Offshore monopile wind-turbine systems face significant design challenges due to complex environmental loading conditions, including wind, waves, current, and their intricate coupling effects. Computational Fluid Dynamics (CFD) simulations of complex nonlinear wave-structure interactions have become essential for understanding offshore systems. However, high-fidelity (HF) CFD analyses are computationally demanding, making it impractical to use them alone for extensive parametric studies. Surrogate models provide a promising alternative, as they can deliver efficient and accurate predictions once trained. This study presents a methodology for combining HF computational models and a data-driven approach to assess hydrodynamic loads on an offshore monopile subjected to waves and current. A three-dimensional numerical wave tank was developed in the OpenFOAM environment and validated using experimental and theoretical data. This model accurately accounts for complex wave-current interactions and the resulting force variations on the monopile. The obtained results indicate that the hydrodynamic forces on the monopile are significantly affected when waves and current act concurrently. To reduce the extensive computational HF simulations, a Gaussian Process Regression (GPR)-based surrogate model was developed using HF simulations sampled via Latin Hypercube Sampling. The surrogate model was trained to predict peak inline and transverse forces for various wave heights, frequencies, and current velocities. The GPR model exhibits high accuracy with = 0.98 (coefficient of determination) and MAE = 0.03 (Mean Absolute Error) for the inline force, and = 0.94 and MAE = 0.01 for the transverse force, thus demonstrating its ability to accurately assess complex nonlinear responses, while reducing the computational demand significantly. These findings provide a practical and applicable solution for offshore structural design, especially during early-stage assessments or probabilistic load evaluations.
扫码关注我们
求助内容:
应助结果提醒方式:
