Tucked away in an inconspicuous corner of Union Pacific’s Omaha headquarters is a set of plain, white metal doors with a small gray sign that reads “LAB.” Behind the doors lies a peek into the railroad’s future.
Inside the lab, it looks like a robot lost its lunch on every surface. Electronic components are pouring out of every nook and cranny.
Dan Rubin’s office sits next to that 6th floor research and development lab. A general director of Information Technologies, Rubin’s title conceals his real job.
“I’m an evangelist,” Rubin said. “People come to us with a problem, my team provides a solution, and I evangelize that solution.”
The people who work in the lab solve problems by designing the internal and external components of the complex electronics that run the railroad. Projects like Machine Vision (a train-sized, erector-set-like portal containing detection sensors, cameras, lasers and strobes) Arrowedge and Positive Train Control require many gadgets to work. Everything from small, hand-held devices called “puck readers” to custom computer motherboards are created from scratch in the lab.
A Different Kind of Puck
“Pucks are transponders,” said Royce Connerley, senior systems engineer, from his equally messy lab workspace. “They transmit an identity that’s used to establish a point in space.”
Where remote-controlled locomotives are operated, pucks embedded in the track communicate with locomotives to mark the location of crossings, switches, the mainline – anything a locomotive shouldn’t run into. Up until now, maintenance crews have used large, clunky, outdated devices to verify that each puck was working.
“The people who do this work wanted something better,” Connerley said.
Initially, Connerley thought the device should be compact and portable. “I thought it should be something you can take anywhere,” he said. But then feedback came back on the first prototype – the new design was causing backaches. “The prototype required crews to bend down to use it. They didn’t like that.”
After several tests, additional feedback and a lot of creative thinking, the new puck reader now looks like a little metal detector with a rubbery, 3D-printed handle containing the display.
“To manufacture the handle, our options were to have an injection mold created, or to 3D print it,” Connerley said. “Considering the quantity needed, 3D printing was less expensive.”
Eventually these new puck readers will become standard issue for maintenance crews throughout the Union Pacific system.
Lasers and Cameras: A Dynamic Duo
Before a train departs, each rail car requires a 13-point inspection. Typically at least 100 cars long, it can take several hours to get a train fully inspected. Union Pacific operates hundreds of trains per day, so the time adds up.
About five years ago while working to prevent mechanical derailments, Todd Snyder, director of advance freight car engineering, realized lasers could be used to inspect trains as they pass. The idea resulted in Machine Vision. These “In Advance” inspections are conducted rain or shine, day or night, without leaving the comfort of a desk chair.
As a train passes through a Machine Vision imaging area, lasers and cameras provide a three-dimensional model of each piece of train equipment. The model and images can be viewed remotely from any Union Pacific computer. So far, the system is able to identify and measure 22 components of a train.
“With In Advance inspections, I know what repairs are needed before the train arrives in the yard,” said Tom Jacobi, vice president of Operating Systems and Practices, who oversees the implementation of technology related to operating trains.
These kinds of inspections accelerate cars through the yard. “By using technology to help facilitate what goes on in a yard, we greatly reduce the risk of missed connections,” Jacobi said. “We improve the customers’ experience.”
So far, Machine Vision has been implemented near three rail yards: in Nebraska, Iowa and Arkansas.
The technology development took ingenuity and cold mathematical logic from Rubin’s team —especially from Yujie Ying, a quiet young woman who wrote some of the algorithms that make Machine Vision work.
Sitting in her 6th floor cubicle, Ying’s computer desktop wallpaper displays her motivation – a large photo of her infant daughter. The baby’s eyes peek over a computer program displaying Machine Vision, the culmination of her mother’s hard work.
The algorithms Ying and her team members wrote comb through the data gathered by Machine Vision as a train passes through the metal structure at 70 miles per hour, displaying images and analyzing characteristics of the train in near real-time.
Various types of data come from wheel detectors, lasers, infrared cameras and line-scan cameras that photograph one continuous line of pixels at a time – snapping 50 thousand photos every second – which are then strung together to form an uninterrupted image of the train. “We can see crystal-clear images of the train as it passes,” Ying said.
Then there’s LiDAR, which stands for Light Detection and Ranging, a surveying technology that’s been used for geo mapping. Machine Vision uses LiDAR to generate three-dimensional images of the entire train.
“Our algorithms look for common defects using photography and LiDAR. For example, we can check to make sure the amount of spring compression is appropriate,” Ying said. Springs are part of a rail car’s suspension system. “If the springs are compressed too much on one side, it throws off the car’s balance, which could cause a derailment.”
The long-term goal is to automate certain aspects of inspections using Machine Vision, but optimizing the system requires another evolving technology: Centralized Automatic Equipment Identification (AEI).
Data for Days
“If we can’t accurately identify a rail car, we don’t really know where it is in the train,” said Gary Baker, general director – Information Technologies. “AEI is critical.”
AEI is a system developed to keep track of rail equipment. Each rail car has an AEI tag – a small radio transponder located on the side of the car containing equipment information. AEI readers sit alongside the track identifying each car as it passes via radio signal.
“Right now, if a train stops for 15 or 20 minutes, the AEI reader sometimes ends up splitting the train because the AEI reader thinks it’s a new train,” Baker said. “We also have this problem in areas that are double-or triple-tracked – the AEI reader ends up getting signals from two passing trains.”
Union Pacific’s engineers are working to solve these problems with a new reader called the T2000.
“Once we get all the T2000s rolled out, everything will be consistent,” Baker said. “The new AEI readers are essentially dumb devices. All they do is collect information and send it to a centralized location in Omaha.”
Centralizing all the data lowers AEI reader replacement costs, adds consistency, allows Union Pacific to leverage that data to improve service and improves Machine Vision’s accuracy.
“We need to be able to get specific as the train goes past Machine Vision,” Jacobi said. “I want the system to tell me there’s something wrong with the 72nd car on a 100-car train. I want it to be that detailed.” Without centralized AEI, the technology can identify the general location of a problem, but not the specific car. “Right now, we have to check five cars in front and five cars behind,” Jacobi said.
Fostering 'Aha' Moments
Like all technology, AEI will eventually become outdated.
“Honestly, in the future, AEI no longer exists,” Baker said. “There’s another research and development project in the works that would make AEI readers obsolete – if it happens.”
That research and development project will remain a mystery, because the concept is still being developed. It begs the question: How does an idea become reality at Union Pacific?
The Technology Steering Group drives the process to fund promising projects. It’s made up of leadership representatives from departments where innovative ideas are most likely to bubble up: operating, engineering, mechanical, safety and information technologies. This process means research and development at Union Pacific is integrated into day-to-day jobs. Employees are able to keep an eye on trends in an environment that fosters those precious “Aha!” moments.
Jacobi has been part of the Technology Steering Group since its inception.
“We created the group because we needed a path to provide funding to explore, validate – or invalidate – projects that could benefit Union Pacific customers, communities, shareholders and employees,” Jacobi said.
The group currently is looking for ways Union Pacific can use drones, 3D printing and gaming simulators to train new engineers.
“The beauty of the future is we don’t know what it is,” Rubin said. “But if we continue to stare at the horizon, we’ll be here to make sure people understand the vision, and believe in it. I wouldn’t say we develop bleeding-edge technology, but we are changing the way Union Pacific does business."