Multi-year (MY) ridge and level ice interactions with sloping and conical structures involve complex ice feature shapes and ice failure mechanisms. The limited available field data makes calibration of associated load models difficult. To account for associated randomness and uncertainty, models may tend to be on the conservative side.
New deterministic algorithms were recently developed to calculate loads more accurately for interactions of MY level ice and MY ridges with an upward sloping structure. This paper presents the application of these recently developed formulations in a probabilistic framework using SILS. SILS is a Monte-Carlo type simulator developed by C-CORE following the general procedures outlined in ISO 19906. Ice and metocean input parameters are defined by the user either as a fixed value (e.g. friction coefficients) or a random distribution (e.g. ice drift speed, floe size). The yearly encounter frequency is first estimated for these ice features for the site of interest. The ice loads are then determined for each of simulated interaction event occurring over the model timespan, using the deterministic load formulations. By simulating a large number of years of ice interactions, design ice loads can be determined that correspond to various low annual probability of exceedances.
This paper demonstrates how complex loading scenarios, modelled in terms of idealized deterministic models, can be incorporated within a Monte-Carlo framework to provide design level loads. During the model implementation and analysis of results, significant improvements were identified and implemented in the deterministic model, resulting in a more robust model and better design estimates. The results provide valuable insights regarding model inputs and behaviour corresponding to the extreme design ice loads. An example of a full probabilistic analysis is presented in the paper to illustrate the models. Here the probabilistic framework of SILS is used to assess a Base Case scenario and a number of sensitivity cases using different environmental inputs and model parameters.