Surrogate infill criteria for operational fatigue reliability analysis

Abstract

Analysis of Offshore Wind Turbine (OWT) fatigue damage is an intense, resource demanding task. While the current methodologies to design OWT to fatigue are quite limited in the way and amount of uncertainty they can account for, they still represent a relevant share of the total effort needed in the OWT design process. The robustness achieved in the design process is usually limited. To enable OWT to be more robust, an innovative methodology that tackles current limitations using a balanced amount of designing effort was developed. It consists of generating a short-term fatigue damage (DSH) using a Kriging surrogate model that accurately accounts for uncertainty using an adaptive approach. The current paper discusses the application of a reinterpolation convergence to build a Kriging surrogate model that replicates DSH in OWT tower components. Different variables involved in the convergence are discussed. The discussion extends then to how the design could be improved by using different convergence scenarios for the Kriging surface. Cross-validation is used to train and validate the surrogate surface. The main goal is to give the designer a rationale on the trade-off between computational time and accuracy using the mentioned approach to design robust OWT towers. Results show that on a design basis two levels of approach may be efficient. In the first, if a very high computational cost is expected, a trade-off between accuracy and computational time must be considered and then, if the intention is to check how robust the current design is, a full convergence of the surface should be pursued.