Model-based diagnosis of systems emphasizing a holistic approach

Diagnosis of technical systems is very important both from the technical and economic point of view. Monitoring and diagnosis start with the observation of symptoms, which depend on the fast time dynamics and on the system's evolution based on the slow time coordinate. Model-based diagnosis can be performed by symptoms directly and by a holistic mathematical model of the entire system decomposed into subsystems, whereas the submodels of the related subsystems can be of different accuracies dependent on their goals. If these models are adjusted at every lifetime required for condition assessment of the operating system/system in service, and if the models are verified and validated, then they represent the best knowledge base available for diagnosis. Adaptation of mathematical models uses system identification tools which can produce verified, validated and usable mathematical models. These models serve for fault detection, localization, assessment and cause finding, for trend prediction and decisionmaking for further use of the system. Consequently, the procedure described combined with an extended knowledge base of the limit states etc. of the system can be seen as an intelligent diagnostic method.