Dynamic mode decomposition for data-driven modeling of free surface sloshing

Abstract

The recently emerged data-driven dynamic mode decomposition (DMD) method was employed to investigate the free surface sloshing dynamics of a partially filled rigid tank excited by horizontal harmonic motions. The volume of fluid algorithm was adopted for liquid–gas free surface tracking, and DMD was utilized for decomposition with physical interpretations of the DMD modes and the eigenvalues. Our results demonstrate that DMD works well for both data reconstruction and future-state prediction in terms of either free surface profiles or the sloshing pressure exerted on the rigid wall. DMD presents an efficient and versatile approach for accelerated reduced-order modeling and future-state forecasting. Our efforts provide the first reference on use of DMD for free surface sloshing problems to the best knowledge of the authors. We publicly share our data and codes for all the implementation.