Feeling The Heat: Railway Defect Detection

On the technology spectrum, railroads would certainly seem to skew toward the brutally simplistic side of things. A couple of strips of steel, some wooden ties and gravel ballast to keep everything in place, some rolling stock with flanged wheels on fixed axles, and you’ve got the basics that have been moving freight and passengers since at least the 18th century.

But that basic simplicity belies the true complexity of a railway, where even just keep keeping the trains on the track can be a daunting task. The forces that a fully loaded train can exert on not only the tracks but on itself are hard to get your head around, and the potential for disaster is often only a failed component away. This became painfully evident with the recent Norfolk Southern derailment in East Palestine, Ohio, which resulted in a hazardous materials incident the likes of which no community is ready to deal with.

Given the forces involved, keeping trains on the straight and narrow is no mean feat, and railway designers have come up with a web of sensors and systems to help them with the task of keeping an eye on what’s going on with the rolling stock of a train. Let’s take a look at some of the interesting engineering behind these wayside defect detectors.

Friction Is The Enemy

A railway car truck, or “bogie” to the British. This is an old-style truck of the Bettendorf pattern; the journal boxes contained oil to lubricate the plain bearings at the end of the axles, and often caught fire in dramatic fashion, leading to the term “hot box.” Source: US Army Field Manual FM 55-20 (public domain).

At the risk of stating the obvious, trains have two essential characteristics that make monitoring systems necessary: they’re heavy, and they’re long. The weight of a train is a problem because even though the basic architecture of a railway reduces rolling friction between a wheel and the ground, it does nothing to reduce friction between the railcar’s axles and the trucks that carry them. That’s the job of the wheel bearings, which like any other mechanical component are subject to wear, damage, and eventual failure, with the potential for catastrophic consequences.

As for the length of a train, that becomes a problem when it puts most of the rolling stock out of the direct visual range of the people running the train. Back in the day, limitations on locomotive power tended to keep trains relatively short, making it possible for conductors and engineers to keep an eye on every car. This was made easier by the invention of the caboose; in its classic configuration with a windowed cupola jutting up over the roof of the car and from its position at the very end of the train, conductors were able to observe the entire length of a train, especially on curves. Given that the wheel bearings of the day were often plain bushings in journal boxes stuffed with oil-soaked fibers, it was generally easy to spot a “hot box” bearing failure by the smoke and flames they emitted, as unsubtle an indicator of trouble as there ever was.

Engineering advances, like replacing plain bearings with roller bearings, made it possible to build ever-larger railcars. Freight cars operating on North American railways these days can have a gross weight of 315,000 pounds (143 tonnes), a mind-boggling amount of weight that’s carried by as few as eight roller bearings. Improvements in locomotive design have also allowed trains built from these supersized cars to get ever longer; the average freight train in 2017 was between 1.2 and 1.7 miles (1.9 to 2.7 kilometers) long, with some railroads regularly operating trains 3 miles (4.8 km) in length. On a train like that, anything more than a dozen or so cars back from the head-end locomotives is out of direct visual range of the engineer and conductor, and is effectively operating completely unobserved.

Eyes On The Rails

Wayside monitoring is the answer to the problems presented by scaling trains up to such massive dimensions. Collectively known in the railroad business as “defect detection,” the sensors and systems installed periodically along railroad tracks automatically scan for any problems with the rolling stock of a train that might result in an accident.

Defect detector with hot box, hot wheel, and dragging equipment detectors, along with the equipment bungalow. Source: Sturmovik, CC BY-SA 3.0

For good reason, the bulk of defect detection is focused on the condition of the wheels and bearings on each car in the train. And since friction is the enemy, most detectors key on the heat of these critical components to assess their condition. A typical wayside sensor installation will include both hot-box detectors (HBD) and hot-wheel detectors (HWD) on both rails. Both sensors are typically based on microbolometer arrays like those in thermal cameras. For HBDs, the sensors are typically mounted on the outside of the rail and pointing up, to get a good look at the bearing boxes on the end of each axle. HWDs are also typically mounted outside each rail, but they are aimed to look directly at the side of the wheel as it passes by. The thermal characteristics of wheels and bearings are quite different — wheels can get much hotter than bearings before counting as a defect — so HBDs and HWDs are calibrated differently.

