The private U.S. freight railway system of 140,000 miles of track is widely regarded as the world’s best. Every day, 24,000 locomotives and 365,000 freight cars carry 40 percent of the nation’s freight. Freight railroads own most of the 22,000 miles of track on which Amtrak, the public passenger rail service, runs.
Seven large Class 1 railways dominate the freight rail business. All are investing in sensors, communications, processing, and integration capabilities to help them better understand their operations, increase throughput, and minimize accidents. Integrating those functions through the Internet of Things is key to their future success.
Railway managers are well aware that their knowledge of their vast networks is inadequate. For example, around half of all the nation’s tracks are still in what the industry calls “dark territory”, invisible to railway controllers save through voice messages from engineers. In these areas, mostly in the West, there are no signals, and the switches still need to be thrown by hand by someone who climbs down off a train.
Even though most traffic is in areas with signals and sensors, most of those sensors just say when a train has passed a certain point, but not its speed, or whether it is accelerating. Railways have been focused in lighting that dark territory, and achieving situational awareness.
According to the U.S. Federal Railroad Association, more than 60 percent of all rail-related accidents are due to human error. And this is the area that is targeted by what is called Positive Train Control (PTC).
PTC was in the news most recently in the May 12, 2015 Amtrak derailment in Pennsylvania—the technology was installed on that stretch of tracks, but not yet operating. If it had been working, it is likely that the crash would have been prevented.
PTC is a sophisticated and predictive system to monitor train location and speed, inform signaling equipment of the train’s presence, and automatically control train speed in case of operator error.
Some railways complain that the system is expensive, and only prevents a limited subset of accidents. For example, there are deaths that happen from collisions at crossings, and people trespassing on the right of way, neither of which will be affected by PTC. Still, there are many promising benefits.
PTC is a good example of IoT in action, collecting information, processing it, and using it for optimizing performance and making crucial decisions quickly and appropriately.
Given a cost of around $2.5 million a mile, freight railroads aren’t laying a lot of new track. Thus, increases in freight volume have to go through the existing network.
The average speed of a freight train on a long-distance run ranges between 20 and 25 miles an hour. Locomotives are certainly capable of pulling trains at much higher speeds, but rail yard congestion, breakdowns, and a lack of precise knowledge on train locations all impose limits.
Norfolk Southern estimates that a one mph increase in average speed could be worth $200 million in additional revenue. The integration of IoT velocity, traffic, and location data will allow for increases in average speeds, and thus volume and revenue.
From reactive to predictive maintenance
A key failure point in a railcar or locomotive is the axle bearings. Union Pacific, for example, has put infrared sensors every twenty miles along its tracks to detect overheating bearings. It also has trackside microphones that listen for the sound of wear. And in the highest-risk, highest-consequence situations, as with heavy coal trains, it uses ultrasound to look for flaws in the wheels in a running train.
This information is sent down fiber-optic cables along the tracks to UP’s data centers in Omaha. There, 20 million sensor readings are analyzed per day. Potential problems are flagged for operators, who then decide whether the problem requires immediate intervention, or can wait until the next maintenance stop.
Railways currently respond to maintenance issues reactively. Only when an issue is detected is it responded to. Like other industries, railways are desperate to move to predictive maintenance, where knowledge of failure modes makes it possible to create algorithms to anticipate and deal with maintenance problems before they manifest themselves in performance failures. The result is less time in the shop and more time on the rails.
The railway sensor problem
Sensors are the foundation of the IoT. They are the source of information about the physical world.
Sensors for rail cars need to be rugged, long-lasting, and capable of working for long periods without recharging. For example, those heavy switches in dark territory are in remote, harsh territory, far from power lines and cellular links.
And rail operators would love sensors everywhere: measuring the pressure of a tanker car, indicating whether hatches are open or closed, noting if a handbrake is on, measuring the temperature of a refrigerated car. At the moment, the technology simply is not there for a comprehensive solution.
Sensors, in railways and elsewhere, are the key technological growth area over the coming decade.
The visible, and the easy to miss
The media is captivated by “transformational” technology, by inflection points, by dramatic and visible changes. A one or two percent increase in railway velocity over several years is hardly worth noticing.
But never underestimate the power of IoT-enabled incremental improvements. For a period of time it may seem that nothing is happening—and then everything will be different.