Neuronska mreža za predviđanje značajki slobodne vožnje brodskog vijka u sapnici

Abstract

Dynamic positioning system is a computer-controlled system which enables maintaining the marine vessel’s position and heading by means of propellers and active thrusters. In order to improve the efficiency of the heavily loaded propeller at low advance velocity, the azimuth thrusters with a nozzle around the propeller are mostly used. The paper deals with the development of the neural network prediction model for the Wageningen ducted propeller series. The aim is to create a prediction tool that provides the estimation of the open water characteristics of the four blade Ka series in combination with the 19A accelerating nozzle. For this purpose, two layered feedforward neural network is used. The proposed neural network model provides the possibility for a more general modelling of azimuth thruster in usually very complex hydrodynamic propulsion model of the dynamically positioned marine vessel.