Welcome to our bi-weekly AI & Data Science meetup!
We discuss items of interest, learn how it works, and apply it to our own projects.
Please note that youth are welcome with parental attendance as this is an unsupervised community event.
Detection of Poor CNC Milling
The goal is to create a system that detects bad milling situations and warns the operator before the catastrophic failures, such as end-mill breakage.
We’re exploring analysis of captured sound in the frequency domain. We suspect there will be a detectable change in the spectral power when bad milling starts. (This is based on YouTube interviews with experienced CNC operators who say they know when they have the right speeds and feeds by the sound of the cut.)
- continuous sound capture, eg via a raspberry pi
- detection of CNC milling (automatic training data capture)
- there are several models that can detect normal home/office sounds
- capture tagged input data – create a means for the operator to tag detected milling operations as good or bad
- train AI systems to detect bad milling from the recorded sound
Code and more details on GitHub.
We succeeded in getting a Raspberry Pi to interface with an ICS43434 MEMS microphone! Screen shot is of an example python script that is capturing and plotting live sound data. Next steps are to document this setup, and see if we can create a system to automatically detect CNC milling operations. Part of our AI…