Combining simulations and machine learning for neutron noise-based core diagnostics

Core monitoring techniques represent methods that allow detecting anomalies in nuclear reactors, subsequently characterizing those anomalies, localizing them (if relevant), and classifying them according to their impact on plant safety and availability. One of the most promising core monitoring techniques relies on the measurement of the inherent fluctuations in the neutron flux in a nuclear core, the so-called neutron noise. Those fluctuations are the results of perturbations existing in the system, such as mechanical vibrations and perturbations in the cooling medium of the reactor, to name a few. Such perturbations, by modifying the probability of occurrence of the corresponding nuclear reactions in the core, will give rise to fluctuations in the neutron flux. The monitoring of such fluctuations thus makes it possible to detect any abnormal behaviour of the system. The main advantage of using neutrons to detect anomalies as compared to using temperature, pressure or flow rate information is that neutrons propagate through the system via the fission and scattering reactions. Neutron detectors can thus “sense” any perturbation even far away from the actual location of the perturbation.