Title: Research on PCB defect detection in intelligent ships based on hybrid CN-YoLov5

Authors: Yuchen Cui

Addresses: Department of Navigation Technology, Wuhan University of Technology, Wuhan, Hubei, China

Abstract: With the progress of intelligent ship engineering, Printed Circuit Boards (PCBs) have become indispensable in ship control systems. However, during PCB manufacturing and subsequent operation, various defects frequently occur, undermining their reliability and threatening the stable operation of ship control systems. Accurately identifying microscopic targets, balancing detection efficiency and precision and detecting small-sized and complex defects are significant challenges. To address these issues, this paper presents Hybrid CN-YoLov5, a novel PCB defect detection technique based on an improved YOLOv5s framework. It enhances the model's target feature - capturing ability by replacing the C3 module with the C3_SAC module and the Conv module with SAConv, and improves small-object recognition accuracy through the integration of the NWD loss function. The incorporation of the CBAM attention mechanism strengthens the model's feature extraction and overall recognition and classification performance. Experimental results show that compared with the original YOLOv5s, the mean Average Precision (mAP) of Hybrid CN-YoLov5 reaches 95.80% (an improvement of 2.50%) and the precision reaches 100% (an increase of 6.79%), indicating its effectiveness for PCB defect detection and great potential in the quality inspection and fault diagnosis of ship-based PCB systems.

Keywords: PCB defect detection; YoLov5; SAConv; global attention mechanism; NWD loss function.

DOI: 10.1504/IJRS.2026.150495

International Journal of Reliability and Safety, 2026 Vol.20 No.1, pp.91 - 111

Received: 25 Oct 2024
Accepted: 19 May 2025
Published online: 15 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article