Probabilistic Analysis of Characteristic Pressure for LNG Tanks

Background. New challenges in LNG shipping, such as ship size growth, trading routes with more severe weather conditions, need for operating with unrestricted filling level and new propulsion systems attract very much attention in marine and offshore oriented community. One of the main concerns is the prediction of loads caused by violent fluid motion in cargo tanks. In the paper we address the problem of determining characteristic extreme values of sloshing pressures for structural design. This involves estimating ship motion in a long-term period, fluid motion in the tank, excited pressures, and relevant structural responses. Method of Approach. Ship motion analysis is based on linear strip theory. In order to investigate the dependence of the sloshing response on sea conditions, an approach based on statistical characteristics of the tank motion is utilized as well as a multimodal approach for fluid motion in a tank. However, an appropriate theoretical/numerical approach, which can be used for a realistic prediction of the most extreme pressure has not yet been developed. Thus, experiments are utilized for the most severe sea states for a chosen tank filling level. Results. Our main contribution in the paper includes the statistical analysis of experimental short term pressure distribution. The choice and fit of probability distribution models is addressed, with due account of different physical mechanisms causing impacts. The models are evaluated. The most critical tank areas for sloshing loads are briefly discussed. Appropriate dynamic response of the tank structure needs to be investigated by accounting for temporal and spatial distribution of sloshing loads. These two factors are also addressed in the paper. The variability of results obtained by processing data from multiple test runs is discussed. Conclusions. The three-parameter Weibull and generalized Pareto statistical models are fitted to the data and evaluated. They prove to accurately describe sloshing excited pressures. However, the highest data points are underestimated by the both distributions. Generalized Pareto model results in more conservative estimates. The threshold level of peaks used in a fit of generalized Pareto distribution was investigated. Based on this, it is set to a level of a 0.85–0.87 quantile of peak values. The big influence of spatial and temporal distribution on the estimates is reported. Uncertainty in measured pressures originating from inherent fluid motion variability exceeds the uncertainty resulting form ship motions' variability. Moreover, generalized Pareto model results in higher variability.