Prediction of Solitary Wave Forces on Coastal Bridge Decks Using Artificial Neural Networks

Abstract

This study proposes an alternative and competitive methodology for predicting solitary wave forces on coastal bridge decks using artificial neural networks (ANNs). It is imperative to accurately predict the on-deck wave forces for the design and retrofit of coastal bridges subject to the potential impact of hurricanes and tsunamis. For this purpose, ANNs are used to determine wave loads based on a valid data set. First, the structural, fluid, and wave variables involved in the bridge deck-wave interaction are briefly introduced. The back-propagation network (BPN) wave force prediction model trained using the back-propagation algorithm is highlighted. A data set with 472 evidence cases is prepared based on extensive computational fluid dynamics (CFD) simulations. Three major input variables, the still-water level (SWL), bottom elevation of the girder/superstructure, and wave height, are selected. Then, the procedures of training the ANNs for the vertical and horizontal forces are presented in detail. Finally, the trained network structures with high predictive skills after substantial training are given with a proposed predictive equation for the vertical and horizontal forces. The results showed that the ANN methodology is robust and capable of capturing the underlying physical complexity in the bridge deck-wave interaction.