A methodology for the reduction of numerical diffusion in sloshing analyses through automated mesh adaptation

The paper proposes a methodology to improve the accuracy of Volume of Fluid (VoF) multiphase problems involving liquid/gas sloshing in fuel/lubricant tanks without penalizing the computational cost of the simulations. In order to correctly track the complex trajectory of the liquid/gas interface and the presence of liquid droplets in the gas phase, the VoF method requires a fine mesh at each interface location to reduce modeling errors. The investigated case is a lubricant tank of a sport car subject to typical race track maneuvers. Due to the geometrical extent and the complexity of the computational domain and to the relevant accelerations, resulting in dispersed liquid structures within the gas phase, the use of a generalized fine mesh would result in computational costs far beyond the industrial practice. A methodology is then proposed to reduce the overall number of computational cells through a combination of local interface tracking and mesh refinement, which is combined with an active control of the time step to comply with Courant-Friedrichs- Lewy number limits. The methodology is at first validated against experimental measurements for a simplified test case, and then applied to the actual oil tank sloshing case, showing a relevant reduction of the numerical diffusion and a consequent higher accuracy.