Recent advances in the smart factory created new opportunities in industrial support, specifically in anomaly detection and asset management for maintenance purposes. Data collected from machines in operation is integrated in cyberspace with advanced cockpit and dashboard visualisation tools, as well as computerised maintenance management systems (CMMS). This paves the way to collaborative environments for Cyber Physical Systems, that assist maintenance operators in making better decisions while dealing with real time data in a dynamic context of interconnected systems. To this aim, models and techniques for data representation and treatment are highly required. In this paper, we propose a state detection service for Cyber Physical Systems, able to identify anomalies based on large amounts of data incrementally collected, organized and analysed on-the-fly. The service combines in a novel way data summarisation and data relevance techniques, to focus the computation on relevant data only, as well as a multi-dimensional model, that organises summarised data according to multiple dimensions, for flexible anomaly detection according to different analysis requirements. A pilot case study in the smart factory is also described, to demonstrate the applicability of the approach.
Big data summarisation and relevance evaluation for anomaly detection in cyber physical systems
Bagozi A.;Bianchini D.
;De Antonellis V.;Ragazzi D.
2017-01-01
Abstract
Recent advances in the smart factory created new opportunities in industrial support, specifically in anomaly detection and asset management for maintenance purposes. Data collected from machines in operation is integrated in cyberspace with advanced cockpit and dashboard visualisation tools, as well as computerised maintenance management systems (CMMS). This paves the way to collaborative environments for Cyber Physical Systems, that assist maintenance operators in making better decisions while dealing with real time data in a dynamic context of interconnected systems. To this aim, models and techniques for data representation and treatment are highly required. In this paper, we propose a state detection service for Cyber Physical Systems, able to identify anomalies based on large amounts of data incrementally collected, organized and analysed on-the-fly. The service combines in a novel way data summarisation and data relevance techniques, to focus the computation on relevant data only, as well as a multi-dimensional model, that organises summarised data according to multiple dimensions, for flexible anomaly detection according to different analysis requirements. A pilot case study in the smart factory is also described, to demonstrate the applicability of the approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
