Several features of an industrial sloshing model test are presented. Data mining is conducted for an experimental database at Seoul National University. 3,000 different test cases have been fetched from the database. The fetched data are 15,000 hours of 6DoF irregular model tests with various cargo hold dimensions, floating units, environmental conditions, and operational conditions in 1/50 – o1/30 scales. The severity of the model test based on the impulse peak pressures is described. A correlation between 6DoF motions and the severities are illustrated. Dynamic characteristics of sloshing events are discussed by unsupervised machine learning techniques.


When the LNG tanks of the floating units are designed, it is important to analyze possible sloshing loads. Since the LNG transportation market has been continuously expanded (BP, 2018), numerous LNG carriers have been constructed. In addition to the LNG carriers, a lot of LNG fueled ships have been built because emission regulations of greenhouse gases have been tightened. Moreover, owing to the increasing size of the floating units (Kuo et al., 2009; Malenica et al., 2017), accessing the sloshing loads becomes more significant.

Because of the highly nonlinear and complex nature, the sloshing impacts are commonly analyzed by conducting a model test (Malenica et al. 2017) in a 1/50 or 1/40 scale. ITTC and the Classification societies suggest several guidelines of the test (ABS, 2009; BV, 2011a; 2011b; DNV-GL, 2006; ITTC, 2014; LR, 2009), and numerous sloshing model test have been conducted as an industrial project (Gavory and de Seze, 2009; Malenica et al., 2017, Ryu et al., 2016). Seoul National University (SNU) has carried out parts of these industrial sloshing model tests and has created an experimental database from those model tests (Ahn et al., 2018; Lee et al., 2018; Kim et al., 2017; Oh et al., 2015; Park et al., 2009; 2014).

The present paper presents three parts of an industrial sloshing model test: a severity of sloshing impact peak pressure, a relation between 6DoF motions and the peak pressures, and dynamic characteristics of the sloshing events. First, an industrial sloshing model test is briefly introduced. A database of the industrial sloshing model test in SNU is described. Second, the severity, which is analyzed based on the impulse peak pressure, of the database is presented. The severities on the location are also presented by the faces of the cargo hold models. Last, several data mining techniques are used to analyze the dynamic characteristics of the sloshing events, which occur during the industrial sloshing model test. Rise times and rise impulse areas are mainly discussed. Clustering techniques for the data are illustrated, and the results are discussed.

You can access this article if you purchase or spend a download.