A Novel Reconfigurable Distributed Anomaly Detection Framework

Intelligent Automation, Inc. (IAI) proposes to develop a novel, distributed framework for the detection of faults and anomalies. The proposed framework is based on dynamic, reconfigurable agents and is capable of automatically representing and managing disparate data sources. Meanwhile, efficient processing requires automatic selection and extraction of features from different data sources for fault or anomaly detection. In our approach, advanced feature-selection techniques are employed to enhance the anomaly detection performance. Finally, since there are many techniques for fault or anomaly detection and each has different data requirements and a different chance of success, including physical model-based methods, rule-based methods, and data-driven methods, we know that these methods need to be integrated. Our distributed approach is open and flexible to accommodate any of these methods as agents, and a decision level fusion technique is proposed for pooling/aggregating detection decisions made by these detection agents. A promising information fusion method called transferable belief model (TBM) is proposed as the core fusion technology in this proposal to achieve more reliable and accurate fault/anomaly detection.