Extreme Response Prediction of Steel Risers Using a Four Parameter Distribution

Abstract

Short-term extreme response estimates are required in many areas of ocean and offshore engineering, such as steel risers design. As in many cases, the response in non-Gaussian, a theoretical solution, is usually not readily available for this purpose. Hermite transformation and Weibull-based models, among others, are some alternatives that have been used in connection with sampled response time series. In this work, a new approach is investigated. Recently, a four-parameter distribution known as the shifted generalized lognormal distribution (SGLD) has been presented in the literature. One of its main advantages is that it covers regions of skewness–kurtosis not covered by other distributions of common use in engineering. In this paper, the performance of this distribution is evaluated in the extreme values' estimation of the utilization ratios of steel riser sections. Three alternatives for using SGLD are investigated in two case studies of different dynamic behavior. The first one is a steel-lazy wave riser (SLWR) connected to a turret-moored FPSO (floating, production, storage and offloading unit) in 914 m water depth, and the second is a SLWR connected to a spread-mooring FPSO in a water depth of 1400 m. The results obtained by the SGLD-based analysis, which considered several simulation lengths, are compared to those obtained by means of an extreme value distribution fitted to episodical extremes obtained from many distinct realizations. The results of a traditional Weibull-fitting approach to the response peaks and those obtained with a Hermite transformation-based model are also presented for comparison.