Contour Methods for Estimation of Multi-dimensional Extreme Riser Response

Abstract

The aim of this thesis is to investigate the extreme response of a flexible riser.A full long term analysis is usually carried out to find the response with a givenreturn period in years, which is a useful quantity in design.Time domain simulations are required to obtain the short term stochasticstructure of the response, because of its non-linear nature.Unfortunately, this is a method that requires excessive computational power.It is desired to find a method to estimate the extreme values without establishingthe full long term response distribution. The Contour Line Method (CLM)provides a solution to this problem, as it decouples the two sources ofrandomness in the stochastic response process; the randomness of the slowlyvarying parameters that define the sea state, and the randomness of the response,given a sea state.Two case studies will be investigated in this thesis; the first concentrates on onedimensionalresponse. First, sea states with a 100 year return period arecalculated from the long term distribution of the simultaneous values ofsignificant wave height and spectral peak period. Short term distributions for 3-hour response extremes are then generated for these sea states. Using the 90 %percentile gives reasonable approximations of the 100-year response from thefull long term analysis. The results are in agreement with existing literature onthe area.This thesis also comprises a new approach, the use of the CLM on multidimensionalresponse, i.e. simultaneous values of axial force and curvature.The scale factor is defined to be the length of the parametric load vector, dividedby the distance to the capacity surface of the structure.Using a scale factor is convenient because it lays no constraint on the global andlocal analysis schedule-wise.The scale factor reduces the multi-dimensional response process to a onedimensionalresponse process that can be predicted with the CLM.A convex hull of a parametric time series is a surface enclosing all the worstload combinations encountered during the time series.The response contour is a ?long term? convex hull with a given return period inyears. It gives information about which load components are causing excessivestresses. A procedure to apply the CLM to predict the response contour ispresented in this thesis.