Modeling beach realignment using a neuro-fuzzy network optimized by a novel backtracking search algorithm

Abstract

Beach realignment is caused by sediment movement across and along the beach due to the ever-changing incident waves. Its driving mechanisms are highly nonlinear and cannot be easily resolved by numerical process-response modeling which may also involve high computational costs. In this contribution, an alternative approach is developed to cope with the nonlinearity of beach realignment, which utilizes a neuro-fuzzy neural network optimized by a novel backtracking search algorithm. The network comprises multiple layers operating in sequence and establishes input–output relationships in terms of first-order fuzzy rules. The proposed backtracking search algorithm improves its performance by effectively modifying the existing mutation and crossover operations of the standard algorithm. A novel experimental setup was deployed in a touristic beach of Santorini island, Greece, to collect high frequency morphological and hydrodynamic information and generate a data set described by few representative input variables. Three additional networks were designed using the same data set. Standard criteria and nonparametric statistical analysis showed that the proposed approach outperforms all other tested methods in the case of modeling beach realignment, achieving also a significant improvement over previous efforts.