SHIP RESISTANCE PREDICTION USING ARTIFICIAL NEURAL NETWORKS

摘要

Ship design is a complex process undertaken in the form of a design spiral. Calculation of resistance of a vessel is one of the key steps of this design spiral. The traditional methods of calculation of resistance involve model tests or regressions equations. The major disadvantages of these methods are that they are time and cost intensive and the generalisation ability is low. In this paper, the feasibility of using artificial neural networks for determining the residuary resistance of a ship has been studied The data for prediction of resistance is obtained from experiments carried out on catamaran hull forms by Insel and Molland. The input features are slenderness ratio, B/T ratio, S/L ratio and Froude number. It was found that a three hidden layered network predicted the resistance of a ship with good accuracy. The mean relative error for different optimisers was also obtained and Adam resulted in the least error. Hence, we can conclusively say that artificial neural networks are a useful tool for predicting the residuary resistance. The main advantage it offers is that neural networks can adapt to new data easily.