Another detector present in most defect detection stations is the dragging equipment detector (DED). These are simply a series of paddles that are set up perpendicular to the rails. The paddles are mechanically connected to switches and are activated by anything that the train might be hanging down from the bottom of the train. The main target here is a disconnected air brake hose, but there are plenty of other hazards, everything from a broken truck to an unlucky animal. DEDs have to be extremely robust, as impacts with dragging equipment can exert a force of 600 g, and most DEDs are set up to work with trains moving in either direction.

Defect detection installations have become quite pervasive alongside North American railroads; there are something like 6,000 HBDs currently installed, or about one every 25 miles (40 km) of track. HBDs and HWDs can be a little hard to spot for railroad watchers, partly because they sit very low next to the track, but also because they need to be at least 300 feet (100 m) from a grade crossing, which is the only place most people get to see railroad tracks up close. Far easier to spot are the wayside bungalows that hold the equipment to which the sensors are connected. Bungalows look like little utility sheds next to the track, usually painted white or silver to reflect light and keep the temperature inside relatively constant. They contain racks for the electronic equipment that processes the signals from the detectors, along with support equipment like computers, power supplies, and backup batteries to keep the system powered even if grid power fails.

Defects On The Air

Once a defect is detected, what becomes of the information? A clue to that can be found on the wayside bungalow, which often has a conspicuous antenna mounted to it. Most seem to be a folded dipole antenna for either the VHF or UHF band, mounted vertically alongside the track and oriented to radiate parallel to it. Inside the bungalow, automated equipment uses a voice synthesizer to compose a report of the train’s condition, including any defects found, and transmits it to the train crew. Reports generally include the identification of the defect detector, the nature of the defect, and the axle or axles where the defect was detected. The train crew can then stop the train and walk back to the problem car and assess the nature of the problem.

Hot boxes, hot wheels, and dragging equipment aren’t the only kinds of wayside detectors in use, of course. It’s very common to monitor for mechanical defects on wheels, such as flat spots caused by an axle locking up and dragging the wheel on the rails. Flat spots cause excessive noise and equipment wear because of the impact they cause, and are detected by wheel-impact load detectors, or WILDs. These consist of a series of strain gauges attached to long stretches of track, which record the high-g vertical impacts generated by out-of-round or otherwise damaged wheels and alert the train crew. Other detectors are focused on the status of the rolling stock trucks, to make sure the axles are in proper alignment with the track and not engaged in “track hunting,” a lateral oscillation of the wheelset on the tracks as it seeks an equilibrium point. Hunting can result in damage to the wheel flanges, the track, and can even cause derailment if the oscillation causes the car to sway too much.

As you can imagine, all this equipment requires a lot of resources to install and maintain. Defect detection systems are widely distributed, with installations often dozens to hundreds of miles apart, meaning that technicians trained to work on them are spread thinly and often have to travel to remote locations to work on the systems. It’s worth it, though — since 1980, trains accidents related to axle and wheel failures have decreased 81% due to the widespread use of hotbox detectors.

29 thoughts on “Feeling The Heat: Railway Defect Detection

  1. If you read 1950’s popular science type magazines, you’ll sometimes run across Timken Bearing ads that say “Watch for a Hot Box!” and have a phone number for people to call if they see a car spewing fragments of hot metal and lighting fires along the right-of-way, which was a pretty effective ad for a company trying to replace the last plain bearing trucks with their roller bearing offerings. Where I live, we have a lot of issues with wildfires, and several have been caused by a train with a failed bearing that was doing exactly this, despite the roller bearings.

    Minor nitpick: 600 g at least momentarily implies a force of 600 grams. I think 600 G’s would get the idea across better.

    1. How do they deal with flat spots in cars? Grind the wheel down to the flat? Or replace it?
      Once I took a sleeper from Paris to Lisbon, and there was a flat spot right below us. It made for an unpleasant ride all night.

      1. If it’s minor, my basic research suggests they just turn it down on a wheel lathe – the wheels aren’t coupled to any other, so the precise diameter doesn’t likely matter much.

