Probabilistic assessment of first-year ice ridge action on offshore structures

Abstract

As defined by the international standard ISO 19906, the representative ice actions should be estimated for ELIE (extreme-level ice event) and ALIE (abnormal-level ice event). These events are defined by their relevant annual exceedance probability levels. Probabilistic methods are often used to obtain a proper characterization of the ice actions from which representative ice actions can be inferred. In this paper, we consider ice actions caused by first-year ice ridges. Ridge action depends on a large number of input parameters. Important correlations between the parameters add to the complexity of ridge action and need to be included in probabilistic assessments to obtain reliable results. Data for establishing input probability distributions are often incomplete, biased, or completely non-existent. To solve this problem, it is common engineering practice to combine data from locations other than the location of interest and make ad hoc extrapolations and assumptions. This often leads to overly conservative estimates of representative ice ridge actions. We propose a framework for Monte Carlo simulation of the first-year ice ridge actions. The goal is to establish a method for which the need of input data is at a minimum. The only needed input data in our simulation is the statistics of the annual maximum level ice thickness and statistics related to ice being present or not. Based on correlations and findings from our previous studies, we are able to simulate other input parameters such as ridge keel draft, ridge frequency and consolidated layer thickness. In this paper, we also discuss the problem of defining the probability distribution for the ice strength coefficient C_R in the context of Monte Carlo simulations. Our approach is dependent on our previous research that is based on data from the Beaufort Sea. Without an appropriate calibration of the correlations between the parameters, we cannot be certain if the simulation can be extended to other locations by simply adjusting the input maximum annual level ice statistics. Nevertheless, we believe that our approach can be used as a tool for preliminary probabilistic assessment of ridge action. Our approach offers good flexibility, and we believe that with suitable data it can be calibrated for other geographical locations and structure types.