Artificial neural network for estimation of harbor oscillation in a cargo harbor basin

A harbor should provide safe mooring for vessels and facilitate clean and unimpeded transfer of passengers and cargo between vessels and land. Therefore, oscillation in a harbor basin must be lower than the value providing safe anchorage. Conventionally, the oscillation level can be determined by physical and numerical model studies. In this study, physical and artificial neural network (ANN) models on a cargo harbor oscillation were conducted, and their results were compared. Physical model studies have been carried out in Karadeniz Technical University Civil Engineering Department Hydraulics Laboratory wave basin. The models were performed for 180 cases with various kinds of wave and breakwater condition. Wave heights were measured in 36 points in the harbor basin. The experimental data were divided into 144 training, 24 testing, and 12 validation patterns in the ANN model. By comparing the results of physical and ANN models, it has been concluded that the maximum and average relative errors computed for validation data set are 16.6 and 12.8 %, respectively.