An automatic detecting method for multi-scale foreign object debris in civil aircraft manufacturing and assembly scenario

Minghao Yu (), Qijie Zhao (), Sheng Cheng (), Hongxia Cai () and Lilan Liu ()
Additional contact information
Minghao Yu: Shanghai University
Qijie Zhao: Shanghai University
Sheng Cheng: Commercial Aircraft Corporation of China Ltd
Hongxia Cai: Shanghai University
Lilan Liu: Shanghai University

Journal of Intelligent Manufacturing, 2025, vol. 36, issue 8, No 28, 5835-5857

Abstract: Abstract Foreign Object Debris (FOD) is a critical aviation safety concern, and the detection of FOD has been highly prioritized by aircraft manufacturers. Currently, FOD detection is primarily conducted through manual vision inspection. However, due to the extensive range of FOD types and scales, manual inspection is susceptible to overlooking and misidentifying some small-scale FOD. To handle these difficulties, a lightweight model for detecting small-scale FOD is proposed based on YOLOv5s. This intelligent FOD detection method is of great significance in promoting the aircraft intelligent manufacturing industry. Firstly, we design the backbone feature extraction network (LSGK-GhostNet), which can automatically extract multi-scale features of FOD while reducing computational resources. Secondly, the Deformable Receptive Field Module (DRFB) is proposed by imitating human vision. It utilizes a three-branch structure to integrate various receptive fields for extracting effective FOD scale and contextual information. Additionally, it dynamically adjusts the spatial distribution of the receptive fields. A refined module (DFPN) is designed to amplify the details of small-scale FOD. In addition, the head of the model for detecting the small target is decoupled to strengthen the localizing accuracy for small-scale FOD. Finally, to improve the detection precision of dense detection and address unbalanced sample training, this paper improves the post-processing method and the localization loss of FOD. Detection experiments of FOD are conducted in laboratory and assembly scenarios, The results indicate that our algorithm notably outperforms other methods, and achieves an average accuracy of 94.1%, with a speed of 34.8 Frames per second (FPS) across seven FOD categories. This paper proposes a pioneering way to detect FOD automatically in the assembly process of civil aircraft.

Keywords: Foreign object debris; Intelligent manufacturing; LSGK-GhostNet; DRFB; FOD detection (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s10845-024-02508-x Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:joinma:v:36:y:2025:i:8:d:10.1007_s10845-024-02508-x

Ordering information: This journal article can be ordered from
http://www.springer.com/journal/10845

DOI: 10.1007/s10845-024-02508-x

Access Statistics for this article

Journal of Intelligent Manufacturing is currently edited by Andrew Kusiak

More articles in Journal of Intelligent Manufacturing from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().