Construction of an hybrid Artificial Neural Network - Fuzzy Logic system to perform Atomistic Kinetic Monte Carlo simulations in iron-copper alloys

Neutron-irradiation enhanced copper precipitation in iron is one of the major causes of the shift in the ductile-to-brittle transition temperature in reactor pressure vessel steels. This process has been studied for years by computer simulation. Molecular dynamics is used to study atomic collision cascades induced by primary knock-on atoms, whereas Monte Carlo simulations are used to model thermal diffusion of created microscopic defects. This thesis aims at building an improved Atomistic Kinetic Monte Carlo (AKMC) simulation to study the thermal diffusion of vacancies and the induced copper precipitation. The vacancy migration energies are calculated with the aid of an Artificial Neural Network, trained with a limited set of Molecular Dynamics evaluated examples. A Fuzzy Logic feedback is constructed to reduce the mean error committed. The improved AKMC algorithm is validated and a series of AKMC simulations are performed to study the evolution of the copper solubility limit in iron with temperature.