Title: Aero-engine blade detection and tracking using networked borescopes

Authors: Shuai Liu; Wei Liang; Yinlong Zhang; Yanfei Wang; Yuanhao Liu

Addresses: State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China; Key Laboratory of Networked Control Systems, Chinese Academy of Sciences, Shenyang, 110016, China ' State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China; Key Laboratory of Networked Control Systems, Chinese Academy of Sciences, Shenyang, 110016, China ' State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China; Key Laboratory of Networked Control Systems, Chinese Academy of Sciences, Shenyang, 110016, China ' School of Information Science and Engineering, Shenyang University of Technology, Shenyang, 110870, China ' State Key Laboratory of Robotics, Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China; Key Laboratory of Networked Control Systems, Chinese Academy of Sciences, Shenyang, 110016, China; University of Chinese Academy of Sciences, Beijing, 100049, China

Abstract: To ensure aviation safety, monitoring aero-engine blade quality is essential. Traditional manual inspection using industrial endoscopes is costly and inefficient due to metallic reflections and blade similarity. This paper introduces an innovative end-to-end convolutional neural network (CNN) approach for detecting and tracking aero-engine blades, leveraging borescope vision to aid in blade quality monitoring. A variable kernel convolution module is developed to tackle the problem of detection inaccuracy caused by morphological changes in blades from various perspectives. Besides, an innovative three-dimensional attention mechanism is developed, which allows the network to effectively capture semantics across the blade's width, height, and channels. Finally, an optical flow-assisted blade tracking strategy is introduced to resolve identity recognition confusion due to high similarity between blades. The algorithm has been tested on the developed aero-engine blade inspection platform. The experimental results have validated the effectiveness of the proposed method.

Keywords: aero-engine blade detection; aero-engine blade tracking; convolutional neural network; CNN; endoscope sensor network.

DOI: 10.1504/IJSNET.2025.144656

International Journal of Sensor Networks, 2025 Vol.47 No.3, pp.148 - 161

Received: 12 Nov 2024
Accepted: 03 Jan 2025
Published online: 25 Feb 2025 *

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