An AR Model Spectral Estimation for the Monitoring of Reactor Internal Vibration

摘要

A reactor vibration monitoring has been performed using neutron noise obtained from excore detectors for safety operation. Traditionally, the spectral estimator based on Fourier analysis has been widely used in the noise analysis of the reactor system. If the bias is too severe, the resolution will not be adequate for a given application. One major motivation for the current interest in the parametric approach to spectral estimation is the apparent higher resolution achievable with these modern techniques. In considering an unbias, a consistency, an efficency, and a minimum lower bound of the statictic estimation , an AR model is appropriate for noise spectral estimation with sharp peaks, but not deep valleys. In order to select an appropriate model order, the lag value of an autocorrleaton function is applied. The Burg method to trace the mode of RPV internal structures is the most successful.