Machine2Learn, NL



Ali Bahramisharif is a co-founder and managing director of Machine2Learn. He holds an interdisciplinary PhD in neuroscience and artificial intelligence, and is a former assistant professor at the University of Amsterdam. He has a track record of working in various high-tech sectors including healthcare, telecom, and manufacturing industries.

​Presentation ​

Automated Quality Control of Glass Production with Artificial Intelligence

Anomaly detection is of pivotal importance in many industries. But once there is a tool that can spot anomalies, what can be done with it? It can certainly help with taking actions to cope with the detected anomalies, but is it possible to do more?

There are some business processes where it is possible to go beyond coping with the detected anomalies: intervene in the process to prevent anomalies. Glass manufacturing is one of them. Once a link between sensor measurements and the anomalies is identified, anomalies occurrence can be controlled by controlling the linked measures.

Since an anomaly detection model can be based on many sources of measures, it is not feasible to use a trial and error approach to find the core sources driving the anomalies occurrence. Luckily, AI can support this quest: it is possible to determine what measures are the ones that have the most impact on the type of anomaly, and the nature of the impact.

Machine2Learn has applied this methodology on data from glass manufacturing, and using historical data has shown to correctly identify about 16% of the anomalies that can explain the occurrence of blisters and bubbles in the final production 24 hours in advance. Root cause analysis has been applied to determine the underlying causes of the anomalies. The results allow for defining measures to prevent the anomalies.