[CallumG]

Hi Foamers,

I'm using olaFlow (https://olaflow.github.io/), a set of wave boundary conditions used with interFoam to simulate some wave conditions and would like to process the wave height (i.e. water level where alpha>0.5) during runtime at certain locations.

I'm currently achieving this via adding the following function to my controlDict:

functions
{
    WG1
    {
        type sets;
        writeControl timeStep;
        outputInterval 1;
        interpolationScheme cellPoint;
        setFormat raw;

        sets
        (
            data
            {
                type uniform;
                axis z;
                start (2.5 5 1);
                end   (2.5 5 2);
                nPoints 1000;
            }
        );

    fields (alpha.water U);
    }
}

This samples the specified line and returns the alpha value against the z position. I then look through the series, find where alpha goes from >0.5 to <0.5, and voilà wave height.

The only issue is the curves I'm getting look like the attached, for regular sinusoidal waves.

When I check in ParaView the waves look correct.

Could anyone offer any advice on what's going on and how to correct it?

Cheers,
Callum

[Phicau]

Hi Callum,
I think I know what is going on. I have a couple of other pointers for you too. These learnings have made me adopt the method that I use (and recommend using) in https://github.com/phicau/olaFlow/bl...ls/breakwater/

- I believe that your postprocessing script may be wrong. When you sort the folders you should sort them by value, not by their name (I have been there before...)

- If the coordinates coincide with a face, sampling may jump between different cells. Add a small number (e.g., 1e-6) to the coordinates to prevent this.

- The uniform sampling procedure requires a lot of storage and gives you the wrong impression of accuracy, because in the end the mesh resolution is what it is, and uniform is just interpolating linearly to all those points. Using lineCellFace/midPointAndFace saves you from choosing a randomly large number of points to sample, as it samples only when the line passes through faces and in the middle point between them (cell centers). You can use interpolationScheme as cellPoint or cellPointFace.

- Integrating the VOF in the vertical direction is a widely accepted to obtain free surface elevation and provides subgrid accuracy, as VOF ranges from 0 to 1. The result is approximated when overturning waves exist, but it is very robust otherwise. Searching for VOF = 0.5 in the context of an algebraic VOF may be problematic too when overturning occurs too.

I hope it helps.

Best,
Pablo

[CallumG]

Hi Pablo,

thanks for the advice! As always, you are correct on every point :-)

Although I had implemented natural sorting, for some unknown reason, my time steps had been mixed up in my post-processing code. Equally, because my mesh is structured, and I had sampled at 1/4, 1/2, and 3/4 domain locations respectively, my sample lines did fall on cell faces!

So now, as you advised, I've repaired these issues and moved to type "midPointAndFace" with a "cellPoint" interpolation scheme. This is now much better!

I'd be interested to know how you would deal with free surface position. I'm not dealing with breaking waves at all and my mesh resolution is reasonably dense around the free surface, so I've opted to maintain the alpha>0.5 as before. It seems to work fine, but would welcome ant better logic.

All the best,
Callum

[Phicau]

My pleasure, glad to know that it worked!
I would continue with the 0.5 threshold, which is generally accepted. I don't expect much variation between looking for the free surface location by interpolating or by integrating the water column; perhaps the first one may show a little bit more of noise (?) since it only depends on the values of 2 cells rather than being an integrated result.

The only condition in which I have applied other values (0.1) is to calculate runup, since 10% of the cell being filled with water seemed wet enough to constitute runup.

Pablo