[Gaetano]

Dear OF developers,

I have a 10M$ question: is there a way to do a run-time line sampling? My problem is that the slope of the sampling line depends on calculations being made. I wrote a post-processing script, but I can't believe I need to store double-digit GBs of data to get a few undred kB ascii file at the end of the post-processing!

Here's my problem step by step:

1. I have a vector: it's the direction of the sampling line;
2. I need to find the points where this line intersect with the boundary: those will be the start and end points of the sampling line (no need for line-patch intersection to pass through;
3. I need to sample a field along that line.

I've been browsing the forum for days looking for an answer, but with really poor results.

I'm trying to do something like this (still not tested, basically taken from here):

    dictionary interpolationDict = mesh.solutionDict().subDict("interpolationSchemes");
    autoPtr<interpolation<vector> > Uinterp = interpolation<vector>::New(interpolationDict, U);

    vectorList positions = ..something;

    vectorList Ui;
    labelList cellI;

    forAll(positions, i)
    {
        cellI.append(mesh.findCell(positions[i]));
        Ui.append(Uinterp->interpolate(positions[i], cellI));
    }

How does it look? Am I on the right path? Or perhaps I'm reinventig the wheel again? Would you suggest me better approaches, if any?

Gaetano

[Gaetano]

Nobody has any clue? I'll be satisfied even with a "not possible" answer.

[ngj]

Hi Gaetano,

I have not been able to find the thread, but I recall a discussion, where some people needed to sample over a line, which was fixed, but the whole of the computational domain was moving, i.e. so was the line.

If I remember correctly they looked at sloshing and they seemed to solve the problem.

I hope that this has helped you a bit.

Good luck,

Niels

P.S. This is definitely not a "not possible" problem, but it might require a bit of work.

[Gaetano]

Thanks for the hint, but it doesn't help me a lot.

Those guys were looking for a line moving with the mesh and not with respect to it, so they simply commented out the mesh updating part in the sample and probe utilities. Here's the post where the discussion started, two posts below they figured out the way.

I know that my problem might require some work (and I'm not afraid at all), but I'm still looking for a way to do that.

In other words: I'm not looking for someone doing my homework for me. I'd like a clue from more skilled users who might suggest me few logical steps to tackle the problem.

I still can't believe that I have to save 30 GBs of time folders, filter out the results of the sample utility via a bash/awk script just to end up with few kBs of neat results...

[jherb]

Have a look at this thread:
http://www.cfd-online.com/Forums/ope...swak-type.html
I think the idea with the probes could be used in your case.

Also have a look at swak4foam and function objects with "type createSampledSet". Search for it in the files named "contolDict" in the Example directory (I use the development version on the port_2.0.x branch): http://openfoamwiki.net/index.php/Co...am#Development
here is one of the examples:
http://hg.code.sf.net/p/openfoam-ext...ontrolDict#l85
I am not sure but you might be able to output the values sampled on this line after each timestep.