Elastic Shape Analysis for Anomaly Detection in Textile Finishing Time-Series Data

Quality control represents a challenge in the industrial textile finishing field. The general unavailability of labelled data from real production plants and the imbalanced nature of the problem suggest tackling it with a semi-supervised approach that monitors the behaviour of the system and identifies whether shifts from the nominal conditions arise. In particular, we utilize techniques from Elastic Shape Analysis to study the time-series data collected from sensors installed in real production plants and to extract features used to define distance metrics that quantify the data variability. The results of this exploratory study support the efficacy of the proposed strategy.