Dynamic mode decomposition for the stability analysis of the Molten Salt Fast Reactor core

•Dynamic Mode Decomposition (DMD) is employed to analyse a fast MSFR transient.•A previously developed OpenFOAM multiphysics model is used for full order simulation.•DMD model is used for stability and dynamical analysis of the original MSFR model.•A preliminary sensitivity analysis is performed to assess DMD method capabilities. The study of innovative nuclear reactors involves the use of increasingly complex numerical models. While such models provide a high-fidelity description of many non-linear coupled phenomena, they are not suited for many-query tasks such as design optimisation, uncertainty quantification, stability analysis or parameter identification due to the required computational effort. For this reason, a variety of techniques have been employed to reduce the complexity and surrogate the response of large nuclear systems. One example is the dynamic mode decomposition (DMD), a data-driven method which builds a low-dimensional eigenvalue-eigenvector representation of the underlying model from numerical data, and allows for non-intrusive analyses of the dynamical properties of the system without knowledge of the model itself. In this work, DMD is applied to the study of a free-dynamics fast transient of the Molten Salt Fast Reactor (MSFR), following a variation of the heat transfer coefficient. The numerical data is provided by a multiphysics model developed using the open-source CFD toolkit OpenFOAM. The aim of this work is to demonstrate the applicability of DMD to the study of large next-generation nuclear systems such as the MSFR. The results show the capabilities of DMD to extract and surrogate the dynamics of the MSFR following perturbation, including the initial non-linear dynamics and the final steady-state. Different values of parameters relevant to the construction of DMD models are tested, to provide some insights on the sensitivity of the method to the selection of the numerical data set and to the size of the reduced model.