Research on a rapid prediction method for multi-physics coupled fields in small lead-cooled fast reactors based on machine learning

Abstract

The small lead-cooled fast reactor (LFR), as a typical representative of the fourth-generation reactors, is widely applicable to island areas, deep-sea environments, and specialized industrial scenarios due to its high power density, modular design, and inherent safety features. The neutron physics, thermal-hydraulics, and structural deformation in LFR are highly coupled, and most existing studies neglect the impact of fuel deformation on the neutron physics and thermal-hydraulics, making it difficult to accurately reflect the operating state in the reactor. However, the coupling analysis of the three fields involves significant computational costs, making it challenging to achieve real-time prediction. To address this issue, this paper proposes a method that integrates the multi-physics field coupling of reactors with machine learning-based rapid prediction techniques. By using measurable parameters of the reactor, rapid prediction of the multi-physics field distribution inside the core can be achieved. The final test results show that the model controls the relative prediction error of each physical field within 1%, with prediction time significantly shortened compared to traditional numerical methods. It efficiently achieves accurate and rapid prediction of the multi-physics field distribution in the LFR.