Analysis and Probabilistic Modeling of the Unstationary Ice Loads Stochastic Process, Based on Experiments With Models of Offshore Structures

Abstract

Experiments with models of platforms and offshore structures with vertical and inclined panels, which were conducted at Krylov Research Center (St. Petersburg), demonstrated that sometimes ice loads time series registered in these experiments cannot be considered as stationary. At the same time until nowadays methods and algorithms of probabilistic modeling were mainly based on the assumption of ice loads time series stationarity. That is because the analysis and modeling for stationary stochastic process is easier than for those unstationary. In the paper the method for determining the presence of unstationarity in ice loads time series, based on statistical analysis, is described. This method employs sample mean normality. Fuzzy C-means algorithm is used to cluster autocorrelation vectors, which are built for different fragments of time series. In the paper ice loads time series, got in experiments in ice tank with offshore structure columns and basement models, are investigated on their unstationarity. The algorithm of unstationary ice loads time series simulation is offered.