It seems that there is a segmenation/clustering tool in IVY. I never tested it but it seems a good starting point.
I have done my segmentation tool for mesh but I will keep it for myself. Because it is not so simple to make such tool, you have to take into account the “unrollability” of the mesh, dimension limitation, topology to be able to rivet sheets …
Are the “coral-like” segments individual meshes that keep their initial topology (e.g. tris), or complex ngon meshes with one face and a number of vertices greater than 4?
Without getting to much into the deepest secrets of your work, do you produce the pattern with diffusion limited aggregation or something similar, since the result differs quite a bit from your previous post?
Yes I use my special script, here with triangular meshes. But this is just a rendering the path is quite difficult to have the real thing. The pattern must be flattenable so I suppress all inner point of the mesh. That is generating deformation so some moves are necessary to lower the tolerances between rivets …
by parallel agent-based search protocol, the non-linear stripes are the result of the indecisive behavior of the agents, which change their attitudes in tight turns. As they come together at seams and make their way to the ground, they find alignment as linear stripes. Across the ballooning expanses, they express the recursive quality of the computational description. Bifurcations and splits in the pattern tighten and spread to produce porosity. Like bubblegum blown just to the point of popping, the skin becomes thinner and more open across the wider expanses of the spheres, allowing light to come through. Where the volumes merge, the stripes between them become more densely packed, lending to the structural performance.
Mainly about parallel agent’s
What I can discern in the upper reference image is that the base mesh is a collection of simple platonic solids. I guess, these spheres are distributed to look like a cloud. Bigger spheres seem to form the cloud center, whereas somewhat smaller ones occupy its periphery.
In the lower image, modified, stilted spheres are added. Some seem to be open meshes, like the columns in Zaha Hadid’s Serpentine Sackler Gallery expansion.
The first challenge is to unify the individual geometries into a single, clean base mesh with a quad tessellation and soft transitions. This could be done by re-meshing the geometries with Starling. Note though, how the topology of the base mesh is already very carefully designed. The stilts seem to feature a vertically oriented diamond tessellation.
The resulting new base mesh is then segmented into developable strips, probably with one of the approaches described by @laurent_delrieu above, to form the blue mesh segments of the substructure.
Note, how the segmentation seems to follow a rule, where there must be unselected mesh faces between two or more adjacent segments. The individual segments seem to be offset and thus aren’t interconnected, which renders this secondary structure unsuitable for development.
And here’s where the structural, primary shell comes into play, which stitches it all together!
It seems to be based on agent-based design, as described in the referenced text.
There are some plugins like Kangaroo 2, the BOID Library that is usually used in combination with Anemone, Quela, Culebra, etc. that can all handle agents, but scripting (e.g. Python, C#, Visual Basic) is always more flexible, especially if you want to incorporate intrinsic rules.
The agents start at the bottom of the stilts and in the connective valleys in-between the previously individual sphere geometries of the base mesh.
Their main mission seems to be to form developable bridging strips over the above mentioned individual strips of the blue substructure, thus rendering the cloud statically sound.
At the stilts, the agents produce very dense strips (light blue) that cover the entire surface of the base mesh and seem to be all interconnected.
In the connective zones, the agents also produce denser strips (violet), but leave some breathing room in-between.
The light blue and violet strips surely form the primary supporting structure, over the blue substructure.
The rest of the base mesh, gets more loosely covered in yellow strips, which probably only have a secondary structural role, apart from holding the strips of the substructure underneath together.
As said by PirateBoy, agents are used to fullfill many constraints
Solidity => maximize intersection at right angle between lower and upper
Fabricability => Folding perpandicular to the plate
Fabricability => maximum length
Fabricability => unrollability