Applying analytical redundancy to safety-relevant data acquisition in nuclear power plant

The modelling study reported here was carried out for examination and testing of four instrumental fault detection methods within the framwork of a non-linear steam generator model. It is shown that the introduction of a time shift between components of the Kalman filter innovation for generation of the covariance matrix can lead to correspondingly delayed error signals, which adds arguments against the suitability of this method. A main point of interest of the project is the question of whether model non-linearities are necessarily to be taken into account in the filter equations. Two alternative methods have been tested and are suggested. These use robust detection filters or state observers and the Fourier analysis. The results obtained are positive, so that there remain three recommendable methods, although it is not possible to carry out a direct comparison of the methods. (orig./DG)