A divide and conquer approach to anomaly detection, localization and diagnosis

With the growing complexity of dynamic control systems, the effective diagnosis of all possible failures has become increasingly difficult and time consuming. The virtually infinite variety of behavior patterns of such systems due to control inputs and environmental influences further complicates system characterization and fault diagnosis. To circumvent these difficulties, we propose a new diagnostic method, consisting of three elements: the first, based on anomaly detection, identifies any performance deviation from normal operation; the second, based on anomaly/fault localization, localizes the problem, as best as possible, to the specific component or subsystem that does not operate properly and the third, fault diagnosis, discriminates known and unknown faults and identifies the type of the fault if it is previously known. Our prescriptive method for diagnostic design relies on the use of self-organizing maps (SOMs) for regionalization of the system operating conditions, followed by the performance assessment module based on time-frequency distributions (TFDs) and principal component analysis (PCA) for anomaly detection and fault diagnosis. The complete procedure is described in detail and demonstrated with an example of automotive engine control system.