Outdoor+indoor airflow

I want to model indoor airflow where the air flow comes from outside the window. I guess this is more of a Rhino/Grasshopper question than Butterfly.

For outdoor airflow, by default Butterfly will create a buildings as a block, i.e. the air does not flow inside the building. What do I need to do if I want to put the indoor example into a wind tunnel?

Note: I want to look into ways to model openings for naturally ventilated space. I will start from an opening (an open window), but later will transition into a more complex opening like louvers etc. If anybody has documents that I can read along this line, please feel free to suggest.


Hi @erydjunaedy

This is still an ‘outdoor’ model. All you need to do is model your space with the window openings included (gaps at the walls) and you will be able to assess wind flow passing through the space. Nothing else really changes in the settings of the model, apart from requiring a bit higher (than the exterior) refinement in the transition and interior NV spaces.

Kind regards,


@erydjunaedy Here is one demo of outdoor+indoor airflow.


Can we not do this in two steps?

  1. We run an outdoor study and measure the velocity received on the opening geometry
  2. We run an indoor study and use the velocity measured earlier as the boundary condition
    Can you please share your thoughts on this @TheodorosGalanos?
1 Like

@devang Sure that is the so-called decoupled approach. Although, you would usually use pressure measurements near openings for your initial conditions in an indoor (only) model.

However, when assessing NV, this isn’t really necessary and it is only used (usually) to lower computational requirements of the combined model. Today’s computers can handle most combined indoor/outdoor cases though, so I feel the extra model is not necessary and would definitely be less accurate.

If you were to assess other aspects of the indoor space however, e.g. fans, temperature sources, etc. then that is the normal approach used.

Kind regards,


Considering the size of the model the number of cells will be a huge number. Did you simulate this on a local machine or over the cloud?

1 Like

@devang Yeah. we run that model on my local machine.We usually used Butterfly for creating OpenFOAM dictionary. And we run parallel using windows 10 subsystem OpenFOAM, this version run very fast than the virtual box version OpenFOAM.


This is the idea of the study I am doing. There are many types of openings used here in the tropics. I do not think we have an accurate (or good enough) models representing those openings.

Yes, the smaller model is the goal. I am not going to have a wind tunnel just to model a half-open top-hung window.

On the accuracy: Sure, I will need a very fine mesh to capture the correct airflow through a window. If I get this correctly when I build the simpler opening model, then I think the accuracy of the simpler model is similar. Well, at least this will be reported later when the study is done: accuracy vs simplicity.

Stay tuned.

Here is a question about the outdoor-indoor airflow,can this simulate the temperature distribution indoor? In other words,can we use the heat transfer recipe to get the temperature distribution when using wind tunnel in a case?

@WuXue Yeah. Your workflow can be done by combined EnergyPlus simulation and CFD simulation.But it is a little difficult for Butterfly beginner.

You’d be surprised. I think that specific model is around 10-13m cells. Of course, I would suggest close to double for a production level simulation, but it already captures quite a lot as it is.

Hey,i am new to use butterfly. I have to do a research of indoor and outdoor wind environment coupling.Could you please share your script or just your idea? Thanks

It is simple to do outdoor +indoor airflow simulation. Create a normal wind tunnel for outdoor airflow, then pick the building with the hole(windows),then run simulation.

Got it. Thanks a lot

I have one more question. In this way, we cannot control the parameters of windows, such as the opening rate. So we cannot perform dynamic simulations to find the best setting. I try to build a Honeybee model to make detailed settings, then use the Butterfly to simulate. However, it does not work. Is there a way that I can achieve this aim?