Reliability assessment of an offshore wind turbine jacket under one ultimate limit state considering stress concentration with active learning approaches

At present, offshore wind turbine jacket structures are generally modeled with beam elements. The local joint flexibility and stress concentration of the joints are hardly considered. In this work, one jacket model is developed to consider joint stress concentration by using super-elements. Another jacket model is also developed with pure beam elements. The stress concentration is investigated between the two jacket models. Moreover, a general framework is proposed to conduct the reliability analysis with active learning Kriging approaches and one typical ultimate limit state is studied. At first, the developed jacket models are validated with the numerical simulation results of National Renewable Energy Laboratory (NREL) technical reports. Secondly, global sensitivity analysis is carried out to reduce the random parameters for the reliability analysis. In the end, the reliability analysis is conducted with active learning Kriging approaches. The results show the probabilities of failure of the beam model are very sensitive to the assumed stress concentration factors. More importantly, it is noticed that active learning Kriging approaches can efficiently and accurately assess the probabilities of failure. Compared with the traditional Monte Carlo simulation approach, active learning Kriging approaches can reduce computational loads by hundreds of times and have the same accuracy.