WIT Press

A Multivariable Neural Network Ship Mathematical Model

Price

Free (open access)

Volume

12

Pages

10

Published

1995

Size

697 kb

Paper DOI

10.2495/MT950871

Copyright

WIT Press

Author(s)

R.S. Burns, R. Richter & M.N. Polkinghorne

Abstract

Conventional techniques to model plants require the utilisation of differential equations. The computation of such equations becomes slow in situations when the plants are highly complex. By taking training data from the real plant, it is possible to design and train a neural network which is capable of achieving a successful plant model using an off-line backpropagation technique. For a marine application, analysis of the results of this study is included which demonstrates how this technique may be applied, and the nature of the performance obtainable. 1 Introduction The classical approach to modelling the dynamic behaviour of rigid bodies is to express their behaviour as a set

Keywords