CRII: CPS: A Knowledge Representation and Information Fusion Framework for Decision Making in Complex Cyber-Physical Systems

Performance monitoring data (e.g., measurements, logs, events) are becoming increasingly accessible and abundant (in terms of cost and availability) in modern distributed complex systems such as computer systems and networks, integrated buildings, industrial systems, transportation networks and power-grids. With efficient exploration of such data, health monitoring, diagnosis and prognosis can be greatly improved beyond the current state-of-the-art. From the perspective of monitoring, anomaly detection and root-cause analysis of such systems, technical challenges arise from a large number of subsystems that are highly interactive and operating in diverse modes. A semi-supervised tool for anomaly detection and root-cause analysis in complex systems has been proposed. The proposed framework is based on a spatiotemporal time series feature extraction scheme built on the concept of symbolic dynamics for discovering and representing Granger-causal interactions between the subsystems. The proposed tool aims to (i) capture multiple operational modes of a complex CPSs within a single data-driven model, (ii) only use nominal data for training and perform inference without requiring labeled anomalous data, and (iii) implement root-cause analysis in a semi-supervised way for a large variety of anomalies (e.g., one failed pattern, multiple failed patterns, one faulty node, and multiple faulty nodes). In the final year of the project, we performed three case studies associated with various CPSs, i.e., bridge health monitoring, cyber-physical security for industrial robots, and non-intrusive load monitoring (NILM) for residential buildings. First, for bridge damage detection, dense sensor networks (DSN) are used to explore spatiotemporal relationships among bridges. For the industrial robots, nominal robotic system characteristics are learnt from historical data and streaming operational data is used to detect and isolate possible anomalies. Finally, the proposed framework is also used to analyze mixed, multivariate time-series to improve the residential home energy disaggregation performance. We demonstrate better or comparable performance with respect to the state-of-the-art methods for all of the above applications.