Reconstruction and prediction of rising bubble by Lagrange DMD in data-driven

Abstract

Dynamic mode decomposition (DMD) is a widely used data-driven modeling method for understanding complex flow systems, however, it is inadequate to deal with translation problems such as bubble rising. The Lagrangian DMD formed by the DMD combined with translation information has shown preliminary promise for solving simple translation problems (convection dominated). This paper aims to use a data-driven method to predict the 2D bubble rising process and try to improve the Lagrangian DMD modeling ability. In the simple case without bubble deformation, Lagrangian DMD shows significantly better reconstruction and prediction results than DMD. However, in a complex case with bubble deformation, Lagrangian DMD cannot provide satisfactory prediction results. After adding velocity prediction to the Lagrangian DMD method, the prediction of bubble rising has been improved.