Fast prediction of the flow induced vibration of wire-wrapped fuel rod in a lead-bismuth reactor based on improved DMD and LSTM network

Abstract

Flow induced vibration of wire-wrapped fuel rod in lead-bismuth reactor has become an increasing concern due to its significant influence on cladding fretting wear and reactor operation. In this study, we proposed a reduced order model to expedite the prediction of wire-wrapped fuel rod vibration. Firstly, the predefined structural motion is implemented in the fluid solver using a user defined function, while the fluid forces are computed. Subsequently, both the motion and fluid forces are decomposed into mode functions and time coefficients by dynamic mode decomposition. A long short-term memory network is then trained to bridge the time coefficients of motion and fluid forces. Finally, the vibration response of wire-wrapped fuel rod under different conditions is efficiently calculated using the Newmark integration scheme. The results demonstrate that 18 modes are sufficient to reconstruct the fluid forces, significantly reducing computation complexity, while the long short-term memory network provides reasonable fit to the original data. In addition, the study reveals that the motion of wire-wrapped fuel rod can be chaotic when the axial flow velocity exceeds 2.0 m/s. The mechanism of chaos is attributed to period-doubling bifurcation, a phenomenon not observed in prior simulations. Moreover, the asymmetric geometry of wire-wrapped fuel rod consistently leads to ‘elliptic’ trajectory, and the root mean square values of vibration amplitude can be fitted by the power function with respect to axial flow velocity. The proposed method offers a valuable and convenient tool to study the flow induced vibration of wire-wrapped fuel rod with significant efficiency improvement.