M2L ® Anomaly Detection Benefits • Identify failures that are missed by QC • Identify glitch in service network • User-friendly • Plug & Play Use case: towards zero defects in glass production Customisation • Anomaly detection model that links the occurrence of blisters to outliers in specific sensor readings 12-24 hours earlier in the process. Business case • Detecting production settings leading to too many blisters, thus reducing the amount of glass that has to be discarded if acted upon in time. Radboud University Spin-off [email protected] Exclusive to M2L ® • Root cause determination • Combining academic and industrial experience • High precision via customisation • Deployable on edge devices • Integration with cloud Examples anomalous temperature sensors, defective land-line connection Challenge • Subtle changes in the glass production process lead to blisters in the final product • Batches of glass with too many blisters have to be discarded • Large but incomplete database of historic sensor readings from a glass furnace