Why is NS high on AI? To turn finders into fixers 

By Jeff Stagl, Managing Editor 

Norfolk Southern Railway is trying to instill a customer-first mindset throughout its organization, and a key foundational driver of that objective is adopting more technology. Or more specifically, artificial intelligence (AI). 

The Class I strives to leverage cutting-edge AI as well as other advanced technologies to set new industry standards for safety and operational excellence. 

Recently, something both encouraging and groundbreaking occurred with NS’ technological pursuits: AI employed at a digital train inspection (DTI) portal autonomously detected a hairline crack in a wheel. The railroad currently uses seven DTI systems at six locations to inspect rail cars on trains moving at operating speeds. 

The AI wheel-crack detection was a milestone in the North American rail industry, says Anil Bhatt, NS’ executive vice president and chief information and digital officer. Identifying such defects in real-world conditions is an immense challenge, and the breakthrough further strengthens the railroad’s ability to prevent potential train accidents and incidents, he says.  

“This is a powerful example of how AI and human expertise come together to enhance safety across our network.” — Anil Bhatt, NS EVP and chief information and digital officer Norfolk Southern Railway

“As soon as the flaw was detected, our railroaders immediately took action, removing the affected car from service and preventing a potential issue before it could escalate. This is a powerful example of how AI and human expertise come together to enhance safety across our network,” says Bhatt. 

It’s vital to develop capabilities that cater to customers’ needs, such as providing reliable service and becoming tech savvy, he believes. 

“What really matters is what you do on a daily basis. The ceiling is high here,” says Bhatt. “We [provide] an awesome customer experience led by technology and we are looking at areas where we can push the boundaries — AI is one area.” 

And AI is helping to push the capability limits of the DTI portals, which NS developed in partnership with the Georgia Tech Research Institution to identify rail-car defects faster. 

The portals are equipped with more than three dozen trackside cameras and stadium lighting that together forge machine vision Inspection technology designed to capture ultra-high-resolution, 360-degree images of passing rail cars that depict minute details. The cameras are synced to the sub-millisecond, taking 1,000 images per car as they pass through the portal at speeds up to 70 mph. 

“When wheels are inspected manually, it’s laborious and time consuming,” says Bhatt, adding that the DTI portals not only speed up the inspection process but help detect high-risk defects. 

AI analyzes the images generated at a DTI portal for potential defects and transmits information to NS’ network operations center. There, the data is reviewed — guided by a robust response protocol — by subject-matter experts who identify and address issues to proactively ensure rail operation safety. Any defect information is made available in a matter of minutes. 

The railroad currently uses seven DTI systems to inspect rail cars on trains moving at operating speeds. Shown: An image of a wheelset captured by a DTI camera. Norfolk Southern Railway

The advanced AI algorithms were developed by NS’ data scientists in close collaboration with the railroad’s mechanical experts. Each portal generates petabytes of data every month.  

“This vast dataset is processed in real-time by AI models that detect emerging conditions with precision, helping our teams take proactive action,” says Bhatt. 

And the AI tool is constantly learning while analyzing images for potential defects, even those taken in challenging weather conditions. Hundreds of years’ worth of learning in train dynamics has been converted into more than 40 algorithms, and more are in development. The algorithms have a very high accuracy rate and a low false alarm rate so defects can be detected sooner. 

“We are trying to make sure that our models improve,” says Bhatt. 

Since the initial cracked wheel was identified by AI, there has been a second case of a cracked wheel caught by the artificial intelligence, he says. 

Last year, 85 Tier 1 critical defects were identified and handled in near real time via NS’ 24-hour wayside desk. Overall, more than 25,000 mechanical maintenance-related defects were identified and handled. 

After adding five DTI portal systems in 2024, NS plans to place an additional three portal sites and three systems online in 2025. The railroad also is working with Georgia Tech to fine-tune the portals, such as to help identify trains that need servicing, says Bhatt. 

In addition, the Class I is striving to employ AI in other departments — such as legal and finance — to make them more efficient. For example, a pilot program is underway to upload a rulebook containing thousands of pages to an AI tool. A user can ask the AI to pinpoint the pages on any topic, such as shift management for a conductor, says Bhatt. 

The whole idea behind technology is that it can augment efforts with people, to work together to solve problems, he says. 

“AI is augmented intelligence,” says Bhatt. 

Since many employees no longer need to spend a lot of time trying to identify issues, they can spend more time solving them. The payoff to NS: avoiding potential safety incidents and network disruptions. 

“Now, we can turn the finders of defects into fixers,” NS officials say.