part 4 - more prototyping
Collaborators: Tianyi Han, Chenghui Nan
The next phase of the project was to continue developing a fabrication process and construction system, this time taking into account a specific site, and more realistic design considerations such as budget and schedule.
The most important change in this iteration is the module geometry. While we continued to use the Non-Random Voronoi logic, the module geometry shifted from a simple cube to a Rhombic Dodecahedron. The dodecahdron can be generated from a cube, so the change was easily implemented. We made this change for two reasons. First, the dodecahedron geometry fills twice as much space as a cube, which means the system is more efficient. Second, while filling more space, it actually makes the module more structrually sound, especially at the vertices of the cube.
From the previous iteration, we learned that the rod-to-node detail was not efficient enough for mass production; we were constantly drilling into the concrete nodes in order to make the rods fit. In this iteration, we developed a process for casting 3-D Printed receptors directly into the nodes. This way, we reached an efficient balance of material properties: the concrete as the cheap, imperfect material, and the 3-D Printed receptors as the precise material.
Left: Rubber mold. Center: 3D-Print Right: Concrete Node, with 3D-Printed receptors and acrylic rods
We also began testing the detail that allows modules to connect to one another. We determined a plastic snap-fit detail would work best, so we developed a 3D-Printed snap that is easy to insert, but slightly more difficult to pull apart. This way, the modules become part of a larger system that is intuitive as a system such as Legos or K'nex.
The snap consists of 3 parts. First, there is a 3-D Printed sphere with inserts in specific directions. Those inserts then accept the 3-D Printed snaps. Finally, the relevant nodes are castwith a 3-D Printed receptor, which accept the snap. The video below shows the first working prototype of the snap fit, which in this case is a vertex-to-vertex connection.