Noise diagnostics of Stationary and Non-Stationary Reactor Processes

Doctoral thesis, 2007

This thesis concerns the application of noise diagnostics on different problems in the area of reactor physics involving both stationary and non-stationary core processes. Five different problems are treated, divided into three different parts. The first problem treated in the first part is the classification of two-phase flow regimes from neutron radiographic and visible light images with a neuro-wavelet algorithm. The algorithm consists of wavelet pre-processing and of an artificial neural network. The result indicates that the wavelet pre-processing is improving the training of the neural network. Next, detector tubes which are suspected of impacting on nearby fuel-assemblies in a boiling water reactor (BWR) are identified by both a classical spectral method and wavelet-based methods. It was found that there is good agreement between the different methods as well as with visual inspections of detector tube and fuel assembly damage made during the outage at the plant. The third problem addresses the determination of the decay ratio of a BWR from the auto-correlation function (ACF). Here wavelets are used, with some success, both for de-trending and de-nosing of the ACF and also for direct estimation of the decay ratio from the ACF. The second part deals with the analysis of beam-mode and shell-mode core-barrel vibrations in pressurised water reactors (PWRs). The beam-mode vibrations are analysed by using parameters of the vibration peaks, in spectra from ex core detectors. A trend analysis of the peak amplitude shows that the peak amplitude is changing during the fuel cycle. When it comes to the analysis of the shell-mode vibration, 1-D analytical and numerical calculations are performed in order to calculate the neutron noise induced in the core. The two calculations are in agreement and show that a large local noise component is present in the core which could be used to classify the shell-mode vibrations. However, a measurement made in the PWR Ringhals-3 shows that it is impossible to use this local component due to practical reasons, i.e. due to the limitations of the possible detector positions. In the third and final part the validity of the break frequency method for the estimation of the reactivity in a subcritical system is investigated. A numerical calculation of the neutron noise in a 2-D subcritical system corresponding to a power reactor during loading conditions is performed. From the auto power spectral density of the numerically calculated neutron noise, the reactivity is estimated by the break-frequency method. The result indicates that the break-frequency method yields only a rough, although conservative, 64estimate of the reactivity.