Novel adaptive Bayesian scheme for real-time simultaneous anomaly detection and system identification

This study presents a novel approach for real-time anomaly detection and system identification. The approach eliminates the need for fixed threshold settings in anomaly detection and provides an efficient solution for simultaneous recognition of multiple anomaly types and identification of time-varying systems. Statistical models for random and gross errors are introduced to represent typical measurement anomalies, and Bernoulli random vectors are used for anomaly detection. Once potential anomalies are recognized, they are either excluded or compensated in further real-time system identification through detect-to-reject and detect-to-fix procedures. An adaptive Bayesian scheme updates both the Bernoulli and model parameters, allowing for real-time simultaneous anomaly detection and system identification. The approach is verified through numerical simulation and laboratory experiment. Moreover, it is implemented in a full-scale monitoring system. The proposed method effectively detects multiple anomaly types and achieves reliable identification results for time-varying systems.