Predicting the onset of void swelling in irradiated metals with machine learning

Radiation-induced void swelling is a serious mode of degradation in nuclear structural materials. Much effort has been spent to predict swelling resistance, with the goal of increasing the void swelling incubation dose so as to postpone the consequences of radiation damage. However, this trial-and-error approach is highly inefficient due to the time- and resource-intensive nature of both experiments and physics-based multiscale simulations. In this work, as a first attempt, machine learning is applied to perform this prediction based on available experimental data. Of the multiple techniques applied, the gradient boosting ensemble method best predicts experimental onset doses for swelling in test datasets, and identifies the main contributing factors such as temperature, Fe and Cr content, and dose rate, which are consistent with established understanding. This work demonstrates the feasibility of machine learning to predict macroscale radiation effects based on material and environmental parameters, and has practical significance in guiding further material optimization for nuclear applications.