ai-based prediction of berm recession under irregular wave attack, toward standardization

Berm breakwaters are an effective and environmentally friendly way to protect coastlines, yet designing them reliably remains difficult because no widely accepted, accurate prediction method exists. traditional empirical formulas often disagree, lack solid physical basis, and apply only in narrow conditions, while recent studies have produced many specialized models that are hard to compare or use confidently in practice. this paper presents a practical step forward: a single, unified prediction tool built with a multi-layer perceptron (mlp) neural network trained on a large, carefully harmonized international collection of physical model test results. by bringing together and standardizing data from different laboratories. when rigorously validated on completely unseen data, the mlp model clearly outperforms existing formulas, achieving good performance with r² = 0.819 and significant improvements in other key parameters such as scatter index (si) and correlation coefficient (cc). these results show that a well-trained ai model can finally provide engineers with a dependable, transparent, and truly universal design standard for berm breakwaters recession.