Characterizing uncertainty in complex systems is steadily growing as a topic of interest. One of the efficient ways to characterize a complex system is achieved by probabilistic sensitivity analysis. In the context of performing it, there are a limited number of methods to quantify the change of the output to its full probabilistic extent. Moreover, in some engineering applications, such as reliability analysis, some established indicators of sensitivity do not fit the best interest of the analysis. This is the case of Kullback-Leibler divergence. Despite applied for probabilistic sensitivity analysis, it has limited interest in certain circumstances. A transformation of this indicator of entropy between two distributions is proposed in the present work. This transformation is used to establish a complementary indicator that is more perceptive, and more efficient for reliability analysis. This new function is applied for the global sensitivity analysis of an offshore wind turbine on a monopile foundation. Results show that, for engineering problems as the one presented, the usage of this transformed indicator produces intuitive results. It allows the efficient identification of relevant states of operation as well as the most influent variables in the design of experiments, resulting in better comprehension of system's behaviour and operational risks.
For complex systems, the applicability of surrogate models has shown the potential to enable accurate assessments using a reduced batch of data and to compile information about large datasets. These behave as black-box functions that replace a series of inputs/outputs. In the present work, a Kriging surrogate is used to predict confidence intervals in an offshore wind turbine tower fatigue design. Uncertainty in fatigue due to loading is highly connected to the mean. One year operational fatigue results is used to validate the results. The Kriging is applied to replicate the yearly states of operation, and successfully predicts intervals of confidence for the long-term fatigue design. Regarding the interest of data analysis, the approach implemented is characterized by its flexibility and capability of approaching any problem that can be characterized by a single variable. Being therefore an interesting tool in decision schemes where large datasets are available or prediction of unknown outputs is required.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.