ANN-based surrogate modeling of composite laminates for structural optimization of wave energy converters

Wave energy converters (WECs) possess considerable renewable energy potential; however, their large-scale deployment remains constrained by structural durability limitations and high manufacturing costs in harsh marine environments. Composite materials offer favorable strength-to-weight characteristics, yet optimizing laminate configurations via conventional finite element analysis (FEA) is computationally intensive, limiting exploration of high-dimensional design spaces. This study proposes a surrogate-based structural optimization framework integrating high-fidelity FEA with machine learning for composite WEC buoy design. A parametric analysis of a truncated spheroidal buoy was performed in ANSYS, systematically varying material type, stacking sequence, ply orientation, and laminate thickness under wave loading to construct a structural response database. An artificial neural network, trained on this dataset (R² = 0.97, RMSE = 1.8 mm) and coupled with a genetic algorithm, identified Pareto-optimal laminate configurations minimizing deformation subject to mass constraints. Optimized designs, validated against independent FEA simulations, exhibited deviations below 5%, substantially reducing computational cost and enabling scalable composite WEC hull design.