algorithmic modeling for Rhino
Playing with optimization concept used for Soumaya Museum's facade.
- 6-dimensional k-means clustering
- ~6000 triangular panels grouped in 15 families
- equilibrium reached after 40-50 iterations
- computation time ~8-12 minutes
- native GH components + Anemone
Does this involve in any kind of machine learning?
You mentioned K-clustering... Does this relate to KNN Algorithm?
Thanks, very interested in re-implementing your algorithm.
that means, you have done this for a result of similar planar triangles in a set of a family?
I try to do like the engineers from Gehry Technologies did. Resulting similar planar hexagonal panels in a set of a family for the soumaya museum. They used a k-means clustering. Is this quit similar as you did?
@Hendrik - 6 dimensions mean that each point has 6 coordinates. Every panel is represented as one of such points (as you need 6 coordinates to describe a triangle). Once you encode the panels as points, you can measure distance between them with the regular distance formula ( sqrt((x2-x1)+(y2-y1) + ...))). With closest point search you can cluster those points, and take some averages... Look up Lloyd's algorithm to learn more about this part.
iam trying very similar and iam wondering what do you mean with 6-dimensional k-means clustering? In my version i compare every attribute from each panel to another. This is i think 1-dimensional k-means clustering, or not?
Have a look here for more specifics on the hex panel optimization done for the museum.
great work !!! I´m kind of new using anemone components, can u upload a tutorial for this ? it would be awesome .
Dario - I saw Alexander Pena de Leon's lecture at 2013 eCAADe(Prague), in which he described this optimization method.
Actually, does anybody knows if the parametric design part of the Soumaya Museum facade was made by the same studio (FR-EE) or it involved any specialized computational design studio? Thnx!
More Soumaya info:
© 2023 Created by Scott Davidson. Powered by
You need to be a member of Grasshopper to add comments!