - I have a rectangular mesh of dimension 256 * 128 * 1 cells ( adapted for a channel flow case) and some VolScalarFields.
2 - I have a neural network which gives a matrix of dimension 256 * 128.
3 - I want to assign the matrix values to the corresponding cell in OpenFoam
4 - I already have found a way to loop through the cells centers in a particular order to assign the matrix values and it works.

void CNN::updateFoamFieldChannelFlow(torch::Tensor &ai_output, Foam::volScalarField &field){
	int it = 0;
	forAll(field.mesh().C(), celli){
		int i = it / 256;
		int j = it % 256;
		field[celli] = ai_output[i][j];
		it++;
	}
}

5 - How can I also access the boundaries of the field[celli], to also change its value ? Is there a better way, more efficient to assign values to a VolScalarField ?

Thanks,
mmany