Edge Intelligence-Based OCS Fault Detection in Rail Transit Systems
Xuetao Shi, Hongli Zhao, Hailin Jiang, Huijun Zuo, Qiang Zhang · Wireless Communications and Mobile Computing · 2023
The Overhead Contact System (OCS) is critical infrastructure for train power supply in rail transit systems. OCS state monitoring and fault detection are indispensable to guarantee the safety of railway operations. The existing human-based OCS state monitoring and fault diagnosing method has some inherent drawbacks, such as poor real-time capability, low detecting precision, and waste of human resources. Edge Intelligence (EI) can perform complex computing tasks offloaded from trains within a little delay, and it is believed to help empower the OCS. In this paper, we propose an EI-based OCS state monitoring and fault detecting system. The latest Computer Vision (CV) model YOLOv5s is used to detect the OCS faults using the collected images. Edge Computing (EC) is used to perform the CV model inference. The EC system receives the OCS images taken by the train cameras and calculates the real-time fault detection results. The consistency and scalability of running jobs on edge devices are also addressed in our approach. Extensive experimental results demonstrate that the proposed EI-based system can detect OCS faults in real-time. The adopted YOLOv5s achieves a high fault detection rate, outperforming other models.