Scholarlink

In Vietnam, the railway industry has always been of interest to society, with its scale constantly expanding and upgrading. According to statistics, there are approximately 10,000 railway level crossings across the nation. At railway level crossings, system of protective equipment is deployed to ensure traffic safety. The operational data of these protective devices is collected and transmitted to the data center for monitoring and management. We developed and evaluated machine learning (ML) models to detect potential anomalies in crossing operations and identify potential safety risks. The system analyzes real-time sensor data including magnetic sensors, barrier status monitors, power supply indicators, and environmental sensors from multiple crossing locations. By using one class support vector machine (OCSVM) models to exploit the collected data, with the dual objectives of early detection of potential risks as well as the stable operation of the equipment we achieved 96% accuracy. This research proposes a methodology for improving the quality of management as well as the application of science and technology in the digital transformation and modernization of the railway industry in Vietnam.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Anomaly Detection for Vietnam Railway Using Unsupervised Learning Based on IoT Device Data Monitoring Railway Level Crossings

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ngo Anh Tuan,
  • Nguyen Nhu Hai Linh,
  • Pham Van Hiep,
  • Vu Thu Diep,
  • Phan Duy Hung

摘要

In Vietnam, the railway industry has always been of interest to society, with its scale constantly expanding and upgrading. According to statistics, there are approximately 10,000 railway level crossings across the nation. At railway level crossings, system of protective equipment is deployed to ensure traffic safety. The operational data of these protective devices is collected and transmitted to the data center for monitoring and management. We developed and evaluated machine learning (ML) models to detect potential anomalies in crossing operations and identify potential safety risks. The system analyzes real-time sensor data including magnetic sensors, barrier status monitors, power supply indicators, and environmental sensors from multiple crossing locations. By using one class support vector machine (OCSVM) models to exploit the collected data, with the dual objectives of early detection of potential risks as well as the stable operation of the equipment we achieved 96% accuracy. This research proposes a methodology for improving the quality of management as well as the application of science and technology in the digital transformation and modernization of the railway industry in Vietnam.