Nexans uses various materials twisted together with either plastic extrusion or mechanical joining. To provide quality control, human operators check for any issues or breakages; however, any stoppages on the production line can be hugely expensive.
Normally, human operators sit and monitor the production line 24/7. Nexans wanted to automate this process using machine learning to continuously check the lines and flag any issues. It would free up staff expertise in decision making and planning, for example.
Bringing Industry 4.0 to the production line
Nexans Rognan worked with Orange Business to record large amounts of training data from its production lines, including phases of disrupted operation. The data, consisting of video footage, sound recording and vibration monitoring, was stored in the cloud via Google Cloud Platform (GCP).
Orange Business data scientists used this information to develop algorithms for predictive maintenance to keep equipment running without interruption. The system would also send out alerts to an operator if issues were spotted in the production process. The visualization for this was developed by Flow Design Bureau, which specializes in industrial flow systems. Operators can also manually trigger data capture if something happens to ensure that training data is kept up to date.
An affordable Raspberry Pi is used as the analysis server, with data capture taking place on a National Instruments controller. This highlights the fact that components for machine learning do not have to be costly. In addition, large amounts of data are continuously sent to the cloud platform for further, long-term analysis using a customized machine learning library.
The pilot program has so far been run on one production line that twists steel around a cable core of fiber-optic cables and one that winds together electrical copper wires. Nexans Rognan plans to adopt machine learning algorithms for other production lines in the future.