A complete sloshing assessment within tanks of a membrane LNG vessel is far beyond the state-of-the-art capabilities of numerical simulations. Therefore, the only tools available today for such an assessment remain model tests at small scale. Nevertheless one must keep in mind that these tests represent an experimental modelling of the reality. Through a parametric study based on numerical simulations of academic liquid impacts, this paper shows that for a perfect similarity between full scale and model scale, not only the excitation motions must be Froude scaled and the Density Ratio between gas and liquid must be kept equal, but also the speeds of sound in both gas and liquid must be Froude scaled. With these constraints, the different phenomena involved locally during a sloshing impact, including the compressibility effects, are balanced in the same way at both scales, regardless of the impact conditions. Therefore, the impact pressures can be strictly derived at full scale from model scale values by Froude scaling. This ideal situation is not realistic as the speeds of sound required for the gas and the liquid at scale 1:40 would be respectively 45 m/s and 270 m/s. After analysing the influence of each parameter involved in the problem, the study concludes that adopting model tests with water and a mixture of Sulfur Hexafluoride (SF6) and Nitrogen (N2) in the right proportion to match the Density Ratio of LNG in equilibrium with its vapour is a good trade-off for the time being. It improves significantly the representativeness of model tests with regard to the solution with water and air.
The different methodologies for sloshing assessment (e;g. Gervaise, de Sèze and Maillard, 2009) rely on the statistical post-processing of the peak pressures obtained during sloshing model tests, even though these peaks are of very short duration (usually smaller than 1 ms at scale 1:40).
