Modelling and Simulations of Neutron Noise Experiments

Abstract

This thesis focuses on neutron noise, which can be used to detect anomalies in the core of light-water nuclear reactors. Neutron noise refers to the small, stationary fluctuations of the neutron flux around its nominal value. These fluctuations arise from processes in the core, and can be monitored and analysed through a technique called neutron noise diagnostics. By identifying deviations in the neutron noise, anomalies that may lead to safety issues can be detected at an early stage, allowing prompt actions to be taken to prevent potential accidents. This work involved modelling neutron noise experiments performed in the CROCUS research reactor (EPFL) using the diffusion based neutron noise solver CORE SIM+. The reactor’s response to two different noise sources was simulated: a laterally vibrating group of fuel rods and a vertically vibrating absorber rod. The former has been modelled and verified in previous studies, whereas the latter has not previously been studied using CORE SIM+. Given its novelty in this context, several modelling approaches were explored for the vertically vibrating absorber to gain insight into how this type of perturbation can be represented. In addition, the study evaluated two computational approaches for computing the noise: the direct and adjoint methods. The influence of the geometric dimensionality was also assessed by performing simulations of the vibrating fuel rods in both 2D and 3D. The resulting simulated noise fields were then compared to the experimentally measured noise to assess the performance of the simulations. The simulations of the vibrating group of fuel rods resulted in a noise field with both a notable point-kinetic component, and a strong spatial component near the perturbation. The simulations of the vibrating absorber produced noise fields with highly localised spatial effects in the vicinity of the perturbation and a negligible point-kinetic contribution. When combined, features originating from both sources appeared in the simulated noise. The direct and adjoint approaches yielded consistent results across all simulations, and the 2D- and 3D-simulations of the vibrating fuel rods gave essentially identical outcomes. The solver accurately reproduces the neutron noise measured at detector positions farther from the sources where the point-kinetic component dominates, whereas near the sources where the spatial effects are significant the deviations from measurements are more pronounced. These discrepancies are partly linked to the underlying approximations of the solver, leading to inaccurate predictions of the spatial gradients near a perturbation. For the case of the vibrating absorber, these discrepancies also arise because the models used are not fully representative of the physics of the noise source. This calls for further efforts to improve the models describing the vertically vibrating absorber.