An intelligent diagnostic system for shipboard equipment using the mode decomposition and machine learning methods

Read full text Summarise with VIRAT
Authors: Lyubov Tyndykar
Cited by: 0
References: 20

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This study presents an integrated signal processing and machine learning framework for automatic status prediction of shipboard equipment. The research investigates the utilization of advanced mode decomposition techniques, with a particular focus on the integration of the Hilbert- Huang Transform (HHT) with state-of-the-art machine learning algorithms. The proposed approach is based on adaptive time-frequency decomposition, employing HHT methods in conjunction with Variational Mode Decomposition (VMD) algorithms. These techniques enable precise analysis of the functional characteristics of operating modes while mitigating the adverse effects of noise contamination. The decomposed components are used as input for neural network training, with careful selection of network architecture to identify characteristic malfunctions of shipboard equipment. Special emphasis is placed on the VMD-KAN-LSTM hybrid architecture, which combines preprocessing based on VMD using the Kolmogorov- Arnold Network (KAN) with the Long Short- Term Memory (LSTM) model. This hybrid architecture effectively captures non-linear interactions between system components and temporal dependencies inherent in the data. The efficacy of the proposed methodology was evaluated through experimental validation using data from a marine reciprocating compressor. Comparative experiments, employing LSTM combined with VMD-KAN and alternative methods, demonstrated that the integration of VMD significantly improves classification accuracy in the presence of interfering noise. Notably, the VMD-KAN-LSTM architecture exhibited the highest diagnostic accuracy among the models tested. The findings emphasize the benefits of decomposition methods for monitoring and predictive maintenance of shipboard equipment and demonstrate that combining time-frequency analysis with machine learning provides a robust approach for ensuring operational reliability and longevity of critical marine systems.

Fields of Study & Subjects

Machine Fault Diagnosis Techniques Technical Engine Diagnostics and Monitoring Fault Detection and Control Systems

Publication details

Journal / source
Vestnik Gosudarstvennogo universiteta morskogo i rechnogo flota imeni admirala S O Makarova
Volume
18
Issue
1
Pages
139–151
Publication date
24 March 2026
Language
EN