Episode Ten – The Leading Edge of Rail

Discover how edge and artificial intelligence are now just as important to the railroad industry as tracks and wheels

Discover how edge and artificial intelligence are now just as important to the railroad industry as tracks and wheels.

The advent of the railroad in North America connected towns, cities, states and the continent – delivering goods and services from sea to shining sea. However, that was over two centuries ago. Today, the 21st century rail industry is embracing new technology to enhance safety, reduce downtime and stay competitive through the use of artificial intelligence and edge technology. In this episode of Technology Powers X, we explore how;

  • The prior, manual processes for railroad operations led to inefficiencies and unnecessary downtime
  • Duos Technology is transforming the industry through leveraging artificial intelligence at the edge to inspect their railcars at full speed – resulting in a process that’s 120x faster than prior methods
  • Railroad operators are turning railcar ‘finders’ into ‘fixers’ which results in an industry that’s smarter, safer and competitive

Additional Resources

Dell Solutions with Intel®‎

Guest List

  • Allan Zarembski Professor of Railroad Engineering, University of Delaware
  • Scott Carns Chief Operating Officer, Duos Technologies
  • David Ponevac Chief Technology Officer, Duos Technologies
  • Derrick Schmenk Director of IT, Duos Technologies
  • CK Kumar Director of PowerEdge Server Marketing, Dell Technologies

Danielle Applestone: Disruptive technologies change lives and livelihoods. Some make history, others become history, but only a few have literally changed the American landscape. Railroads began redrawing the American map nearly two centuries ago. With the advent of rail, a trip that might require a long meandering day on a canal boat could be made in an hour. Steel rails connected towns and cities, then states, then a continent. Railroads also changed our relationship with time. Time wasn’t working for railroads. So they changed it. They adopted the use of standard time with its four zones across the continent, Eastern, Central, Mountain and Pacific years before it became US law. Trains and timepieces were woven together in folklore, grand clocks became centerpieces in palatial marble train stations. A pocket watch was as much a part of a conductor’s uniform as their pillbox hat. Trains raced against time, setting and shattering records. To generations of children, the engineer was a near mythical figure. Then in the last half of the 20th century, post-war society underwent rapid change. Highways appeared, literally paving the way for the rise of trucking and car travel and yet rail endures, impacting the everyday lives of almost every American. It was bold technology that built the railroads and today, it’s technology that’s keeping them relevant. Take a seat in first class and I’ll show you how.

I’m Danielle Applestone, engineer and entrepreneur, and I love a good pair of overalls. This is Technology Powers X, an original podcast from Dell Technologies. In this episode, technology powers 21st century rail. It’s hard to imagine life as we know it without railroads. I’m pretty sure the only reason my tiny hometown persisted was because the Mena Kansas City-Southern Depot was built there in 1920 and today along 200,000 miles of track throughout North America, they’re part of everyday life.

Allan Zarembski: Much of the food you eat is brought by train. The railroads are the biggest mover of grains in the United States, including grains for export.

Danielle Applestone: Allan Zarembski is professor of railroad engineering at the University of Delaware.

Allan Zarembski: The iron ore that’s used to manufacture your various steel products, almost guaranteed moved by train, manufactured goods, refrigerators, ovens, anything that’s big and bulky, almost guaranteed moved some or part of the way by train, if not most of the way by train with only the last maybe a couple of miles going by truck.

Danielle Applestone: Both time and railroads continued to change, yet today as in the 1820s, trains build their value on time and constant motion. Scott Carns is chief operating officer for Duos Technologies.

Scott Carns: One of the metrics the railroad uses is called velocity. All railroads measure their velocity, whether it’s a transit railroad or a freight railroad. And as long as their vehicles or trains are moving, they’re profitable. When they have trains that are not moving and they’re stopped and they’re being repaired or they’re parked for whatever reason for a duration, that’s referred to as dwell time, but that costs them money. It costs them money in the fact that their prime vehicle is not operational, but it’s also the personnel that are involved to service and maintain those rail cars are occupied doing that.

