Signalverarbeitung mit analytischer Redundanz

Complex systems normally are observed using the output signal of sensors which describe the inner state of the system. This statement is true only in those cases, where the output signals of the sensors are faultless and thus are reliable enough to portray the inner state. Faultlessness of the sensors means that there is no drift, no bias or a permanent or interrupted malfunction. To guarantee reliable operation of the monitoring system, often redundant sensor systems are used, which is called hardware redundancy. Since the sensor signals of complex systems do not depend only on one internal state of the system, the inner correlation of the sensor output signals can be used to replace the hardware redundancy by software redundancy, which is called analytical redundancy. Using these correlations, faults of the sensors and faults of the system can be separated. In the paper, methods for the detection, localisation and quantification of the different faults are presented. As an example, t he method for fault detection presented in this paper is applied to a steamheater used in power plants.