In a story from The Wall Street Journal, reporter Jared Council writes that American freight railroads are increasingly deploying artificial intelligence (AI) to further improve the safety and efficiency of the nation’s rail network.
AI, the story notes, allows railroads to more safely and efficiently inspect and monitor track and equipment, including the diagnosis of track, ballast and railroad tie conditions. Harnessing the power of technology to supplement traditional inspections, generally done by workers walking the tracks, is especially valuable due to the sheer scope of a 140,000-mile network, which is almost always in use by railroads as they move the economy.
“BNSF, which operates 32,500 miles of track in the western U.S. and Canada, uses AI to identify defects on the wheels of its locomotives and cars,” the story says.
“Specifically, it uses machine learning to analyze more than 750,000 images of wheels per day, captured by seven stationary ‘machine vision’ systems that use RFID tags to identify which trains have defective wheels. Company officials said this application was deployed last October and is utilized on about 250 trains a day. Since that point, a company spokeswoman said, ‘We have identified about 30 confirmed cracked wheels that would have otherwise been harder to detect’.”
The story also spotlights that “CSX operates roughly 21,000 miles of track, primarily in the eastern U.S., and it uses AI to identify wooden rail ties — the support beams that run perpendicular to the rails — that will likely need replacing soon.”
“Ties that are decaying, for instance, can throw off the alignment of rail tracks, which in turn can cause derailments,” the story reads. “Using radar and other scanning equipment attached to specialized vehicles that travel about 20 miles per hour, CSX captures images of rail ties and uses computer vision to grade the health of the ties based on decay, shape irregularities or other issues. Kathleen Brandt, CSX’s senior vice president and CIO, said the company plans to work with vendors to enhance the AI application by feeding it images from X-ray scans that can ‘look inside the tie for health conditions.’ CSX officials they believe the application has helped decrease infrastructure-related train derailments. The company reported four such derailments to the Federal Railroad Administration in the first quarter of 2018 and none in the first quarter of this year.”
Norfolk Southern is also part of the action, using “machine-learning algorithms to help predict which of its locomotives are likely to experience ‘cooling water’ leaks, which could lead to on-track breakdowns and delay other trains scheduled to use that track. It does this by analyzing real-time mechanical data, such as water pressure, as well as maintenance history to point out which locomotives are at high risk of a leak in the next 24 to 48 hours. Lost cooling water is a top cause of locomotive failure, the company said, downing 415 of its roughly 3,300 locomotives in 2018.”
Smart Public Policy Helps Spur Rail Innovation
The thesis of the story — that technology aiding workers is more effective than the human eye alone — supports the rail industry’s request for more use and acceptance of automated track inspection. Automated track inspection systems provide an objective method to evaluate track conditions and to identify defective conditions in the track or conditions that could lead to defects in the track. The story also underscores the need for a modern regulatory paradigm for freight railroad operations and validates the railroads’ call for policy rooted in a sensible principles:
- Barriers to deployment must be overcome outside of waivers.
- Railroads and railroad equipment manufacturers should be permitted to create voluntary standards.
- Regulations should be performance-based, rather than prescriptive.
- Regulations must be federal to avoid a patchwork of state and local rules.
- The Federal Railroad Administration must engage in the process.