Comparison of two adaptive identification methods for monitoring and diagnosis of an experimental nuclear reactor

This paper deals with the comparison of two adaptive methods based upon sensitivity equations for use in the surveillance and diagnosis of an experimental nuclear reactor. The surveillance and diagnosis are obtained by a real-time comparison of reference parameters and the actual parameters given by the adaptive algorithm. The first algorithm uses an on-line, steepest descent method. Results obtained with this algorithm using experimental data from a reactor are given using two different criteria. The second algorithm uses both sensitivity equations and a recursive least squares method. An example is given using the experimental model of the same reactor. Both algorithms described in this paper are easily implementable on a mini computer and are not sensitive to a priori knowledge of the statistical properties of the noise. These algorithms are also suitable for the surveillance of nonlinear processes.