Structural materials in nuclear reactors operate under extreme conditions, including high temperatures, intense irradiation, and corrosive environments. These conditions drive complex microstructural evolution and degradation of mechanical properties. Accurate modelling of these behaviours is essential for predicting component lifetimes and ensuring reactor safety. In recent years, machine learning (ML) has attracted growing interest as a powerful tool for accelerating material modelling and capturing complex environmental effects by exploiting large datasets generated from experiments and simulations. This review critically evaluates the application of ML techniques to the modelling of nuclear structural materials and provides readers with a comprehensive guide to their use in this field. We begin with a conceptual overview of machine learning in material modelling, and highlight key modelling dimensions and representative ML architectures that underpin current applications. The existing literature is then analysed by grouping studies according to material class, while the ML methods employed and their respective application purposes are presented in detail. Finally, ML applications are presented and discussed in relation to the extreme environmental conditions encountered in nuclear systems, followed by a critical evaluation of current challenges and future directions.
Tasdemir et al. (Fri,) studied this question.