By the Wayside

Written by Marybeth Luczak, Executive Editor

RAILWAY AGE, DECEMBER 2021 ISSUE: Wayside inspection equipment is helping railroads ensure that rolling stock operates safely and efficiently.

Machine vision and artificial intelligence (AI) are among the tools enabling railroads and transit agencies to automatically and proactively monitor rolling stock condition—identifying mechanical problems before they become serious failures. Transportation Technology Center, Inc.; Duos Technologies; Trimble® Beena Vision® Solutions; and Cogniac shared with Railway Age how it’s done.

Machine vision camera technology captures pictures of moving railcars from a variety of angles. With a Duos Technologies inspection system, for instance, the images are processed through trackside AI servers that look for railroad-defined issues such as engaged brake pistons. “The camera images that contain the brake piston are processed through the algorithm,” Duos Chief Commercial Officer Scott Carns explains. “The results are pushed from our system to the end user. This all happens for each railcar in milliseconds as each passes through the system at track speeds.”

All Class I railroads have machine vision systems on their networks, ranging from brake shoe and wheel inspection systems to complete inspection portals “that image the entire train in 360 degrees,” TTCI Scientist Matt Witte says. Among the benefits: “superior and more consistent performance of repetitive inspection tasks compared with the inherent human limitations. We want to utilize our skilled workforce where they can provide the most benefit in repairing rather than inspecting railcars.” The inspection systems also provide greater throughput and have the potential to boost average train velocity by reducing yard dwell time for manual inspections. 

— “The completeness of the data stream from machine vision inspection introduces the ability to trend component condition,” says Matt Witte, TTCI. —

“The completeness of the data stream from machine vision inspection introduces the ability to trend component condition,” Witte adds. “The end result is a way to predict the point in time where preventative maintenance can be performed on a railcar just before the end of its useful life is reached.” 

Other advantages include reducing safety risks to personnel, providing statistics for compliance, and maintaining fleet-wide visibility of component condition, Trimble Beena Vision Director of Sales Americas Timothy Francis points out.

Of course, there are technology challenges as well. “Development of algorithms to automate defect detection are not always easy in the railroad environment,” TTCI’s Witte explains. “Variations on the designs of cars and components make observation, detection and evaluation of baseline acceptable condition a much greater challenge than detecting product defects in a manufacturing environment where machine vision inspection is common. And reliable inspection requires a clear view of the components to be inspected. Common variations in railcar hardware, such as placement of hand brake components, can block the optical pathway between the target and the camera. When coupled with adverse environmental conditions such as rain or snow, this obscured line of sight can affect the performance of the machine vision system.”

Duos Technologies

Railroads are using wayside inspection technology like Duos’—which incorporates a canopy to help mitigate environmental factors—to increase safety and improve velocity, but application ultimately boils down to their specific pain points, Scott Carns says. And system positioning tends to fall well outside of rail yards and instead on main line track, he adds. “Our metrics have shown that a good rule of thumb is approximately one system per 1,000 route-miles.”

How are railroads capturing data? “With the volume of data generated from our systems, we currently deploy trackside data centers,” he says. “All of the data is captured, processed and stored at the edge. From there, we have developed Application Programming Interfaces (APIs) that integrate directly into existing railroad systems for maintenance reporting and bad order tagging. Currently, because this type of technology is very cutting edge, the general philosophy is to leverage the systems as a new data source into existing processes throughout their organization.”

Duos’ latest technology is ObliqueVUE, a crosstie-mounted system that captures bi-directional images from eight cameras and “sees” at least 25 new inspection points, Carns notes. “We’ve also started implementing the latest in 8K line scan and 5MP area scan cameras that allows us to image at speeds above 125 mph.”

What’s in the pipeline? For one customer, Duos is currently expanding the number of imaging perspectives from nine to 34, allowing it to “see” more than 100 inspection points on each railcar. 

“Another very exciting opportunity is the creation and near-term implementation of a ‘fast lane’ for railroad border crossings,” Carns says. “The intent here is to provide all of the inspection data for various federal agencies in partnership with the railroads to allow them to change the current process on how trains transition across our land border crossings in the United States. Ultimately, it will substantially improve velocity, allow higher throughput and assist the various law enforcement agencies responsible for this effort.”   

Trimble Beena Vision Solutions

Trimble’s non-contact wayside measurement and inspection technologies assess rolling stock condition from component level to full train inspection. Rail operators use the generated data to prioritize train maintenance and derailment prevention, Timothy Francis says. “Furthermore, the correlation between data generated by different detectors provides additional opportunities to analyze and understand the condition of any individual rolling stock or the whole fleet.”

Uptake of condition monitoring continues to grow, Francis says, noting that the company has installed numerous systems at all of the Class I’s and at multiple worldwide operators, such as Aurizon, BHP, Rio Tinto, and FMG in Australia; SNCF in France; VR in Finland; Vale, VLI and MRS in Brazil; and Etihad in United Arab Emirates. 

Trimble’s latest offering is Trimble® TreadView®, an automatic non-contact optical inspection system that images and inspects the wheel surface—wheel tread, flange and plate surface areas—at main line operating speeds and in all ambient light and weather conditions. “The high-resolution images and high-density 3D data of the wheel surface are used to determine any external surface abnormalities of the wheel tread,” explains Francis. “Processed data and images from the Trimble TreadView system are integrated into the Trimble CMMS™ (Condition Monitoring Management System) software to provide web-based access for data visualization, alarm management and data analytics.”  

Deriving actionable insight from the wayside data is the key to improving railroad efficiency and reliability. “Via faster enhanced data preparation and contextualization, and the application of intelligent analytics, data can be modeled to show trends and patterns that would otherwise remain hidden,” Francis says. “Trimble is working within the rail industry to create meaningful predictive analytics that reduce risk and boost efficiency.”


“Cogniac’s solution allows railway companies to evaluate, in near real time, train wheels and tracks at speeds of up to 60 mph,” says the company, which deploys its AI machine vision platform “to enable railroads to process the images they capture on location and within a minute, providing near-instant evaluation of critical assets on the move.”

Cogniac is currently working with a Class I, and, on a monthly basis, monitors 22 million wheels and 32,500 miles of track. The company “processes hundreds of thousands of images taken by cameras on the front of more than 450 moving trains each month. The images are sent  by Cogniac’s EdgeFlow and evaluated for splits, cracks or missing bolts. A human subject matter expert is alerted if any images are flagged as defective. The human supervisor can then make a decision on how to proceed with the defect identified.”

— “Since January 2020, Cogniac’s vision system has stopped more than 100 trains where there was a potentially devastating issue that could lead to a derailment, saving the railway up to $350 million in damages.” —

Cogniac has also worked with this railroad to install trackside gantries, taking high-resolution wheel images. The images are processed, and those showing cracks or other issues are sent to a human inspector for secondary review. “Since January 2020, Cogniac’s vision system has stopped more than 100 trains where there was a potentially devastating issue that could lead to a derailment, saving the railway up to $350 million in damages,” Cogniac says.

What’s Next?

“The next big hurdle for the industry is regulatory modernization,” TTCI’s Witte tells Railway Age. “Historically developed regulations are not necessarily structured to take full advantage of the safety, repeatability and efficiency benefits offered by automated inspection. Through the AAR committees, the railroads are working to demonstrate and quantify the safety benefits of automated inspections. TTCI supports this effort by objectively evaluating the performance of inspection systems in a controlled environment where known defects can be included safely in the population of test subject railcars.”

Tags: , , , , ,