Danielle Applestone: Enter technology. In some of the loneliest stretches of rail on the continent, a number of intriguing new arch shaped structures are springing up, destined to become as much a part of railroads as trestle bridges and mountain tunnels. Their purpose, to provide comprehensive inspection of trains without there needing to stop or even slow down. Scott Carns.

Scott Carns: I would best describe it as a very large barn. It’s a 40 by 60, 40 foot wide, 60 foot long by 42 foot high structure that straddles the rail track. So it’s almost like we’re building a tunnel that this train drives through. Inside the tunnel, we have a trust structure and the trust structure holds all the camera equipment, all the lighting for illumination and all of that is then interconnected into the small data center that we put there. The data center is actually an eight by 12 bungalow that’s fully environmentally controlled, has two cabinets, essentially stacked full of Dell server equipment and storage subsystem. So we have the PowerEdge servers. We have the PowerVaults in there. It’s very interesting because we are capturing images of a train, a basic system has nine different camera angles of a rail car that we pick. One rail car, the pictures, essentially, it’s almost a terabyte of information, just that single rail car. And so you take that times 150 cars, that’s one train transaction. And so then you multiply that times 40 trains a day going through that system, we have huge storage subsystems that are inside there. They’re holding all this information, in most cases, up to a period of 6 to 12 months.

Danielle Applestone: Within that portal, the immediate task is to provide concise, actionable feedback on the train’s condition. Are the wheels doing all right? How are the brakes? Is their oil leakage? Are any wheels chipped or worn? As much as steel wheels on rails, information has become the commodity that keeps trains moving. David Ponevac is chief technology officer at Duos.

David Ponevac: We basically built a data acquisition system. That’s the most low level way of looking at it, where we capture image data, vibrational data, laser data, LIDAR data, anything that we can get our hands on, and we try to make sense of it. We try to come with an actionable intelligence, actionable event out of that data.

Danielle Applestone: When Duos began its corporate life, an early task was to design a means of operating drawbridges remotely. Then one day Scott Carns got a call.

Scott Carns: A friend of mine at Union Pacific Railroad in Omaha reached out to me back in I think it was 2007, and he had a problem where US Customs was randomly stopping their freight trains to conduct inspections about 90 miles East of El Paso, Texas. Well, train tracks are static. They don’t move. There’s no alternate routes or getting off a ramp and moving the train this way. So when you stop a train in Texas, it impacts train traffic nationwide. And so my friend asked me, he said, hey, can you guys come up with a way to help US Customs do these inspections without stopping our freight trains? And so we developed the first generation of the rail car inspection portals for that application to help them do essentially contraband and human smuggling inspections in Texas.

Danielle Applestone: A technology to prevent downtime attracted interest throughout the rail industry. Suppose it could be used to inspect trains while they’re moving. It was intriguing and it was no small ask. David Ponevac.

David Ponevac: You’re trying to teach an algorithm to behave like a human and to look for things the way human would. And back then, there was just no good way of doing it or no reliable way of doing it. We were getting accuracies anywhere between 60 to 80%, which is good, but it’s not as good as the customer would like. So we eventually throughout the journey as we went through blob analysis, histogram analysis, all these old types of approaches, AI popped up on our radar and we said, well, this is a way to go. This is the way of the future. So we tested on a small sample set of data, and we started experimenting with AI and we were getting 95% accuracy. We were getting very decent results and this has shown us that AI is the way.

Danielle Applestone: The best way to appreciate the benefit of this no stop, tech savvy, AI-driven inspection portal is to understand the traditional painstaking inspection procedure. Scott Carns.

Scott Carns: Traditionally, a rail car inspection when a train arrives at an inspection point, which is typically a rail yard, and picture a train is on average 150 rail cars. Each rail car is approximately 60 feet long and 20 or so feet tall. And they have a crew of people that walk around the train when it’s parked with flashlights and chalk, and they look for defects. They have a checklist that they have to go through and there’s literally hundreds of points on a rail car that get just visually inspected. And when they find a defect, they take that piece of chalk and they mark it or circle it on the car and then they write it down and say car XYZ123 needs repair and then it goes into a repair shop and the technician will come out and fix that defect on the car. And so that’s the traditional methodology of rail car inspection.

