As part of the Horizon 2020 CORTEX project led by Chalmers, and building upon some earlier successful tests on the applicability of using machine learning for anomaly characterization (reported at two conferences – see here and here), Chalmers has just delivered to the CORTEX partners a unique set of simulation data. The simulations cover an extremely wide range of possible scenarios of anomalies, such as vibrations of fuel assemblies, core barrel vibrations, control rod vibrations, perturbations generated at the inlet of the core and propagating with the coolant, and perturbations generated inside the core. All the simulations were performed using the CORE SIM+ neutronic core simulator (an extension of the CORE SIM tool), developed as part of CORTEX by Dr. Antonios Mylonakis.
The simulations cover all possible frequencies of the aforementioned scenarios and all possible locations of the perturbations. They represent more than 3 Tb of data. The machine learning experts from the University of Lincoln will now, together with Chalmers researchers, extend the machine learning techniques earlier developed so that the different scenarios can be identified from very few in-core and ex-core neutron detector signals, the anomaly characterized and localized. It is the first time ever that such a large set of simulation data representative of a commercial reactor and corresponding to all possible scenarios of stationary fluctuations were created. With this delivery, the CORTEX project has reached a new milestone.
