Machine Learning Surrogates for Predicting Line Tensions in Catenary and Semi-Taut Mooring Systems

Abstract Mooring systems are safety-critical components essential for the station-keeping and operational integrity of floating offshore wind turbines (FOWT). While various mooring configurations have been developed, identifying the optimal system for intermediate water depths remains a significant challenge. This paper investigates two representative configurations—catenary and semi-taut mooring systems—focusing on their influence on the dynamic responses of FOWT in intermediate water depths. A comprehensive sensitivity analysis is performed under representative environmental loading conditions to quantify key response metrics, including fairlead tension and platform offset. This analysis identifies dominant nonlinear features and regime-dependent behaviors inherent to each mooring concept. Given that systematic design optimization requires a high volume of computationally expensive time-domain simulations, a machine-learning-based surrogate model is developed. Trained on the generated simulation database, the surrogate emulates mooring responses as a function of line length and loading conditions, demonstrating high fidelity in capturing underlying physical trends and nonlinear tension characteristics. Finally, this study proposes a framework for integrating the surrogate model with multi-objective optimization algorithms. This approach is recommended as an efficient means to explore the design space and identify critical trade-offs between line length, maximum tension, and platform offset, bypassing the need for exhaustive high-fidelity simulations in future design iterations.