In this study, numerical studies on the sloshing phenomena in two- and three-dimensional rectangular tanks were carried out to predict the impact loads inside the rectangular tanks by Eulerian and Lagrangian approaches. The two-phase Eulerian solver is based on the CIP/CCUP (Constraint Interpolation CIP/CIP Combined Unified Procedure) method and THINC-WLIC(Tangent Hyperbola for Interface Capturing-Weighted Line Interface Calculation) method is used to capture the air-water interface. The Lagrangian one is an improved version of MPS(Moving Particle Simulation) method, which is named as PNU-MPS(Pusan National University modified MPS). These two methods were applied to the sloshing problems and the characteristics between the methods were compared each other through the numerical simulations. The results of simulations were also compared with those of experiments and the overall comparisons were reasonably good. In general, Lagrangian approach showed better results on predicting the impact pressure acting on the structure than Eulerian one.
It has long been recognized that local impact loads due to sloshing can be accurately predicted only if the non-linear free-surface motion inside the tank is correctly simulated. There are several techniques to handle such problems, including SOLA-VOF (Volume Of Fluid) (Hirt and Nichols, 1981), Level-Set (Sussman, Emad, Peter, and Stanley, 1994), Marker-Density Function (MDF) (Miyata, and Park, 1995) and so on. Most of them are the techniques which capture the free-surface on the grid system and have been widely used. When using Eulerian gird-based methods, however, the convection terms are appeared in the governing equation and needed to be dealt carefully, since the scheme to handle them affect the accuracy and stability of the simulation very much. For both accuracy and stability, some up-winding manners with higher order have been generally employed as the scheme for the convection terms. However, the conventional up-winding schemes including artificial diffusion added into the central differencing scheme can cause often render decisive influences on the simulation results.
