So we needed to figure out how to stream the G-code. None of this functionality would work with just sending one G-code file to the printer. This would allow someone to decide which components need to be disassembled, but would use the computer control to do the precise work. We also plan on allowing a human to override the what the computer is doing.The system will try to melt it, then use the algorithm to see if it was successful, and may need to try to melt it again for longer. We may need to heat a solder blob multiple times to melt all of the solder off.
That should decrease the amount of calibration the end user needs to perform. We plan to use our machine learning algorithm to look at camera data multiple times to make sure we’re aligned with the soldering blob. Then it needs to move the point of the soldering gun to each solder blob and heat it up.
The user would load in the piece (red in our mock-up)
We developed a 3d prototype in blender to construct the parts we needed. We pitched it to our peers and got some great feedback. The basic idea was that you’d load a part, flip the entire system over, it would unscrew every screw it found, and the pieces would fall into a bin below it. (just using different heads for different jobs).Īs we started work on the project, we shifted focus from construction to deconstruction. This could be useful for very fine movements with applications in areas like watchmaking, painting model trains, soldering of circuit boards, etc. Originally we came together to start a project we called “Mini-Maker” which would use a 3d printer type setup combined with a really cool method being used on the JWST that enables quick jogging to a location and then very precise movement, all with just one motor per dimension. Three of us ( and came together to see what we could do to extract useful parts from other devices. It’s getting harder (and more expensive) to get components for our projects.