Estimating Characteristic Load Effects on Floating Structures in Ice

The present paper presents research results from the IceStruct Joint Industry Project (Det Norske Veritas, 2012), in which a simplified methodology for determining characteristic ice load effects on floating and fixed structures was developed. The simplified methodology was calibrated to the results of a probabilistic model and the work presented herein describes how the characteristic mooring line tension for a spar located in an environment with drifting level ice and ice ridges was determined. The probabilistic model includes a description of several stochastic variables (ice ridge keel depth, consolidated layer thickness, parameters describing the strength of the keel etc.) where some variables are dependent on each other. A computational response model of a spar structure was used to determine the maximum mooring line tension during each ice ridge interaction event. Estimating the characteristic mooring line tension involves estimation of low probability levels of a multivariate stochastic system. Direct Monte Carlo simulation is usually not applicable to estimate such low probability levels as it would require a lot of simulation time. When solving similar problems for the design of floating structures in waves, response surfaces are often applied, and this methodology was also selected for the current work. In this example, six governing stochastic variables were identified through sensitivity studies, resulting in a response surface that was embedded in a seven-dimensional space. This response surface was approximated and represented by a Beziér surface, which was obtained by evaluating the computational response model for a number of points on a predefined grid in the variable space. The Beziér surface is well suited for representing response surfaces as it is fast to evaluate and does not introduce artificial fluctuations on the surface. The characteristic mooring line tension (corresponding to a given annual probability of being exceeded) was then estimated by a direct Monte Carlo simulation on the Beziér surface.