Vibrating assortments of non-OEM hairsprings to work with old balance wheels, etc
I have a large collection of old balances. It looks like most balances were made by a few groups of subcontractors. This is more true of the chronographs. Surprisingly Val-23 staffs trend at a premium. On the other hand I have a box of elgin staffs for post war stuff still in the plastic bubble packs.
I tend to buy lot assortments, so the unused parts tend to wind up in the general parts box. Without packaging there is often no way to identify what it is. I am sure I have 30T and 26X parts.
I still think there should be a way to use micro-controllers that could take some of the grunt work out of this. Most of it is a statistical grading process. Basic robotics does seem to be taught in high school. mostly though with the competitions to destruction. However some AP students use the maker space to do some fairly complex challenges.
What does not seem to be taught is hydraulic and pnuematics. Robotics seems to be based on servo control. Physics also works different at the macro level where air currents and gravity do not react the same way as they do at the automotive part size level. The tiny inertial masses are hard to visualize. Hand are also more adept at shoveling food in the mouth and scraping meat from bones, than controlling tiny parts what fly across the room.
I suspect the AI tail eating dragon probably ate the bestfit catalog. The bestfit catalog does remain copyrighted. So what comes out of generic replies is probably garbage. I have better ways of wasting time than attempting to drive a chatbot.
We did get an example of an optical flow scanner, at the makerspace, which was not quite optimized for the omega plate I had. I downloaded an open source version. Also have a box of lenses.
Such is a non trivial process. So far I have yet to meet anyone else with motivation. No discussion on any of the threads I started. Why I tend to reply to the more general inquiry threads which get more views.
I have been sort of outlining and researching my 'Child History of AI.' Looking at the old programs from 40 50 years ago. Which are quite proficient at identifying and sorting parts.
Spending a few hours here and there a week is not the most progressive. This week I was revisiting JPEG huffman codes. Found some you tube tutorials, and comparing them to code I wrote in the 1990s. JPEG destroys correlation in image processing. (why there is a strong interest in bayer demosaicing.) Dumping the high frequency data is not a bad thing, yet the training data is mostly garbage. So it is real easy to get distracted in the trees. I still have not found a way to get the Tissot pictures out of the Quark Immedia database, which may have additional code compression as while I can see the source disassembly calls a standard JPEG library, but the Tags are not in the data. So the exercise was yet another dead end.
More to the subtopic at hand. A lot of the traditional ways have been lost. Or are too time consuming. There is also a lot of baggage and dogma around, that stuff has to be done the way it was done in the 18th or 19th century. Or the way it was industrialized in the 20th.
It is like we want some swiss gnome to do it, and remain behind the curtain.