Evaluating Machine Learning and Deep Learning Algorithms for Predictive Maintenance of Hydraulic Systems

Citation

Al-Khulaqi, Ayat and Palanichamy, Naveen and Haw, Su Cheng and Raja, S Charles (2025) Evaluating Machine Learning and Deep Learning Algorithms for Predictive Maintenance of Hydraulic Systems. International Journal on Advanced Science, Engineering and Information Technology, 15 (1). pp. 52-59. ISSN 2088-5334

[img] Text
Ayat+216-AAP.pdf - Published Version
Restricted to Repository staff only

Download (1MB)

Abstract

Hydraulic systems are essential in industries like aerospace and petroleum. However, equipment degradation can lead to failures over time, resulting in costly downtime. Condition monitoring and predictive maintenance can be implemented to predict the equipment's failure before the machine's total failure. Current data-driven methods for predicting faults in hydraulic systems are insufficient due to inaccurate predictions. Our primary objective of this research is to investigate the potential of different classifier models’ predictive capabilities in enhancing the reliability of hydraulic systems. This paper implements two machine learning models (ML), random forest (RF) and categorical boost (Catboost), and one deep learning model (DL), long-short-term memory (LSTM), to predict the maintenance needs of hydraulic system equipment using the data from ZeMA gGmbH. The results of the models are evaluated through different metrics, such as Precision, Recall, F1-Score, and Accuracy. The outcomes of the experiments validated the paramount importance of the RF model, which has proven to be the most efficient and successful in accurately predicting the instances of equipment failure before the occurrence of total system failure. Critical hydraulic system condition components revealed their varying performance across different components, with LSTM excelling in predictiveness of the valve, RF dominating pump predictions, and overall reliability observed for Accumulator and Stable Flag. The experimental findings demonstrate that the proposed method for predicting the state of hydraulic systems outperforms alternative approaches.

Item Type: Article
Uncontrolled Keywords: Hydraulic systems; condition monitoring; predictive maintenance; machine learning; deep learning.
Subjects: Q Science > QA Mathematics > QA71-90 Instruments and machines
Divisions: Faculty of Computing and Informatics (FCI)
Depositing User: Ms Suzilawati Abu Samah
Date Deposited: 30 Apr 2025 01:21
Last Modified: 30 Apr 2025 01:21
URII: http://shdl.mmu.edu.my/id/eprint/13709

Downloads

Downloads per month over past year

View ItemEdit (login required)