How AI Will Impact the Rail Industry

There are over 140,000 miles of railroad tracks spanning the US, transporting people, goods, and critical materials. We’ve all ridden on a train or counted the minutes waiting for one to pass by—but have you ever considered how and by whom railroads and trains are maintained? Guidehouse Insights believes AI and advanced analytics will revolutionize the inspection and monitoring of railroads, offering significant opportunities in terms of cost savings, worker safety, and efficiency gains. 

The magnitude of the global railroad network, and the age of installed infrastructure, requires constant and thorough inspections and maintenance. The cost to maintain railroad tracks and equipment is steep, with 20%-33% of US rail operating revenue reinvested in CAPEX and maintenance. Compared to other modes of transportation, this is one of the highest revenue percentages required to maintain the network, amounting to almost $25 billion annually.

Inspection Methods Improved by AI

Traditional railway and railcar inspections require on-track occupancy, meaning specialized equipment, operated by a worker, must be on-track to perform tasks. These inspection methods are costly from an equipment and labor perspective and present safety concerns for workers. On-track inspections also often result in logistics issues, delays, and unscheduled downtimes.

Using traditional methods, an estimated 20 miles of track can be inspected per day. Understanding this constraint, in 2020 the Federal Railroad Administration (FRA) revised rail integrity and track safety standards to reduce costs, incentivize innovation, and improve safety for the rail industry and the public. As part of this revision, the FRA approved several new, improved inspection methods, such as continuous rail testing. These methods can allow up to 80-160 miles of track to be evaluated per day.

As regulations evolve, the use of innovative technologies including AI, machine learning, and machine vision for completing and automating rail inspections has increased. Machine vision—enabling sensors and cameras installed either on the track or railcars—is used to autonomously inspect various components of the cars as they pass by. Traditionally, railcars would be removed from service and technicians would manually inspect their integrity by climbing on the roof. Using advanced image analysis, railcars can be continuously inspected, thereby reducing train downtime and the risk to technicians.

Machine learning and neural networks have also been used for detecting faults and defects in images of tracks, wheels, and cars. Using an extensive library of images, trained machine learning algorithms can flag defects on collected data for review by technicians or automate maintenance requests for immediate crew attention. Machine learning’s ability to test previous assumptions, analyze outcomes, and refine the model based on prior outcomes drives greater accuracy and will be an increasingly important tool for rail-based inspections going forward.

AI and the Future of Railway Inspection

Rail inspections are becoming increasingly automated with more frequent data collection happening across the network. The FRA's revised regulations have focused on the accurate and timely collection, transmission, and analysis of data and on maintaining its integrity. The use of advanced analytics solutions can improve the entire process from gathering high resolution data to analyzing it.

The focus on developing innovative inspection methods and data quality outlined in the revised regulations indicates that the rail industry is on the brink of a digital transformation. The demonstrated use of AI and advanced analytics prove that this technology will be critical for aiding the industry on the path toward digitization while providing financial and safety benefits to the rail industry and the public.