The improved multi-scale convolution method for fault diagnosis of civil aircraft hydraulic systems based on QAR data
- 20 March 2025
- journal article
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 36 (4) , 045006
Abstract
As a core system of the aircraft, the stability, response speed, and efficiency of the hydraulic system are directly linked to flight safety, control performance, and operational reliability. Hydraulic system failures can lead to catastrophic accidents. This paper proposes a diagnostic method for civil aircraft hydraulic systems based on quick access recorder (QAR) data. First, based on guidance from an airline’s maintenance records, eight typical hydraulic system faults were identified. Next, under the guidance of the hydraulic system’s working principles, 28 sensor parameters most relevant to hydraulic system faults were selected from over 3000 parameters collected by the aircraft’s sensors. Subsequently, using preprocessed QAR fault data, an improved deep learning model based on multi-scale convolution (MSC) was developed to diagnose hydraulic system faults. The model incorporates an enhanced MSC design, which employs cross-channel convolution on top of MSC and integrates a multi-head attention mechanism, enabling the model to extract multi-dimensional features from QAR data more quickly and accurately. Experimental results on the QAR fault dataset demonstrate high prediction accuracy and fast diagnostic speed, achieving an accuracy of 98.2%. This model effectively and accurately diagnoses hydraulic system faults, contributing to flight safety and improving maintenance efficiency.Keywords
Funding Information
- National Natural Science Foundation of China (52172387)
- Joint Fund of National Natural Science Foundation of China and Civil Aviation Administration of China (U2033202, U1333119)
- Postgraduate Research & Practice Innovation Program of NUAA (xcxjh20230708)
- Aeronautical Science Foundation of China (2022Z071052001)
- Fundamental Research Funds for the Central Universities (ILA220321A22)
This publication has 34 references indexed in Scilit:
- A multi-rate sensor fusion and multi-task learning network for concurrent fault diagnosis of hydraulic systemsDigital Signal Processing, 2024
- A fault diagnosis method for hydraulic system based on multi-branch neural networksEngineering Applications of Artificial Intelligence, 2024
- Rotating machinery fault diagnosis based on one-dimensional convolutional neural network and modified multi-scale graph convolutional network under limited labeled dataEngineering Applications of Artificial Intelligence, 2024
- A fault diagnosis method with AT-ICNN based on a hybrid attention mechanism and improved convolutional layersApplied Acoustics, 2024
- A rolling bearing fault diagnosis technique based on recurrence quantification analysis and Bayesian optimization SVMApplied Soft Computing, 2024
- Multi-source heterogeneous information fusion fault diagnosis method based on deep neural networks under limited datasetsApplied Soft Computing, 2024
- Superimposable neural network for health monitoring of aircraft hydraulic systemEngineering Failure Analysis, 2024
- Failure analysis of aviation hydraulic rotary jointsEngineering Failure Analysis, 2023
- Attention-based long short-term memory fully convolutional network for chemical process fault diagnosisChinese Journal of Chemical Engineering, 2022
- Analysis of premature failure of aircraft hydraulic pipesEngineering Failure Analysis, 2020