Prediction of extreme and tolerable wave overtopping discharges through an advanced neural network

Abstract

This paper presents an Artificial Neural Network (ANN) to predict the wave overtopping discharge at coastal and harbour structures for a variety of wave conditions and complex geometries. The goal of this work is to provide a robust tool in both extreme and tolerable overtopping conditions, starting from the ANN recently developed by the authors for wave reflection, overtopping and transmission. Optimisation of the existing ANN is analysed: (i) by training the ANN also on very low values of the overtopping discharge: (ii) by the set-up of an architecture consisting of a classifier-quantifier scheme; (iii) through the modification of the weight factors included in the boot-strapping resampling technique. The accuracy of the optimised ANN is proved predicting new data and datasets.