AI-powered inspection system transforms Qinghai-Xizang Railway safety

An AI-powered inspection system that identifies defects in seconds is transforming Qinghai-Xizang Railway safety management at the Xining East Rolling Stock Depot in Qinghai province as the railway marks its 20th anniversary on July 1.

Using 5G to transmit high-definition images from trackside cameras, the system can detect more than 300 types of carriage defects in real time. A full freight train inspection now takes just three to five minutes, reducing inspectors' workload by 70 percent while raising detection accuracy to more than 98 percent.

Two decades ago, inspectors walked nearly 20 kilometers a day at elevations above 3,000 meters, enduring altitude sickness, freezing temperatures, and intense ultraviolet radiation as they manually examined every carriage component. The introduction of the TFDS trackside imaging system, a 3D detection system for freight cars based on point cloud data, in 2016 moved inspections indoors, laying the foundation for today's AI-powered system.

Engineers have adapted the technology to the plateau's harsh environment. Deicing equipment keeps cameras operating in temperatures as low as -40 C, reducing weather-related false alarms by more than 95 percent. Intelligent technologies have also been introduced to overhaul workshops and a big-data monitoring network linking Golmud, Nagchu, and Lhasa.

"Over the past two decades, railway safety management has evolved from manual inspections to intelligent risk prevention," Lan Yuzhu, director of the workshop, said. He said AI and big data will continue to strengthen the long-term safe operation of the world's highest railway.