CNN-Transformer Bridge Mode for Detecting Arcing Horn Defects in Railway Sectional Insulator
Jin He,Ran Duan,Mianxiong Dong,Yonggui Kao,Ge Guo,Jun Liu
DOI: https://doi.org/10.1109/tim.2024.3373084
IF: 5.6
2024-01-01
IEEE Transactions on Instrumentation and Measurement
Abstract:The sectional insulator with arcing horns is indispensable for electrified railways' overhead contact system (OCS). When these arcing horns are damaged, broken, or detached, arcing can occur due to unstable contact between the pantograph and the catenary, potentially causing burns and damage to OCS components or even complete system failure. Unfortunately, no specialized technique or method for detecting arcing horn defects exists. To address this critical issue, this article proposes a novel CTBM-DAHD network that utilizes convolutional neural network (CNN)-transformer bridge mode to recognize arcing horn defects in realistic application scenarios, such as rainy, foggy, sunny, and night-time conditions. Notably, the network can accurately detect obscured and long-range minor arcing horn defects, significantly improving recall and precision of defect recognition. The experimental findings demonstrate that, compared to the state-of-the-art networks, the CTBM-DAHD network achieves outstanding performance while maintaining lower computational costs and a reduced number of weight parameters. It surpasses the best CNN-transformer bridge fusion network by 3.5% and outperforms the dual-branch vision transformer by 1.5%. Moreover, the CTBM-DAHD network has been successfully deployed on over 500 high-speed trains in China, detecting more than 1000 arcing horn defects. These results affirm its effectiveness in recognizing arcing horn defects in complex and natural environments.