To evaluate the sloshing loads from the peak pressures of the model test, a proper statistical model and its parameter estimator should be properly selected. The three–parameter Weibull distribution with 5 different estimators and the generalized Pareto distribution with 3 different estimators have been studied based on the 2,321 model test cases. Three different goodness–of–fitness tests have been used to test their fitness. Expected results of each statistical model are described and possible selection of the statistical model is suggested to obtain well–fitted distribution to the sloshing peak samples.
Assessing sloshing loads on the LNG vessels has been an important issue of the shipbuilding industry (Kuo et al., 2009; Royal Dutch Shell, 2020). Owing to the complex and stochastic nature of sloshing, conducting a model test has been the most common approach to evaluate the sloshing loads (Gervaise et al., 2009; Karimi et al., 2013; 2014; Malenica et al., 2017). Extensive studies have been conducted to develop a proper methodology of the model test (ABS, 2009; BV, 2011; DNV·GL, 2016; Karimi et al., 2013; 2014; Kuo et al., 2009; Graczyk and Moan, 2008). Korean shipyards have carried out numerous model tests for their products and cargo hold development following these suggested procedures. Seoul National University has supported to these model tests greatly commissioned by the industries and a huge database of the sloshing model test has been created (Ahn et al., 2019; Kim et al., 2013; Kwon et al., 2018; Oh et al., 2015).
One of the issues in conducting sloshing model test is to find a proper statistical model. Several extreme statistics have been described in the international guidance notes and a few studies (ABS, 2009; BV, 2011; DNV·GL, 2016; Kim et al., 2014; Graczyk and Moan, 2008), but no concrete solution has been made. The most commonly used statistical models are the three–parameter Weibull distribution (WE3) and the generalized Pareto distribution (GPD) (DNV·GL, 2016; Kim et al., 2014; Graczyk and Moan, 2008). As a parameter estimator, the method of moments has been a typical choice for two statistical models. However, the sloshing loads hugely vary according to the parameter estimators and the statistical models.