Marine system modelling using artificial neural networks: an introduction to the theory and practice

A.P. Roskilly, Ehsan Mesbahi, R.W.G. Bucknall, C. Elliott, G. Armstrong

Research output: Contribution to journalArticle

Abstract

Computer analysis and simulation is of vital importance to ensure the successful design and development of marine engineering systems. In many areas, conventional modelling methods have produced system representations which are inaccurate and/or computationally expensive in terms of processing power and memory requirements. The application of artificial neural networks in certain cases can be used to identify complex marine system characteristics and produce favourable results in comparison to mathematically derived models or look-up tables. This paper provides an introduction to the artificial neural network theory that is relevant to marine system modelling applications. The computational elements that combine to form artificial neurones are outlined and the network topology is examined for static and dynamic modelling purposes. The training procedure is crucial to the performance of a neural network, not only in terms of adapting the network but also in the way that data is presented. Again, these areas are discussed in this paper so that they are relevant to marine system models. Finally, a simple case study is used to demonstrate the application of these learning methods. A comparison is made between a mathematical formulation, a look-up table and an artificial neural network approach to obtain the friction factor as part of a simulation of pipeline fluid flow. This case study demonstrates the ability of well trained neural networks statically to model complex non-linear functions with extreme accuracy and speed.
Original languageEnglish
Pages (from-to)185-201
JournalTransactions - Institute of Marine Engineers
Volume10
Issue number3
Publication statusPublished - 1996
Externally publishedYes

Fingerprint

Neural networks
Marine engineering
Circuit theory
Neurons
Flow of fluids
Pipelines
Topology
Friction
Data storage equipment
Processing

Cite this

Roskilly, A.P. ; Mesbahi, Ehsan ; Bucknall, R.W.G. ; Elliott, C. ; Armstrong, G. / Marine system modelling using artificial neural networks : an introduction to the theory and practice. In: Transactions - Institute of Marine Engineers. 1996 ; Vol. 10, No. 3. pp. 185-201.
@article{0a3f73d1a0984a63b9c36cf7b38f6e46,
title = "Marine system modelling using artificial neural networks: an introduction to the theory and practice",
abstract = "Computer analysis and simulation is of vital importance to ensure the successful design and development of marine engineering systems. In many areas, conventional modelling methods have produced system representations which are inaccurate and/or computationally expensive in terms of processing power and memory requirements. The application of artificial neural networks in certain cases can be used to identify complex marine system characteristics and produce favourable results in comparison to mathematically derived models or look-up tables. This paper provides an introduction to the artificial neural network theory that is relevant to marine system modelling applications. The computational elements that combine to form artificial neurones are outlined and the network topology is examined for static and dynamic modelling purposes. The training procedure is crucial to the performance of a neural network, not only in terms of adapting the network but also in the way that data is presented. Again, these areas are discussed in this paper so that they are relevant to marine system models. Finally, a simple case study is used to demonstrate the application of these learning methods. A comparison is made between a mathematical formulation, a look-up table and an artificial neural network approach to obtain the friction factor as part of a simulation of pipeline fluid flow. This case study demonstrates the ability of well trained neural networks statically to model complex non-linear functions with extreme accuracy and speed.",
author = "A.P. Roskilly and Ehsan Mesbahi and R.W.G. Bucknall and C. Elliott and G. Armstrong",
year = "1996",
language = "English",
volume = "10",
pages = "185--201",
journal = "Transactions of The Institute of Marine Engineers",
issn = "0020-2924",
publisher = "Marine Management (Holdings) Ltd",
number = "3",

}

Marine system modelling using artificial neural networks : an introduction to the theory and practice. / Roskilly, A.P.; Mesbahi, Ehsan; Bucknall, R.W.G.; Elliott, C.; Armstrong, G.

In: Transactions - Institute of Marine Engineers, Vol. 10, No. 3, 1996, p. 185-201.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Marine system modelling using artificial neural networks

T2 - an introduction to the theory and practice

AU - Roskilly, A.P.

AU - Mesbahi, Ehsan

AU - Bucknall, R.W.G.

AU - Elliott, C.

AU - Armstrong, G.

PY - 1996

Y1 - 1996

N2 - Computer analysis and simulation is of vital importance to ensure the successful design and development of marine engineering systems. In many areas, conventional modelling methods have produced system representations which are inaccurate and/or computationally expensive in terms of processing power and memory requirements. The application of artificial neural networks in certain cases can be used to identify complex marine system characteristics and produce favourable results in comparison to mathematically derived models or look-up tables. This paper provides an introduction to the artificial neural network theory that is relevant to marine system modelling applications. The computational elements that combine to form artificial neurones are outlined and the network topology is examined for static and dynamic modelling purposes. The training procedure is crucial to the performance of a neural network, not only in terms of adapting the network but also in the way that data is presented. Again, these areas are discussed in this paper so that they are relevant to marine system models. Finally, a simple case study is used to demonstrate the application of these learning methods. A comparison is made between a mathematical formulation, a look-up table and an artificial neural network approach to obtain the friction factor as part of a simulation of pipeline fluid flow. This case study demonstrates the ability of well trained neural networks statically to model complex non-linear functions with extreme accuracy and speed.

AB - Computer analysis and simulation is of vital importance to ensure the successful design and development of marine engineering systems. In many areas, conventional modelling methods have produced system representations which are inaccurate and/or computationally expensive in terms of processing power and memory requirements. The application of artificial neural networks in certain cases can be used to identify complex marine system characteristics and produce favourable results in comparison to mathematically derived models or look-up tables. This paper provides an introduction to the artificial neural network theory that is relevant to marine system modelling applications. The computational elements that combine to form artificial neurones are outlined and the network topology is examined for static and dynamic modelling purposes. The training procedure is crucial to the performance of a neural network, not only in terms of adapting the network but also in the way that data is presented. Again, these areas are discussed in this paper so that they are relevant to marine system models. Finally, a simple case study is used to demonstrate the application of these learning methods. A comparison is made between a mathematical formulation, a look-up table and an artificial neural network approach to obtain the friction factor as part of a simulation of pipeline fluid flow. This case study demonstrates the ability of well trained neural networks statically to model complex non-linear functions with extreme accuracy and speed.

M3 - Article

VL - 10

SP - 185

EP - 201

JO - Transactions of The Institute of Marine Engineers

JF - Transactions of The Institute of Marine Engineers

SN - 0020-2924

IS - 3

ER -