Application of extreme value theory for the assessment of turret moored floating structures

Abstract

In the design of mooring systems, it is a common practice to use a 100-year design environment to calculate extreme responses. Statistical inference is executed on the environmental data to produce a 100-year environment. This 100-year environment is then simulated to calculate the loads that occur when the vessel is under the influence of the 100-year environment. An alternative for this method is response based design. In response based design, measurements with a 3-hour interval of the environment over a long period of time are used to simulate the behaviour of the FPSO. The simulation provides a data-set of dominant loads for the mooring system over the time period. With extreme value theory the tail of this data is fitted on a generalized Pareto distribution. With this distribution a 100-year extrapolation can be made that results in the 100-year extreme loads. This could result in a more realistic value of the extrapolated loads in comparison with the conventional method. In (Oostra 2015) this method is applied for a specific case and extrapolated for 1 load parameter (Line tensions). In this thesis it will be studied if response based design can be used as a general tool to calculate reliable 100-year return levels for the dominant load parameters. This resulted in the following research question; “Is it possible to apply extreme value theory in an efficient way during the design stage with the use of hindcast data for the assessment of mooring configurations? To study this possibility a method reconstruction is performed to check the reproducibility of the three main steps that need to be taken in extreme value theory. These steps are first pre-processing of the raw environmental, secondly the simulation that provides the responses the third and final step is performing statistical inference on the response data to produce extrapolations. The robustness of the method is proved by executing a parameters study. In this parameter study several cases with variations in input parameters, configurations and environments are studied. Most cases gave reliable extrapolations. An exception is the 3X3 mooring system which has a significant difference in mooring system stiffness for in-line and in-between line translations. It is also noted that the accuracy of the fit depends on the arbitrary decisions made by the user during the process. To limit these arbitrary decisions, the possibility of using a Generalized Extreme Value (GEV) distribution is studied. Monthly and annual load data is fitted to a GEV distribution. This resulted in reasonable extrapolations and passed all goodness-of-fit tests for the annual data. In comparison with the conventional method the extrapolations where approximately 30% lower. The monthly data didn’t result in a reliable fit. To produce reliable extrapolations with monthly data, time dependent parameters need to be included to compensate for the seasonal variations. The final step was a case study. In this case study extrapolations for all dominant load parameters for a 3X3 mooring configuration are obtained. Four variations of the mooring system are studied. For all four variations good extrapolations are produced. It can be concluded that the performance of the methodology isn’t influenced by the individual mooring line make-up. For four out of five dominant load parameters reliable extrapolations are obtained. Extrapolations for the horizontal offset are less reliable due to the difference between in-line and in-between line stiffness in 3X3 mooring configurations. This problem can be coped with if directionality is included. To include directionality, the mooring configuration is divided into several sections. For these sections 100-year return levels are calculated individually. The answer to the research question is: It is possible to produce reliable 100-year extrapolations for shallow water by fitting responses from quasi-dynamic simulations of environmental data to a generalized Pareto distribution for the most dominant load parameters, with or without some adaptions to the methodology depending on the mooring lay-out.