A Relevance-Based Data Exploration Approach to Assist Operators in Anomaly Detection

摘要

Data is emerging as a new industrial asset in the factory of the future, to implement advanced functions like state detection, health assessment, as well as manufacturing servitization. In this paper, we foster Industry 4.0 data exploration by relying on a relevance evaluation approach that is: (i) flexible, to detect relevant data according to different analysis requirements; (ii) context-aware, since relevant data is discovered also considering specific working conditions of the monitored machines; (iii) operator-centered, thus enabling operators to visualise unexpected working states without being overwhelmed by the huge volume and velocity of collected data. We demonstrate the feasibility of our approach with the implementation of an anomaly detection service in the Smart Factory, where the attention of operators is focused on relevant data corresponding to unusual working conditions, and data of interest is properly visualised on operator's cockpit according to adaptive sampling techniques based on the relevance of collected data.