Danielle Applestone: In railroad lore, the ballad of John Henry tells the story of the steel driving folk hero hammer in hand, engaged in a man versus machine contest against a steam drill. Recently, Duos staged its own 21st century version of the contest.

Scott Carns: We actually did a test last summer up in Canada, where we took a human and then we took our system and literally sat side by side and said, go inspect. So when the human completed his inspection on one car, we had inspected 120 cars. And so right there shows the disparity between the human inspection and an automated inspection using technology. And so the advantage of us taking our system, our Railcar Inspection Portal, you immediately see an improvement in the reliability and the velocity of what you can inspect. So literally in that eight minutes that it takes somebody to walk around a car with a piece of chalk, we do 120 cars in that same period of time. You dramatically improve the process overall.

Danielle Applestone: One of the apparent oddities of the Duos inspection portals is their remote locations. Why would they be so removed from major stations and repair facilities?

Scott Carns: Because that’s where the trains are. That’s where the tracks and the trains are. It made no sense to process this data in a rail yard, which would have resource and communications infrastructure there, because at that point it’s too late, so you lose the benefit of doing the early analysis or inspection of the train. The strategy of deployment is that we strategically position these systems approximately 50 to a hundred miles away from an inspection point so when the train does arrive at the inspection point, they already know what they need to do to fix it as opposed to spending the 8,10,12 hours per train that it takes to conduct an inspection and then start repairs.

Danielle Applestone: Considering the massive amounts of data gathered at the portal site, a question arises. How do you get all that information from such remote locations to the nearest maintenance facility? Derrick Schmenk is director of IT at Duos.

Derrick Schmenk: We have evaluated. It cost gets exponentially higher. The infrastructure to support the internet circuit connectivity required is essentially not available at this current time. And because of that availability, you would be set to install that connectivity yourself, which would entail running miles of fiber cable. So in addition to that, you’re looking at a network security firewall routers, etc., that have that same capability as far as bandwidth. So all that exponentially increases the cost.

Danielle Applestone: The answer was to use edge technology, storing and processing the data on site at the edge.

Derrick Schmenk: We utilize a repurpose shipping container that is built by a company Edge Presence that is essentially a data center in a box. When you walk through the door, you’re going to see two 45U cabinets slapped full of servers and storage and power control systems, etc. Our most important essentially server is utilized in a PowerEdge XR2, which is what we utilize for our image acquisition servers. The reason that we use that is because it’s a military spec server that can take on vibrations, extended temperature parameters, and other environmental factors that can affect that process.

Danielle Applestone: The task of creating devices every bit as durable as they are complex presents its own set of challenges. CK Kumar is director of marketing for PowerEdge servers at Dell Technologies.

CK Kumar: It’s a hard process. Think about the trains when they’re moving, they’re moving whether it is winter, summer, it’s raining, it’s fog and mist. But what happens then is when you put technology around that, the technology is now subject to the same weather conditions. Now you have to make the technology weatherproof and still be able to function extremely well and continuously, and that’s a hard problem to solve by itself. But beyond that, the ability to take this high volume of data and be able to quickly analyze it and get the right kind of insight to take action on it.

Danielle Applestone: An operative word there is quickly and there in lies the benefit of edge technology. In order for the insights gathered to be a maximum value, timing is everything. David Ponevac.

David Ponevac: And the difficult part, the complexity is that we need to process all the data within five minutes of the train leaving the portal. So your usual train will vary between 100 rail cars to 200 rail cars. So within five minutes, you not only have to acquire all the imagery, you also have to process all the imagery and then the processed data, the actionable events have to be collated, put together and transmitted to the customer because there might be a defect that can cause a derailment and they need to stop the train right now, right there then. So, that is the complexity. That is where most of the computational power needs to be from collection to transmission across the network to inference and to the final actionable events. So the sensors that we run. If you think about your regular, your run of the mill website that we put together for a customer, you’re talking about a few dozen of cameras, lasers, radars, vibrational sensors. So the amount of data is quite large on the image side on a slow day, and this is per site, you’re talking about 700 gigabytes of images alone that are generated. Altogether, you’re probably talking about, 1.2, 1.3 terabytes of data that we have to analyze every day for each site. And this data is the retention period, it varies from customer to customer. For certain customers, we have to keep it for two years, some only needed for 90 days. So the storage requirements vary, but they are significant.

