AI in Ship Design: A State-of-the-Art Review

The ageing global fleet and demanding sustainability goals are driving a growing role for machine learning (ML) and artificial intelligence (AI) in naval architecture. This review examines state-of-the-art AI and ML applications in ship design, focusing on ship hull hydrodynamics and structural design. It highlights the use of advanced models such as deep neural networks (DNNs), support vector machines (SVMs), decision tree ensembles, graph neural networks (GNNs), and physics-informed neural networks (PINNs). DNNs and GNNs demonstrate superior capabilities in capturing complex hydrodynamic and structural behaviours when ample data and computational power are available, whereas simpler models (e.g., SVMs or tree ensembles) often excel with limited data and, in the case of trees, provide greater transparency in predictions. PINNs integrate fundamental physics into neural models, producing physically consistent predictions and alleviating data scarcity, albeit with increased implementation complexity and computational cost. The advantages and limitations of each approach are considered with respect to data availability, computational resources, and design objectives. Ongoing challenges include data scarcity due to limited and confidential datasets, high cost of obtaining high-fidelity data, and the lack of robust validation frameworks for AI models. The review concludes with recommendations for developing generalizable, physics-integrated AI frameworks and underscores the importance of context-aware model selection for future ship design advancements.