Prediction of breaking wave height by using artificial neural network-based approach

Abstract

This study applies artificial neural network-based methodology to predict the breaker height (Hb). 1120 data samples from 53 experiments covering a wide range of seabed slopes are used to examine 36 existing models and artificial neural network (ANN) models. The results indicate that the developed models are more reliable than existing models not only in overall cases but also in any slope range (horizontal, gentle, intermediate, and steep slope). Especially for ANN model 1, root-mean-square-relative error reduces 3.2%–5.9% in comparison with the three best existing formulas in general. Additionally, there is more than 90% of calculated data that differ from measured data by less than 20% as using ANN model 1, this result is better than all existing formulas. In case of classification according to surf similarity wave, ANN model 1 is also better at estimating breaking wave height for both spilling and plunging types. The sensitivity analyses for seabed slope, water depth, and deep-water wavelength with various models are also carried out. The effect of slope, breaker depth, deep-water wavelength, and deep-water wave height on Hb is quite similar to existing models. Based on the outstanding results, ANN model 1 is acceptable and highly recommended to apply in estimation of breaker height.