Analysis of long‐term loading characterization for stress‐cycle fatigue design

Abstract

The present paper discusses the application of stress‐cycle (SN) fatigue design procedures for offshore wind turbines, applied to the tower and substructure, and emphasizing the long‐term loading characterization. Three main methodologies are compared to assess SN fatigue long‐term design, direct scale‐up of loads and cycles to the design lifetime, probability fit for the loading distribution, and truncation and fit of the tail region. Different characteristic SN slopes are researched. The comprehensive comparison developed shows that limited interest exists in increasing the complexity of the loading assessment for SN fatigue design. In particular, the notion of representative sample has key influence on the fatigue calculations. Probability analysis techniques are not an adequate substitute of a representative sample. If definition of a representative sample for design is unpractical due to its large cost, direct scale‐up of cycles from a time t to a larger time T can be applied with uncertainty assessment. Bootstrapping of load ranges is implemented to tackle the limitations imposed by approaching a design time much larger than the evaluated time. It takes advantage of the fact that SN fatigue design is a statistical problem of mean value. Bootstrapping proved to be an efficient estimator of uncertainty with relatively constant confidence bounds, of particular interest to perform SN fatigue design with small samples. New insights on its application for fatigue design are presented. Due to its simple application and nonparametric character, it can be applied in combination with other techniques, such as extrapolation of loads and cycles.

Flood adaptation decision‐making for vulnerable locations using expectation–quantile–investment analysis

“…The two‐parameter Weibull model has wide applicability (Rinne, 2008) and support in which is of interest for an application of risk that relies in estimates of damage. Nonetheless, as highlighted in Teixeira, Nogal, and O'Connor (2019), the upper unbounded threshold for the support of any transfer models may impose some limitations and needs to be accounted for. In the present, an upper bound of loss is imposed by the event that has a recurrence of 500 years (38.8 M€).…”