Hacker News Reader

Comment by Duanemclemore | Hacker News Reader

This is phenomenal. I wish I had it to share with a student I worked with last year. Even without this, he came to similar realizations and explored similar territory with significant results.[0] He began by comparing and contrasting traditional Japanese joinery with CNC milled joints. He realized that the critical difference was that Japanese carpentry joints are almost always entirely squared off (due to the flat saw and chisels). Yet the inside corners created with CNC were always rounded. So he let this critical difference be the animating concept of the project.

So if you enjoy this, you might enjoy Luke's work too!

[0]https://issuu.com/lukemurrayarc/docs/portfolio_2023 starting at page 26-27.

View on HN

Here is what immediately comes to my mind.

    - Figure out a way to parameterize joints
    - Create an automated evaluator that uses finite element analysis
    - Evolve joints with a genetic algorithm

Oooo... that's a great idea! I would love to see the output of that project!

You left out one step/criteria:

- constrain joints so that they can be cut with a tool on a CNC with the material secured on the machine in a convenient fashion in a reasonable period of time

which is something I've been working on.

constrain joints so that they can be cut with a tool on a CNC with the material secured on the machine in a convenient fashion in a reasonable period of time

Could one make that a part of the evaluator/fitness function?

Yes, but it needs to be included as a consideration up-front and not left to afterthought or accident.

Yea, but wood isn’t actually uniform, so a finite element analysis might find a joint that is stronger on paper, but the particular piece of wood it needs to work on makes it weaker in practice.

It might be neat to do it automatically based on a picture of the wood, but that is a much bigger problem.

When working with wood in an automated/industrial environment, there are two options:

- design for the worst-case scenario for a given practice setup (note that wood is graded on how many knots there are and how tight they are expected to be) and work around that, eliminating pieces w/ problematic grain - use engineered lumber where appropriate

This is a marked contrast to some styles of hand-woodworking --- I can still recall walking with a great-uncle in the woods when we came to a walnut tree where the trunk had an odd angle to it (apparently the main trunk had broken off vertically in a storm and a branch had continued the upward growth) --- he immediately took a colorful handkerchief out of his pocket and tied it around the tree, looked carefully at a couple of local landmarks visible through the trees, and then confident of being able to return to the spot, we went to his cabin, fetched a saw and ax and proceeded to cut the (smallish) tree down and sectioning it off, taking it back to his cabin where he covered the ends of the central portion around the angle in melted paraffin.

Almost a decade later during another visit, he announced he had something to show me and got out a rifle which I remembered as having a broken stock --- he had of course restocked it (once it had dried sufficiently) with the tree which had grown at an angle which was perfect to make the stock out of.

Reply on news.ycombinator.com