Prediction of wave transmission over an outer submerged reef of tandem breakwater using RBF-based support vector regression technique

Abstract

The development of a mathematical model to determine transmitted wave height over a submerged reef of the tandem breakwater is complex. Therefore, it is necessary for researchers to adopt the physical model study to determine the parameters that influence the performance of breakwaters quantitatively. Physical modelling is laborious, expensive and lengthy in the procedure which makes it inconvenient for immediate needs. From the history, the development of the soft computing model shows that the soft computing techniques can be applied successfully to the prediction of the wave characteristics by making use of experimental data available. Similarly, attempt is made to predict the wave transmission over a submerged reef of tandem breakwater based on the data of Subba Rao developed in 2004 on a tandem breakwater in a 2D wave flume available at NITK Surathkal India using Support Vector Regression (SVR) model with different kernel functions. The non-dimensional input parameters used for the development of the models are five in number. Those inputs are incident wave steepness (Hi/gT2), relative reef crest width (B/L<inf>o</inf>), relative reef submergence (F/H<inf>i</inf>), relative reef crest height (h/d), depth parameter (d/gT2) and the output as (H<inf>t</inf>/H<inf>tmax</inf>). The 202 data points (70%) are used for training and the 86 data points (30%) for testing out of 288 total data points. The statistical parameters are computed using the predicted and observed data points after training and testing the SVR models. The RBF kernel gives good correlation to the prediction of transmitted wave heights during testing with RMSE as 0.09 and MAE as 0.07. Therefore, the SVR with RBF kernel function can be adopted as an alternative technique to predict the wave characteristics such as wave transmission over a submerged reef of the tandem breakwater. © Springer Nature Singapore Pte Ltd. 2019.