Using funding from the Yorkshire Innovation Fund, Sheffield Hallam University's Dr Fabio Caparrelli and Teks Sarl are collaborating on a new R&D project that aims to realise a camera-based system to monitor the operation of a CNC milling machine that produces customised mechanical parts for advanced automotive and aerospace applications. The monitoring system aims to automatically detect faults during the operation (eg wearing of a milling tool), allowing the machine to run 24/7 without direct assistance by a human operator. Any faults detected by the system would automatically shutdown the machine as well as alert the operator by email or SMS.

The system’s hardware is based on inexpensive components available off the shelf. In particular, a Raspberry Pi board and a small CMOS infrared camera (RaspiCam Noir) are used in a protective case located closely to the CNC machine. The Raspberry Pi camera is used to constantly monitor the machining of a part.

When the milling tool starts to wear out, it will normally produce a significant amount of sparks. The sparks are detected by a customised computer vision algorithm running on the Raspberry Pi from the images captured by the attached infrared camera.

To remove the image background and simplify the task of detecting the sparks (as well as making it more robust), a small visible-light filter is used in front of the camera lens to extract the sparks information out of the whole image. This is possible as the light generated by the sparks will contain some infrared component that is not blocked by the visible-light filter (see picture below).

By limiting the search area to a small region of interest around the end of the milling tool, it is possible to speed up the vision algorithm and produce results in real-time. Using standard image filtering and thresholding, the system is able to robustly monitor the task continuously and shutdown the machine when a fault is detected. This is carried out using a relay activated by the Raspberry Pi board.

As the board can be connected to the Internet through an Ethernet connection or WiFi, the system is able to email the operator when a fault occurs. The system is also controllable through a web browser, making it possible to manage it by a PC, smartphone or tablet. Finally, the system is able to alert the operator by SMS by simply adding a USB modem to the Raspberry Pi board.

System features

• Remotely accessible over the internet
• Can alert operator by SMS, email, etc
• Automatic shutdown of CNC machine
• Inexpensive (well less than £100)
• Programmable solution
• Customisable for different tools

Benefits and outputs

• Allows operator to use the machine 24/7
• Reduced machine idle time
• Increased production and sales
• Potential to become a commercially viable industrial product