Smarter cleaning.
Built around the machine.
A proof of concept exploring a lower-cost, custom-built approach to automated machine tool cleaning.
A client asked us to investigate an alternative to a costly machine vendor cleaning solution. We built a small prototype addition to their existing hardware, combining a Jetson Nano, cameras and custom machine learning software.
- Proof of concept
- Camera and Jetson Nano setup
- Video review and labelling
- Custom machine learning software
- Machine and door interlock integration
A small addition.
A practical new capability.
The setup connected to the customer’s machine so the cameras began recording when a job started. Those recordings provided the examples needed to develop a system that could recognise when cleaning was appropriate.
Examples first.
Recognition next.
We reviewed and labelled the recordings, identifying examples where the compartment contained no parts, no job was running and a cleaning cycle was needed. We also labelled recordings where cleaning was unnecessary.
These examples formed the basis for the custom machine learning software, helping it distinguish between conditions that required cleaning and those that did not.
The right conditions.
A short cleaning cycle.
The prototype combined recognition with the machine’s operating state to identify when a job had finished, the compartment was empty, the door was closed and cleaning was needed.
When those conditions were met, the setup could trigger a short automated cleaning cycle. The machine integration prevented the door from being opened while cleaning was in progress.
The prototype demonstrated the potential for a lower-cost, custom-built alternative to the machine vendor’s solution, integrated with the customer’s existing equipment.
This work was a proof of concept, demonstrating the approach through a prototype.