[vishalsacharya]

Hi,
I am looking at implementing a generic BC that also involves a second gradient at the boundary. Typically, if you look at zeroGradient or fixedGradient BC, the this->internalField value is available and then used to apply the values at the patch. For a second gradient, I need to take the gradient in the cell and then change its value first to a new gradient and then take the new gradient along with this->internalField to change the value at the patch. How do you generate gradient values in boundary cells only at the time of BC application so that there is low memory usage? A simple starting point is a zero second gradient BC upon which the generic BC can be built, but I am unsure how to get the gradient at the boundary internal field cells first. Also, the boundary patch face Sf() vector and the cell center C() at the boundary are also needed so that the correct component of the gradient along the cell-center-to-patch-center is used when applying/making changes.

Thanks in advance!

[massive_turbulence]

Originally Posted by vishalsacharya

Hi,
I am looking at implementing a generic BC that also involves a second gradient at the boundary. Typically, if you look at zeroGradient or fixedGradient BC, the this->internalField value is available and then used to apply the values at the patch. For a second gradient, I need to take the gradient in the cell and then change its value first to a new gradient and then take the new gradient along with this->internalField to change the value at the patch. How do you generate gradient values in boundary cells only at the time of BC application so that there is low memory usage? A simple starting point is a zero second gradient BC upon which the generic BC can be built, but I am unsure how to get the gradient at the boundary internal field cells first. Also, the boundary patch face Sf() vector and the cell center C() at the boundary are also needed so that the correct component of the gradient along the cell-center-to-patch-center is used when applying/making changes.

Thanks in advance!

Does this sound similar to your question? Gradient at the boundary

[vishalsacharya]

Originally Posted by massive_turbulence

Does this sound similar to your question? Gradient at the boundary

Andrew, It is not similar to this thread. That thread concerns the discrepancy in applying gradient BCs at the boundary at a different method compared to that which is used in the internal cells for grad from fvSchemes. What I am asking for is on how to make a gradient field just for boundary cells that can then be used to make a second gradient calculation.

[massive_turbulence]

Originally Posted by vishalsacharya

Andrew, It is not similar to this thread. That thread concerns the discrepancy in applying gradient BCs at the boundary at a different method compared to that which is used in the internal cells for grad from fvSchemes. What I am asking for is on how to make a gradient field just for boundary cells that can then be used to make a second gradient calculation.

This might be on the right track, line 162 from fixedGradientFvPatchField.H

  //- Return gradient at boundary
             virtual Field<Type>& gradient()
             {
                 return gradient_;
             }

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

I would assume you make the gradient initially a fixedGradient.
There are other patch fields as well. You just have to find the right type and inherit from its class.

[vishalsacharya]

Originally Posted by massive_turbulence

This might be on the right track, line 162 from fixedGradientFvPatchField.H

  //- Return gradient at boundary
             virtual Field<Type>& gradient()
             {
                 return gradient_;
             }

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

I would assume you make the gradient initially a fixedGradient.
There are other patch fields as well. You just have to find the right type and inherit from its class.

This function in that BC just returns the user supplied value for the "gradient" keyword in the BC file. I am trying to get the gradient from the cell center of the boundary cells and this would vary cell to cell and needs to be available at the time the BC is applied. It does not seem like OF has been designed with this possibility in mind.

[massive_turbulence]

Originally Posted by vishalsacharya

This function in that BC just returns the user supplied value for the "gradient" keyword in the BC file. I am trying to get the gradient from the cell center of the boundary cells and this would vary cell to cell and needs to be available at the time the BC is applied. It does not seem like OF has been designed with this possibility in mind.

Are you asking for the gradient of the pressure at the boundaries basically?

[vishalsacharya]

Let us say we are looking at a quantity 'f' and solving for this and I want to apply BC to this.
I want to do d/dx(df/dx) = 0 at the boundary. To apply this, I need to first know the df/dx (gradient) at cell center in cell next to boundary and then use this value along with value of f at cell center to update f at the patch.
d/dx(df/dx) = 0 is the same as applying
f_patch = f_cell + df/dx_cell * (x_patch-x_cell)

[vishalsacharya]

I figured it out.
The best way is to compute gradient and store it in a variable and then pass it as a word to this BC. You can then use lookup object to fund the value of the gradient and then apply it.
The other faster way is to do it in the main code, after solve() has completed. For this, the BC must be set to calculated first and then you can access the boundaryField and internalFieldRef() values and update as needed. For some reason this way of applying BC is much faster computationally... ?