This started out as a basic point attractor exercise by dividing segmented chords into equal parts, then a point was placed at each of these locations. A selection process was set up to reduce and then make a specific selection of your choice of pairings. A similar solution was then attached in order to actuate the said points in the z-direction in a positive or negative direction to influence the underlying attraction for the NURBS curve.
Layering
The next big breakthrough was to add multiple layers of these arrangements with different overall lengths as well as segmented spacing. This was important since it added the level of richness in the final solutions and a higher number of correct solutions through the Generative design algorithm processor as can be seen below.
In the beginning, there were issues in terms of getting sufficient acceptable solutions that produced a correct geometry the both Revit and Dynamo could accept. Below you can see a picture of the issues we faced earlier where most solutions were not acceptable nor did they create a shape with our algorithm. If you compare both have 12 pages however, the upper has a filter on the surface.area command and makes sure that only the positive solutions are thus the only contained within those pages, unlike those below which also have 12 pages. Big improvement, with the same size of generation sample per cycle - nearly 233% more acceptable solutions were created overall. For me, this is where you see a real performance drop compared to what I have seen in other programs such as Grasshopper in Rhino - Rhino seems to accept these types of generative designs and also improve them overall over the iterations.
Finally, it was time to get into the selection process of just randomly checking all over the different pages and getting a feel for patterns that were occurring that you liked within the filters and isolating those in order to get the desired looks and curves. Then a further visual inspection of multiple solutions is done through the GUI until you are happy with a selection and open in Dynamo. The power of getting access to a data set of such a vast number of correct solutions is incredible. I even stumbled upon a variant that resembles one of Zaha Hadid`s work Heydar Aliyev Centre in Baku, Azerbaijan that can be seen below.
