Parametric Failure Detection & Estimation in Linear Systems

An algorithm is reported for the sequential detection and estimation of parametric failures (deviations) which affect the system matrices of a linear, discrete time system. This represents the generalization of a technique, previously reported by the authors [1, 2], for the detection and estimation of soft failures in the statistical parameters of System inputs and measurement errors. In general, a parametric failure may be defined as a discrepancy between the true and design (model) values of any parameter characterizing the system behavior. The proposed scheme maintains the attractive features of the schemes in [1, 2] such as the use of a single Kalman filter. As before, the concepts of Sample Numbers (SN) and their average (ASN) obtained from the Wald Sequential Detector (WSD) are used for rapid detection and estimation of a failed parameter. Capabilities of the scheme are illustrated via simulation results for a servo control system.