Enhancing data-driven design for offshore jacket substructures through synthetic data

Offshore jacket substructures are essential for supporting wind turbines in deep water due to their high stability and load-bearing capacity. However, as turbine sizes grow, the conceptual design of these complex structures becomes increasingly challenging. While Machine Learning has shown promise in predicting key design parameters, existing models are limited by the scarcity and low variability of real-world data. This study addresses this limitation by augmenting an existing dataset of 100 jacket samples with synthetic data generated using promising generative model for tabular data. The augmented dataset is used to train and evaluate supervised learning models, aiming to improve their predictive accuracy and robustness. Preliminary results demonstrate that synthetic augmentation can reduce overfitting, enhance model robustness, and reveal complex input-output relationships. This work highlights the potential of synthetic data as a valuable resource in data-driven conceptual design workflows for offshore jacket substructures.