Bayesian network-enhanced surrogate-assisted reliability-based design optimisation for offshore wind turbine support structures

Offshore wind turbine support structures are subject to significant environmental and structural uncertainties, requiring reliable and efficient design strategies. This study proposes an integrated framework combining Bayesian networks (BN), finite element analysis, multi-response surface models (MRSM), and reliability-based optimisation. Key uncertainties, including environmental loads, material variability, geometric imperfections, and soil-structure interaction, are quantified probabilistically. BN is used to capture dependencies among environmental variables, and correlated samples are propagated through a high-fidelity finite element model of the NREL 5 MW monopile turbine. An MRSM surrogate is constructed to approximate structural responses and validated with high accuracy. Finally, a SORA-based optimisation scheme, coupled with a hybrid MIGA and NLPQL optimiser, is applied for reliability-constrained design. The proposed framework improves computational efficiency while ensuring accurate uncertainty propagation, providing an effective tool for robust offshore wind turbine design.