This set of scripts and data files can be used to re-generate the Adaptive Screening method and its application to a case study. Both are described in the paper "Designing for dangerous waves – a new ‘Adaptive Screening’ method to predict extreme values for non-linear responses of marine structures to waves", which is presently under review.
Predicting extreme values of strongly non-linear ship responses (such as wave impact loads) is crucial for ensuring safety and performance of maritime structures. However, this is challenging due to the complexity and rarity of the responses. Existing methods are limited, as they are either suitable for only weakly non-linear responses or are otherwise very computationally expensive. The paper above in combination with this dataset introduces a new event-based multi-fidelity method called ‘Adaptive Screening’ to efficiently predict extreme values of strongly non-linear waves and wave-induced responses. It combines elements of screening, multi-fidelity Gaussian Process Regression, and adaptive sampling. A case study with second-order wave data validates the effectiveness of the new method in the paper. The input necessary to reproduce the case study is also included in the dataset. The paper showed that Adaptive Screening outperforms the conventional brute force method (which is also included in the dataset) by providing more accurate extreme values for this problem, while also significantly reducing high-fidelity simulation time. Adaptive Screening depends on a response-dependent low-fidelity indicator variable. We also show that the quality of the used screening indicator does not have a significant influence on the accuracy of the predicted results, but the high-fidelity simulation time required to obtain converged results increases with decreasing indicator quality. Still, the method is foreseen to be more efficient than the conventional method by a factor of five. The weakly non-linear test case shows that Adaptive Screening is very promising for the strongly non-linear responses it was designed for. Using the attached dataset, all results in the paper can be reproduced.
%I 4TU.ResearchData