A Neural Network Tool for Predicting Wave Reflection, Overtopping and Transmission

Abstract

This contribution presents a new Artificial Neural Network (ANN) tool that is able to predict the main parameters describing the wave-structure interaction processes: the mean wave overtopping discharge (q), the wave transmission and wave reflection coefficients (Kt and Kr). This ANN tool is trained on an extended database (based on the CLASH database) of physical model tests, including at least one of the three output parameters, for a total number of nearly 18,000 tests. The selected 15 nondimensional ANN input parameters represent the most significant effects of the structure type (geometry, amour size and roughness) and of the wave attack (wave steepness, breaking, shoaling, wave obliquity). The model can be used for design purposes, leading to a greater accuracy than existing formulae and similar tools for complex geometries for the prediction of Kr and Kt , and it has a similar accuracy as the CLASH ANN for predicting q.