A new method for on-line surveillance of nuclear power reactors based on decision theory

In this paper a new method for the automatic monitoring of reactor operational states based upon decision theory is introduced. The observation space is represented by spectral density functions estimated in discontinuous time intervals. After an empirical mathematical modelling of the process and its hierarchical structuring, a combined identification, estimation and detection algorithm is described. The classification of each subsystem is given via a sequential hypothesis test. Furthermore, the possibility of an adaptive learning or training is demonstrated. After a logical, system specific association of the resulting pattern, a diagnostic prognosis about the system state is made which contains integral information about a time interval extending in the past. The implementation of this method is illustrated by an example and the preliminary results are presented. (Auth.)