A Practical Estimation Method of Resistance for Planing Hull with Artificial Neural Networks

Abstract

This paper proposes a practical design system for the resistance estimation of high speed mono-hull craft. It uses regressive approach with artificial neural networks, which learn the teaching sample data of many pairs of input/output vectors, generalize it, and establish the nonlinear mapping relation between its i/o.The system consists of a tank test database, mapping neural networks, and a descriptive neural network. The tank test database is based on the Series 62 which has 1725 data records. The input of the Mapping neural networks is a set of speed, length, chine breadth, weight, and center of gravity position. While, its output is a set of residuary resistance, running trim, rise of center of gravity, and wetted surface. The descriptive neural network denotes the density of the tank test data. It indicates the designer how reliable the estimation is.The learning process of these artificial neural networks is based on back propagation method. The proposed system is easily implemented into the spreadsheet software, very easy to use, and has the practical accuracy, if the learning of the networks is once established.