Danielle Applestone: Just as data, AI and technology are affecting positive change in railroads, they’re also having a positive impact on operations. Scott Carns.

Scott Carns: Yeah. The biggest operational benefit that the railroads are pushing is they’re augmenting their workforce by taking the guys that were finders, they’re finding the problems to changing them into fixers. And so by doing that, you’re maintaining your labor force, but you’re also increasing the velocity of your system of your network. A railroad, as I mentioned earlier, they’re static tracks. Every railroad owns a fixed amount of route. It’s very similar to a computer network. If you look at a map, it’s basically a network of hub and spoke of tracks and the faster they can make their trains go, just like you want data to go fast across your network, the more efficient and a more profitable they are in the long run.

Danielle Applestone: As both railroads and technology change and evolve, the inevitable question is what’s next for this sort of data-based maintenance technology? For David Ponevac, it distills down to two words, more and faster.

David Ponevac: You went from using people to analyze and look for defects to using machines. Now we are only on the forefront of it all. Right now, we have maybe 30, 35 algorithms that we are doing concurrently. There’s hundreds that we needed to do. Hundreds because there’s hundreds of things that can go wrong on each particular rail car. So the future is to do it faster and do more of it. And that’s the way I’m looking at it. And that’s the way the industry is heading and our technology is following that as well, where we want to replace all the manual inspection points, where we want to basically replace everything that has a possibility of defecting and use AI to spot those issues automatically at track speed.

Danielle Applestone: The principle of rail travel remains simple, steel wheels, steel tracks, engine, brakes, signals. Yet, the engineering and the technology behind railroads are neither simple nor static. As Professor Allan Zarembski observes, their evolution is vital to the industry’s survival.

Allan Zarembski: Can we run railroads the old-fashioned way? Sure. But number one, it’s going to be much more expensive and railroads have to compete like everybody else against trucks. I mentioned that nowadays, a lot of the goods, like the containers, if they’re moving more than three, 400 miles, they’re going to go by rail. If railroads were not as efficient as they were, then maybe the trucks would move, their breaking point may be 1,000 to 1,200 miles because trucks have their own efficiencies.

Danielle Applestone: The juggling act of railroad management with its conflicting tensions of efficiency, safety, and competition is made that much trickier when you consider how many people, how many businesses depend on its smooth operation and how any expected downtime can cause a supply chain reaction. CK Kumar.

CK Kumar: If you think about the transportation of the logistics industry, especially when it comes to rail cars, freight trains, the system needs to keep running. They cannot stop. And every minute that the traffic is not moving tends to adversely impact, not just the business, but it also impacts several industries down the chain.

Danielle Applestone: You have to admire an industry that not only transformed the world but remains a vital part of business and culture nearly two centuries later. In its daily dance with time, railroads are keeping and often setting the pace of transportation. Generations ago, they did it by finding ways to move faster. Today, by embracing bold new technologies, enhancing safety and reducing downtime, they do it by finding ways to move smarter. This is Technology Powers X, an original podcast from Dell Technologies. For more information on Dell EMC PowerEdge Servers, please visit DellTechnologies.com/Servers. To learn more about this episode, our speakers and to read the transcript, visit DellTechnologies.com/TechnologyPowersX. I’m Danielle Applestone. Thanks for listening.



Evan Morrison

About the Author: Evan Morrison

Evan has a passion for using digital mediums to showcase the impact that technology can have across the globe. While working at Dell Technologies, Evan has produced content and web experiences across various lines of business and continues to explore new ways to tell these amazing stories. Evan is a graduate of Syracuse University and currently lives in Burlington, Vermont.