One of the cool things I learned at my local Fab Lab was snap tight construction: no nails or glue, just slots and tabs.
Designing a successful pressfit object is not trivial. The slots and tabs have to have enough space to slide together, but tight enough to hold. This is a matter of fractions of a millimeter, and the results depend on the material and the fabrication. Back in the day, I fiddled around with the original code for a cool little design calculator that is essentially a little expert system to design 2D pieces that snap together to make boxes. (Current version here.)
This winter researchers at the University of Tokyo report a cool design system that creates complicated 3D snap fit joints in wood, tagged ‘Tsugite’ [1]. These joints might be used in furniture or in architecture.
Obviously, 3D is harder than 2D, and there are a lot more possibilities that might be realized. In addition to designing a potential joint, strategic corners need to be rounded to be able to snap together.
Different solutions may perform better. E.g., a given size tenon has a limited strength in different directions and under different forces. The tool has manual inputs to enable the human to explore trade offs. The user may also consider the aesthetics of the joinery.
This is pretty cool.
This initial version only designs ‘square’ joints, abstractly constructed of cubic voxels. The system could extend to design non-rectangular joints, at least in principle.
I think the most important limitation of this first version is that it only designs individual joints. This is important because you really need to design the whole configuration of joints (e.g., a chair or roof) together. Still, this system will help get the parts of an initial design worked out.
As far as I can tell, this system has only a rudimentary knowledge of the properties of wood. A given design might be great for one species of wood and not work at all for another. For that matter, it is often important to account for the grain and imperfections in an individual piece of wood.
So, there is still need for human expertise, at least so far!
One thing that might be interesting to do with this system is to try some machine learning. Perhaps it can learn some of the heuristic knowledge that human wood workers know, and perhaps it can learn some new and unexpected designs. Given the many centuries of human attention, I’d be surprised if a computer can discover something really new, but it if did, it would be pretty significant indeed.
Nice work, all.
- Maria Larsson, Hironori Yoshida, Nobuyuki Umetani, and Takeo Igarashi, Tsugite: Interactive Design and Fabrication of Wood Joints, in Proceedings of the 33rd Annual ACM Symposium on User Interface Software and Technology. 2020, Association for Computing Machinery: Virtual Event, USA. p. 317–327. https://doi.org/10.1145/3379337.3415899
- University of Tokyo, Simple Software Creates Complex Wooden Joints That Interlock With No Nails, Glue, or Tools Needed, in University of Tokeyo – SciTechDaily, October 31, 2020. https://scitechdaily.com/simple-software-creates-complex-wooden-joints-that-interlock-with-no-nails-glue-or-tools-needed/
Robert McGrath's Blog 