Author(s): Alberte Castro; Francisco Taveira Pinto; Gregorio Iglesias
Keywords: Artificial intelligence; artificial neural network; shoreline erosion; submerged breakwater; wave reflection
Abstract: The reflection process at submerged breakwaters is investigated by means of an artificial neural network (ANN) model. This model estimates the reflection coefficient based on three dimensionless products which are functions of the wave parameters and the characteristics of the submerged breakwater. The data used for training the model were obtained from an extensive experimental campaign in which seven models of submerged breakwaters were tested under irregular waves combinations. To choose the neural network architecture best suited for this problem, the performances of 400 ANN models involving 10 different architectures are assessed. Having selected the most appropiate architecture, the model was succesfully trained and validated. Excellent agreement was achieved between the model's results and the experimental data. The new artificial intelligence model can be used as a virtual laboratory to predict the reflection coefficient without a need for a physical model test.