Investigation of a discrete ordinates method for neutron noise simulations in the frequency domain

Doktorsavhandling, 2021

During normal operations of a nuclear reactor, neutron flux measurements show small fluctuations around mean values, the so-called neutron noise. These fluctuations may be driven by a variety of perturbations, e.g., mechanical vibrations of core components. From the analysis of the neutron noise, anomalous patterns can be identified at an early stage and corrected before they escalate. For this purpose, the modelling of the reactor transfer function, which describes the core response to a possible perturbation and is based on the neutron transport equation, is often required. In this thesis a discrete ordinate method is investigated to solve the neutron noise transport equation in the frequency domain. When applying the method, two main issues need to be considered carefully, i.e., the performance of the numerical algorithm and possible numerical artifacts arising from the discretization of the equation. For an efficient numerical scheme, acceleration techniques are tested, namely, the synthetic diffusion acceleration and various forms of the coarse mesh finite difference method. To reduce the possible numerical artifacts, the impact of the order of discrete ordinates and the use of a fictitious source method are studied. These analyses serve to develop the higher-order neutron noise solver NOISE-SN. The solver is compared with different solvers and used to simulate neutron noise experiments carried out in the research reactor CROCUS (at EPFL). The solver NOISE-SN is shown to provide results that are consistent with the results obtained from other higher-order codes and can reproduce features observed in neutron noise experiments.