Probabilistic fatigue surrogate model of bimodal tension process for a semi-submersible platform

Abstract

Abstract Mooring-line tension typically involves a bimodal process in which a wave-frequency component and a low-frequency component are induced by dynamic wave loading. The non-Gaussian characteristics make fast frequency domain estimation inaccurate. Although the time-domain fatigue damage is believed to be the most reliable method, it is computationally expensive; thus, in the present study, surrogate models including artificial neural network and kriging models are constructed for fatigue-damage prediction of the bimodal tension process to improve the accuracy and efficiency compared with the simple frequency- and time-domain approaches. A parametric study is conducted to investigate the spectrum parameters and correction factor for the bimodal tension process, and the fatigue damage under arbitrary wave conditions is evaluated using interpolation techniques for long-term fatigue analysis. The fatigue failure probability of mooring lines under dynamic environmental loads is calculated using the surrogate models and various spectral fatigue techniques. The results indicate that with a finite database of time-domain fatigue damage, the proposed surrogate models-based approach can provide a significantly more accurate assessment of the fatigue failure probability than spectral-based approaches.