[typerhack]

Hi,

I am an architect. I am trying to analyze my project's building form. I learned the basics of OpenFoam (OF). I also found the wind around building tutorial, but that is not what I am trying to do. I am designing a form base on parameters that can change. I am trying to run an optimization on 10000 different forms. I don't know which solver I should use (maybe airflow?) and how can I measure the success of the form. for example, does OF returns any value for form or the mesh face in other to improve the form? (I want to create a data table based on that value and parameters) The goal here is not structural analysis is form finding.

any help would be appreciated.
thanks,

[yambanshee]

I would think for something like that you'd want to monitor the forces being exerted on the building, and minimize that.



something along the lines of this:

functions
{
    forces1
    {
    // Mandatory entries
        type            forces;
        libs            ("libforces.so");
        patches         (walls); //rename patch to the walls of your building

    // Optional entries

    // Field names
        p               p;
        U               U;
        rhoInf            1.225;
        rho             rhoInf;

    // Reference pressure [Pa]
        pRef            0;

    // Include porosity effects?
        porosity        no;

    // Store and write volume field representations of forces and moments
        writeFields     yes;

    // Centre of rotation for moment calculations
        CofR            (0 0 0);

        executeControl    runTime;
        executeInterval   10;

        writeControl    runTime;
        writeInterval   10;
    }
}

The best result will likely be something shaped like a semi-circle with the long side perpendicular to the wind, or even something such as a symmetrical airfoil. If you need to account for wind from all sides, a cylinder is likely going to win

[typerhack]

Thank you. that helped me a lot. Sorry for the bother but I don't know very much about this stuff. I am trying to create a good fit for the bounding box of the mesh. I want to read some researches about it but I don't know what to search. would you mind giving me some advice?

Originally Posted by yambanshee

I would think for something like that you'd want to monitor the forces being exerted on the building, and minimize that.



something along the lines of this:

functions
{
    forces1
    {
    // Mandatory entries
        type            forces;
        libs            ("libforces.so");
        patches         (walls); //rename patch to the walls of your building

    // Optional entries

    // Field names
        p               p;
        U               U;
        rhoInf            1.225;
        rho             rhoInf;

    // Reference pressure [Pa]
        pRef            0;

    // Include porosity effects?
        porosity        no;

    // Store and write volume field representations of forces and moments
        writeFields     yes;

    // Centre of rotation for moment calculations
        CofR            (0 0 0);

        executeControl    runTime;
        executeInterval   10;

        writeControl    runTime;
        writeInterval   10;
    }
}

The best result will likely be something shaped like a semi-circle with the long side perpendicular to the wind, or even something such as a symmetrical airfoil. If you need to account for wind from all sides, a cylinder is likely going to win

[yambanshee]

Originally Posted by typerhack

Thank you. that helped me a lot. Sorry for the bother but I don't know very much about this stuff. I am trying to create a good fit for the bounding box of the mesh. I want to read some researches about it but I don't know what to search. would you mind giving me some advice?

The idea about the bounding box is to ensure that your boundaries (inlet, outlet, etc) are far enough from the object you are interested in (the building in this case) that any unrealistic boundary conditions are mitigated by the time it gets to your object of interest. It becomes a question of how well do you know what exactly is happening at the boundaries. If you know the exact velocity/pressure profile, your boundary may as well be right next to the object of interest. If you are simply saying that there's some average inlet velocity and we're going to set the entire inlet domain to this velocity, that will probably be quite unrealistic, and so it's best to put your boundaries quite far.


It is especially important for your downstream boundaries, as you'll have no idea what happens to the flow after interfacing with your building, so by doing something like setting a fixed pressure value downstream and near to the building, you'll be forcing the flow to have that pressure value, and that will affect the solution upstream dramatically.


Short answer; giving a exact 'value' for how far your boundaries should be is very subjective. When testing aerofoils, all boundaries are generally at bare minimum 10 chord lengths away, but such a large domain in 3d can get very computationally expensive. The most correct thing to do would be to pick a value that 'looks good', and get results for this. Next, expand the domain out and get a new set of results. Keep repeating this until the results stop changing with larger domains, and then you know your simulation is not dependent on the domain size. The same method can (and should in any meaningful paper) be applied to mesh density, turbulence boundary conditions, and all other things that are assumed.

[typerhack]

Thanks for the information. I get it that it is experimental. But is there no guide? I am really confused. The things I want to do are simple:

1- I have some meshes exported in .stl format.
2- I want to import them to OpenFoam.
3- I want to run simpleFoam.
4- I want to calculate the drag coefficient for these meshes.

I do not have any specific condition and I'll leave everything at the default. I still don't get how to set up the case. is it enough only to use blockMesh and snappyHexMesh?

Originally Posted by yambanshee

The idea about the bounding box is to ensure that your boundaries (inlet, outlet, etc) are far enough from the object you are interested in (the building in this case) that any unrealistic boundary conditions are mitigated by the time it gets to your object of interest. It becomes a question of how well do you know what exactly is happening at the boundaries. If you know the exact velocity/pressure profile, your boundary may as well be right next to the object of interest. If you are simply saying that there's some average inlet velocity and we're going to set the entire inlet domain to this velocity, that will probably be quite unrealistic, and so it's best to put your boundaries quite far.


It is especially important for your downstream boundaries, as you'll have no idea what happens to the flow after interfacing with your building, so by doing something like setting a fixed pressure value downstream and near to the building, you'll be forcing the flow to have that pressure value, and that will affect the solution upstream dramatically.


Short answer; giving a exact 'value' for how far your boundaries should be is very subjective. When testing aerofoils, all boundaries are generally at bare minimum 10 chord lengths away, but such a large domain in 3d can get very computationally expensive. The most correct thing to do would be to pick a value that 'looks good', and get results for this. Next, expand the domain out and get a new set of results. Keep repeating this until the results stop changing with larger domains, and then you know your simulation is not dependent on the domain size. The same method can (and should in any meaningful paper) be applied to mesh density, turbulence boundary conditions, and all other things that are assumed.

[yambanshee]

check the tutorial case located at incompressible>simpleFoam>windAroundBuildings. It'll be very close to what you hope to achieve.


Depending on the shape of your building(s) it could be as simple as blockmesh -> snappyHex with some default values. In my experience, the most challenging part of CFD is usually meshing. I've spent the better part of half a year just developing one mesh for my thesis, and it's no where close to what I hope to have as an end product, and only marginally comparable to meshes that can be constructed using commercial software.


But yes, to get okay results; snappyHex with a similar setup to the tutorial will be fine. It'll give you a rough idea of what to expect.