Machine learning powered predictive modelling of complex residual stress in laser-welded fusion reactor steel Eurofer97
engineering components. Within the nuclear fusion energy industry, a remote laser welding technique will be extensively employed for the assembly and maintenance of in-vessel components, predominantly constructed from Eurofer97 steel. The welding process introduces significant residual stresses, which interacts with the microstructure, resulting in mechanical property degradation and a shortened lifetime. In response to these challenges, a predictive model, utilising machine learning, was firstly developed for the localised omnidirectional residual stress prediction at a finer scale. The model captures intricate details of material behaviour and the interplay between residual stresses, microstructural features, and mechanical properties. The successful framework establishes a solid foundation for predicting residual stresses, and can be further developed by its nature of highly adaptable to a broad range of the joints and material systems in the nuclear fusion energy industry with the particular ML/DL algorithms and features, thus enhancing the reliability of key in-vessel engineering components.