Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks

Magnetic signatures represent the magnetic field generated by a ship’s ferromagnetic components and provide valuable information for identifying vessels not only in naval operations, but also in civil passages. The topic of accurate modelling of these signatures is relevant to this day, b...

Full description

Saved in:
Bibliographic Details
Main Authors: Kajetan Zielonacki, Jaroslaw Tarnawski, Miroslaw Woloszyn
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10947760/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849730753323073536
author Kajetan Zielonacki
Jaroslaw Tarnawski
Miroslaw Woloszyn
author_facet Kajetan Zielonacki
Jaroslaw Tarnawski
Miroslaw Woloszyn
author_sort Kajetan Zielonacki
collection DOAJ
description Magnetic signatures represent the magnetic field generated by a ship’s ferromagnetic components and provide valuable information for identifying vessels not only in naval operations, but also in civil passages. The topic of accurate modelling of these signatures is relevant to this day, but also the complexity of the model necessary to accurately predict the ship’s magnetic field. This paper presents the implementation of a deep, recurrent neural network (RNN) designed for classification of compliance between the original magnetic signature of a ship and the one obtained from a model. Therefore, the quality of the model can be analyzed using a classifier during the modeling process. The necessity to introduce a tool for signature compliance classification arose during numerical modeling of a ship in Finite Element Method (FEM) environment as well as during reverse modeling based on data coming from measurements. Another application is the use of a shallow RNN for classifying ships by their size and type. A sufficient amount of data is rarely available and therefore data augmentation solution is necessary. The process of obtaining a large dataset of signals from a multi-dipole model and using an interpolation technique for generating training, validation and test data is comprehensively described. Methods used for selecting the best network structure and hyperparameter tuning using grid search and random search in order to achieve a satisfactory classification accuracy are thoroughly explained. Features, advantages and limitations of developed algorithms are derived strictly from the nature of neural networks.
format Article
id doaj-art-e0669aafab5349de8b4993da4f42f33f
institution DOAJ
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-e0669aafab5349de8b4993da4f42f33f2025-08-20T03:08:46ZengIEEEIEEE Access2169-35362025-01-0113595145953010.1109/ACCESS.2025.355733110947760Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural NetworksKajetan Zielonacki0https://orcid.org/0009-0004-7481-3247Jaroslaw Tarnawski1https://orcid.org/0000-0002-5744-5671Miroslaw Woloszyn2https://orcid.org/0000-0002-2663-2255Faculty of Electrical and Control Engineering, Gdańsk University of Technology, Gdańsk, PolandFaculty of Electrical and Control Engineering, Gdańsk University of Technology, Gdańsk, PolandFaculty of Electrical and Control Engineering, Gdańsk University of Technology, Gdańsk, PolandMagnetic signatures represent the magnetic field generated by a ship’s ferromagnetic components and provide valuable information for identifying vessels not only in naval operations, but also in civil passages. The topic of accurate modelling of these signatures is relevant to this day, but also the complexity of the model necessary to accurately predict the ship’s magnetic field. This paper presents the implementation of a deep, recurrent neural network (RNN) designed for classification of compliance between the original magnetic signature of a ship and the one obtained from a model. Therefore, the quality of the model can be analyzed using a classifier during the modeling process. The necessity to introduce a tool for signature compliance classification arose during numerical modeling of a ship in Finite Element Method (FEM) environment as well as during reverse modeling based on data coming from measurements. Another application is the use of a shallow RNN for classifying ships by their size and type. A sufficient amount of data is rarely available and therefore data augmentation solution is necessary. The process of obtaining a large dataset of signals from a multi-dipole model and using an interpolation technique for generating training, validation and test data is comprehensively described. Methods used for selecting the best network structure and hyperparameter tuning using grid search and random search in order to achieve a satisfactory classification accuracy are thoroughly explained. Features, advantages and limitations of developed algorithms are derived strictly from the nature of neural networks.https://ieeexplore.ieee.org/document/10947760/Deep learninggated recurrent unitmagnetic signaturesignal processing
spellingShingle Kajetan Zielonacki
Jaroslaw Tarnawski
Miroslaw Woloszyn
Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks
IEEE Access
Deep learning
gated recurrent unit
magnetic signature
signal processing
title Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks
title_full Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks
title_fullStr Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks
title_full_unstemmed Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks
title_short Ship Magnetic Signature Classification Using GRU-Based Recurrent Neural Networks
title_sort ship magnetic signature classification using gru based recurrent neural networks
topic Deep learning
gated recurrent unit
magnetic signature
signal processing
url https://ieeexplore.ieee.org/document/10947760/
work_keys_str_mv AT kajetanzielonacki shipmagneticsignatureclassificationusinggrubasedrecurrentneuralnetworks
AT jaroslawtarnawski shipmagneticsignatureclassificationusinggrubasedrecurrentneuralnetworks
AT miroslawwoloszyn shipmagneticsignatureclassificationusinggrubasedrecurrentneuralnetworks