Advances in Modelling Wave-Structure Interaction Through Artificial Neural Networks

Abstract

This contribution describes a new Artificial Neural Network (ANN) able to predict at once the main parameters representative of the wave-structure interaction processes, i.e. the wave overtopping discharge, the wave transmission coefficient and the wave reflection coefficient. The development of this ANN started with the preparation of a new extended and homogeneous database (derived from CLASH database) which collects all the available tests including at least one of the three parameters, for a total amount of 16,165 data. Some of the existing ANNs were compared and improved, leading to the selection of a final ANN, whose architecture was optimized through an in-depth sensitivity analysis to the learning and training ANN features. The new ANN here proposed provides accurate predictions for all the three parameters, resulting in a tool that can be efficiently used for design purposes. Keywords: artificial neural network, database, wave overtopping, wave reflection, wave transmission INTRODUCTIONThe design of coastal and harbour structures requires a systematic analysis of all the processes of wave-structure interaction, which takes into account the combined effects of wave overtopping, wave transmission and wave reflection. Indeed, all these phenomena should be considered as different outcomes of the same physical process, and therefore should be investigated contemporarily. However, based on the traditional approach most of the existing formulae are targeted to represent just one process and are fitted on specific (more or less wide) databases, addressing usually one or few structure types and having therefore a specific (more or less narrow) validity field.The development and/or use of an Artificial Neural Network (ANN) is therefore particularly recommended in case of complicated structure geometries and variable wave conditions. This kind of predictive method requires however a homogeneous and "wide-enough" database: the number of data should be sufficient for training the ANN based on a number of ANN parameters and including a sufficiently wide number of data for all range of possible output values. There are specific cases that also an ANN cannot deal with, such as very complex walls, perforated caissons and double promenades, see for details EurOtop (2007) and specifically the methodology released within the PC-OVERTOPPING calculator (http://www.overtopping-manual.com/calculation_tool.html).One of the most successful ANNs is the method available from CLASH (2004) and EurOtop (2007) for wave overtopping, which was specifically developed to predict the overtopping discharge for a wide range of coastal structures, including complex geometries (Van Gent et al., 2007). Each of these ANNs actually proved to be able to overcome some of the limits imposed by the traditional empirical formulae, but they are still restricted to reproduce only one of the processes involved in the wave-structure interaction. Nevertheless, the assumption that all the processes are physically correlated implies that a unique set of phys...

Prediction of mean wave overtopping at simple sloped breakwaters using kernel-based methods

“…on an extended database (Zanuttigh et al 2014) mainly derived from the CLASH database. However, in terms of the accuracy of wave overtopping prediction, the recent version does not show a significantly better performance in comparison to CLASH-ANN (Formentin et al 2017;Pillai et al 2017).…”

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

“…The ANN developed within the CLASH (2004) project and proposed by EurOtop (2007) for the prediction of the average overtopping discharge, q, is the ANN by Van Gent et al (2007). Further analysis and other ANNs have been delivered during and after CLASH (2004): the ANN by Verhaeghe (2005) for q and the ANN developed by the authors (Zanuttigh et al, 2014). The last one predicts the main wave-structure interaction parameters: besides q, the wave reflection, K r, and transmission, Kt coefficients.…”

Review on Applications of Machine Learning in Coastal and Ocean Engineering

“…Subsequently, errors in the CLASH database were corrected, and more than 17,000 datasets were expanded through the Innovative Technologies for Safer European Coasts in a Changing Climate project (Zanuttigh et al, 2014). The calculation for the overtopping rate and the estimated results for uncertainty were presented using an ANN model.…”