A Convolutional Neural Network Developed to Predict Speed Using Operational Data
S.R.A. de Geus-Moussault (TU Delft - Ship Design, Production and Operations)
Mark Buis (Vrije Universiteit Amsterdam)
Herbert Koelman (NHL Stenden University of Applied Sciences)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
In the shipping industry power prediction methods are commonly used. An option is to predict the power based with a theoretical analysis. However, with a purely theoretical approach it is not possible to evaluate all operating conditions. The second, simulation methods, are able to describe all the necessary quantities in detail. Nonetheless, simulation requires relatively high computational power. Thus, the current power prediction methods used in the shipping industry are insufficiently all-encompassing or accessible. Therefore, a machine learning approach is developed to calculate the ships speed over ground using neural network and convolutional neural network techniques. For training and validation of the model operational data from a fall-pipe vessel is used. The developed method combined with ship motion could result in an optimal power usage, and thus leads to reduced fuel consumption and emissions. The method could also be used for optimised routing. Although in this case study applied to one single vessel, the developed model is generally applicable, providing ship management companies the possibility to train the model with operational data from their fleet, therewith, offering the possibility of reduced fuel consumption and thus emissions on a global level.