PrintRite Uses TensorFlow Machine Learning to Watch for 3D Printing Failures

Printing even a relatively small object can easily take several hours on an FFF (Fused-Filament Fabrication) 3D printer. If you’re printing something large enough to take up all of your printer’s build volume, that can take multiple days. For long prints, it’s impossible to constantly monitor your printer for failures, and that’s why PrintRite will do it for you using TensorFlow machine learning.

PrintRite can either run on its own, or as a plugin for the popular OctoPrint remote printer management software. It utilizes Google’s TensorFlow machine learning system to identify failures in the 3D printing process. When it does, it will stop the print and notify you so that you can try to salvage the job — or put out the fire if things have gone especially awry. Currently, it can detect a fire or a completely failed print, but eventually it will be able to do a lot more as a library of successful print training images is built up.

To start using PrintRite, you’ll need both a Raspberry Pi Model 3 B+ and a Raspberry Pi Zero. The full-size Raspberry Pi will run OctoPrint and TensorFlow, and the Raspberry Pi Zero handles the feed from a Raspberry Pi Camera module via MJPEG-Streamer. The camera does have to be positioned in an exact location with a specific mount for the overall PrintRite system to be a success and to ultimately detect a range of failures, so you’ll want to work with the creator, Eric, to determine the best mounting solution for your printer.