The symptom observation matrix for monitoring and diagnostics

Monitoring of complex technical systems in operation/service is a task conc erning safety, reliability and risk management. Additionally, the monitorin g and diagnosis of such systems is also recommended for the reduction of co sts, and to increase the life of systems under consideration. The fundamental quantities for monitoring and diagnosis are symptoms contai ned in measured data of a time-variant system. Symptoms are observable and sensitive with respect to faults or damage. Questions such as how to find s ymptoms, how they should be evaluated for information condensation in order to detect, and if possible to localize, faults or damage, are discussed, a nd, additionally, the choice of the measuring matrix and the test forces ar e treated in the context of the state space formulation of time-variant mod els. For the purpose of information condensation and the evaluation of symptoms, the symptom observation matrix is introduced and manipulated by the singul ar value decomposition. The properties of the observation matrix with its s ingular values and singular vectors show their applicability for monitoring and diagnosis. (C) 2001 Academic Press.