波浪中船体縦曲げ荷重の極値確率分布に関する研究

Abstract

Statistical analysis of the extreme wave-induced hull girder vertical bending moments in short-term sea states is conducted. A series of time-domain numerical simulations of the wave-induced hull girder vertical bending moments are performed based on nonlinear strip method to obtain peak values of the vertical bending moments. The Peaks-Over-Threshold (POT) method is applied to detect peaks exceeding thresholds, i.e., 60%, 80%, 90%, 95%, 97.5%, 98.5% and 99.5% of percentile values, then the peaks are fitted to generalized Pareto distributions. The extreme value in N peaks is estimated by using order statistics based on the generalized Pareto distributions. The simulation results are also fitted to Weilbul distributions and the extreme value is estimated for comparisons. It is found that the POT method gives good estimations of the extreme value in general. Fitting to Weibull distribution is advantageous in a point that it contains smaller statistical uncertainties, especially when the goodness-of-fit is high. It is shown that the POT method gives better estimations on the extreme values when strong nonlinear effects such as whipping are extremely significant.