LAUSR

Abstract A truncated mooring and riser system is the foundation of hybrid model testing of deepwater floating platforms. Existing truncation design methods are mainly based on multi-objective optimization algorithms, which make them complicated and time-consuming. This paper presents an empirical truncation design method, which can instantaneously determine the properties of truncated mooring systems and risers using data-driven models generated based on supervised learning with an artificial neural network. Features of the input parameters are constructed specifically for the truncation design and filtered according to their rank of importance to improve the performance of the data-driven models. Moreover, the relationship between the truncated properties and overall parameters, such as the truncated depth, is investigated to propose methods for expanding the application range of the data-driven models. The accuracies of the data-driven models are evaluated statistically using test samples, in comparison with the accuracies of the optimization algorithm. Finally, the significant features determining the truncated properties are discussed to reveal the regulations of the truncation design. The proposed empirical method can serve as a more efficient tool for the applications of hybrid model testing and the reveal of key features is helpful for understanding the truncation process.