Kriging models for aero-elastic simulations and reliability analysis of offshore wind turbine support structures

The existence of uncertainties in material properties, environmental loads and soil properties as well as the presence of nonlinearities introduced by the control systems have a remarkable influence on the dynamic response of offshore wind turbine (OWT) support structures. The reliability computations of these structures need to consider implicit expensive-to-evaluate limit state functions, implying large computational costs. This paper addresses these limitations by proposing a computationally-efficient reliability framework for OWT support structures, based on the use of a kriging model to approximate the response of the system, capturing both the dynamic behaviour of the structure and inherent uncertainties. The surrogate model is built with sample points from stochastic fully coupled simulations in the time-domain. A thorough sensitivity study is performed on the influence of number of sample points, the seeds used to obtain each point, the range of the variables and the inherent variability in turbulent wind and stochastic waves. The framework is used to evaluate the reliability of the NREL 5 MW turbine model, mounted on a monopile with a flexible foundation for the severest Design Load Cases (DLCs) from the IEC 61400-3. The results agreed with the general literature showing that the structure is far from failure.