| Summary: | Fusion power is generated when hot deuterium and tritium nuclei react, producing alpha particles and 14 MeV neutrons. These neutrons escape the reaction plasma and are absorbed by the surrounding material structure of the plant, transferring the heat of the reaction to an external cooling circuit. In such high-energy neutron irradiation environments, extensive atomic displacement damage and transmutation production of helium affect the mechanical properties of materials. Among these effects are irradiation hardening, embrittlement, and macroscopic swelling due to the formation of voids within the material. To aid understanding of these effects, Bayesian neural networks were used to model irradiation hardening and embrittlement of a set of candidate alloys, reduced-activation ferritic-martensitic steels. The models have been compared to other methods, and it is demonstrated that a neural network approach to modelling the properties of irradiated steels provides a useful tool in the future engineering of fusion materials, and for the first time, predictions are made on irradiated property changes based on the full range of available experimental parameters rather than a simplified model. In addition, the models are used to calculate optimised compositions for potential fusion alloys. Recommendations on the most fruitful ways of designing future experiments have also been made. In addition, a classical nucleation theory approach was taken to modelling the incubation and nucleation of irradiation-induced voids in these steels, with a view to minimising this undesirable phenomenon in candidate materials. Using these models, recommendations are made with regards to the engineering of future reduced-activation steels for fusion applications, and further research opportunities presented by the work are reviewed.
|
|---|
