Optimization of an Observation System for a Failure Diagnosis.

The article reports on optimization of an observation system for a failure diagnosis. Anomaly diagnosis of control systems has become important from the viewpoints of improvement of availability and protection from disasters. Information about the anomaly is usually gathered through an observation system which is instrumented for a control purpose and treated appropriately. This article discusses optimization of the observation system which enables fast and accurate detection of the anomaly for the dynamical system represented by a state variable. It assumes that the treated system is a discrete linear one and that the anomaly is modeled by an unknown jump adding to the state at an unknown time.