I am doing research on which shape performs better for an office high-rise building, taking into consideration also variations in window to wall ratio and orientation. In order to do this, I set up 4 different floor plans to test: round, square, rectangular and triangular. Everything seems to work out fine for the square, rectangular and triangular plan layouts, but for some reason, there seems to be an issue with the round shape and windows are added also to the core walls. I tried to find the issue, but I just cannot figure it out. I also installed the Ladybug Legacy tools to verify whether it can be solved with the new components, but it doesn’t work also with the 1.3 version. I`m quite new with the new legacy tools, so I also cannot figure out how to get rid of the doors in the façade. I tried to organize the file in case any of you has the patience and kindness to help me with this. I would really appreciate any sort of suggestion, as I have run out of ideas.
There may be some exceptions I’m not aware of, but in general the underlying geometry algorithms in LBT only deal with planar, not curved surfaces. EnergyPlus also shares this assumption, so curved geometry is typically represented by a mesh or a set of facetted surfaces for energy workflows. I’m not sure what the constraints are for Radiance, however, in both cases, even relatively coarse “planarization” will not lead to much loss in accuracy, since the slight differences in shape and orientation is averaged out over multiple surfaces (similiar to how finite difference methods can approximate analytrical solutions). So, in your case, I would reccomend converting your circle into some N-sided polygon, where N is around 8 +/- 2 depending on your tolerance for long simulation times.
Thank you for your reply, it was really helpful. I would have thought about this issue. I changed the circle to a polygon as you suggested and noticed that it only works if the number of segments is a multiple of 8. It works with 8, 16, 24 segments. But it doesn’t work for example with 10, 11, 12 and so on. That`s very weird. Do you think that a polygon with 16 segments will require a long simulation time? I think the results will be quite inaccurate for a polygon with 8 segments, given that I want to test the performance of a round geometry. Thanks a lot again for your input, I spent a lot of time rewriting the script in different ways and nothing worked. Thanks!
No problem. I think you should just run both the 16-sided and 8-sided polygon and see for yourself if the simulation time is too long, and if the difference in results is large enough to justify doubling the surfaces. If you’re just doing one simulation its fine to have an inefficient model, but most of the time we run multiple iterations (for sensitivity, uncertainty, optimization etc) where efficiency will matter.
Agree with @SaeranVasanthakumar.
The other thing i would check, is why some of the faces have small windows and other have one big one for the whole face.
The more windows you have the longer time it takes to simulate. If you are only interested in energy simulations you can strive for lesser windows.
-A.
