|dc.description.abstract||The objective of this master's thesis is to perform a long term analysis of a non-linear response problem using the Peak Over Threshold (POT) method. Offset of a semi submersible was selected as suitable response problem to be investigated.
The characteristic responses obtained from the POT analysis is verified and compared to long term responses resulting from an all sea states analysis and original design conditions.
The key idea of the POT method is to establish a long term distribution of the largest response during a random storm. Each storm is modeled as a sequence of stationary 3-hour sea states. The long term distribution of the maximum response is obtained by merging the conditional distribution of the maximum given the most probable maximum with the long term distribution of the most probable maximum. Response contours are created using the Inverse First Order Reliability Method (IFORM), and the largest response on this contour is taken as the q-probability response.
The all sea states approach combines the short term variability of the 3-hour maximum response described conditionally on all realizations of sea states with the long term variability of the weather characteristics under consideration, i.e. significant wave height (Hs), spectral peak period (Tp) and wind velocity (W), to get the long term distribution of the largest response during a random 3-hour sea state. Here, a proper distribution of the weather characteristics is estimated by processing hindcast data for Haltenbanken. The marginal distribution of $H_s$ is established together with the conditional distribution of $W$ given $H_s$ and the conditional distribution of $T_p$ given $H_s$ and $W$. The q-probability responses are calculated utilizing IFORM.
The traditional design conditions assumed to give a conservative estimate of the 100-year, i.e. ULS, response is taken as the most probable largest response during a storm with 100-year wind speed and 100-year waves. Similarly, the 10,000-year response is taken as the most probable largest response in a sea state modeled with 100-year wind and 10,000-year waves, since waves are proven to be the most important weather characteristic.
This study demonstrates that the POT method is less conservative than the all sea states approach and original design conditions. The practical implication of less conservative q-probability responses obtained from a POT analysis is a more optimized design of offshore structures without compromising the safety aspect.||