Estimation of Hydraulic Characteristics around Artificial Reef using Neural Network

Abstract

Takehisa SAITOH, Hiroto SUNAHARA, Masato ICHIKAWA, Tadashi FUKUMOTO, Hajime MASE, and Hajime ISHIDA This study aims to develop a highly accurate method for an estimation of hydraulic characteristics around an artificial reef by using an artificial neural network, instead of a conventional method using a regression analysis. The wave trapping artificial reef is chosen as the structure of an artificial reef, and transmission coefficient and wave setup behind the reef are focused as hydraulic characteristics. For the training of the artificial neural network, the Levenberg-Marquardt method with Bayesian regulation is employed. Predictions of transmission coefficient and wave setup from the trained network agreed well with the experimental results comparing with predicted ones from the method using a regression analysis.