        I would also like to hear from someone more knowledgeable on if more serious flat spots are repairable and/or when replacement becomes preferable. I have a friend in the rail industry and might ask them if they happen to know (though they are not a mechanic/technician).

        Some things are plainly irreparable, though http://www.tsb.gc.ca/eng/rapports-reports/rail/1996/r96t0095/r96t0095.html (found a reference to this accident while doing a little research)

        1. I’m definitely wrong about them not being coupled – I’m not really a train guy, sorry. Not sure why the simple concept of an axle didn’t spring to mind xD. Still, there are probably standards for pairing wheels of similar diameter, and limits on how different they can be.

        2. Typically we keep L/R wheels within 1mm of circumference. Look at the videos from the poster of the video attached. All of this machining can be automated and eat through wheel flat spots.

  2. Ok so the crew gets a hot box warning. It’s a 3mile long train so you might have to find a place to park (or if your a BNSF crew just stop anywhere including major highway grade crossings at rush hour) get out and possibly hike a 6 mile round trip to go see. Worse if the railroads get their way with single person crews who aren’t noted to be in the best physical condition. What could go wrong? The current problem is the bearing failed just after the hot box warning but before the train got stopped and a couple million tons just don’t stop all that fast

    1. My eldest son was a train driver. Once he was driving a heavy good train and had to cross a busy major urban station. He was eventually cleared to cross it but at the end of the station, was alerted by a side worker that the second of the two connected pulling engines was emitting fumes and sparks.

      Of course, the end of the train was still before the station’s other end, so he blocked the whole traffic, with a lot of high-priority passenger trains coming… The problem was fortunately well known : the second engine was prone to keep his brake closed when pressure to release was not applied enough fast (train’s brake system security asks that braking work by lowering pressure, so in case of failure, the train stops instead of runaway).

      So he had to reapply full brake, wait until everything is firmly blocked and then reapply full pressure to open the brakes (and us a hammer to ensure the involved bogie brake are not still blocked). Everything under the anxious look of the station stationmaster, asking every few second if he can repair quickly…

    2. The warning tells the axle number and car number where the problem was observed. So, yes, it still may be a hike to get there, but at least they’ll know where they’re headed.

  3. Well, trains driving require also some skills, especially when this long ! For example, since brakes are air operated, there is some (long) delay between the first and the last car start — and finish — to brake, or accelerate… If you stop too early or worse reapply power before the whole train is stabilized, you can break the link between two cars…

    1. This is why we have electronically controlled valves down the length of the train to activate the brakes…

      Also doesn’t the brake activation propagate at the speed of sound, without such electronic aids?

  4. How about monitoring on each bogie. Data bus through the brake line path (low power RF) powered by some form of motion energy recovery to make it easy to maintain.

    For many years motorists nearing a small town just north of here were greeted with a sign on the little metal hut right by the road “Battleground Defect Detector” and wondered if only good people were in town.

    Now we know!

    1. Rail cars are held to their trucks/bogies by weight alone and are supposed to be matched to them by serial number in case of an accident. They don’t articulate too much so running a bus from the sensors to the rail car should be straight forward.

      Using the brake line path is smart. If you could make the data and power connections through the Gladhand pneumatic couplers that would make the system transparent to the yard crews and conductors who have to make and break trains all day.

      The biggest hurdles I see are isolating all the noise, both electronic and physical, from the sensors so you’re only monitoring what you want. As for power, whatever motion energy recovery systems are there will need to be as tough and dirt simple as everything else on the railcar.

      On the whole though I firmly agree we need telemetry on every piece of rolling stock out there with load monitoring on anything even remotely flammable, HAZMAT, or both. Hell even grain is an explosion hazard. Given the amount of rail theft that occurs around switching yards I want sensors and better locks on the doors of box cars too.

  5. “exert a force of 600 g”
    You mean an acceleration of 600 g. A force of 600 g would be 600 grams, not much, wheras you are talking about an acceleration of 600 x the earth’s surface gravity. That acceleration only becomes a force when it knocks against an object with mass.

  6. “For HBDs, the sensors are typically mounted on the outside of the rail and pointing up, to get a good look at the bearing boxes on the end of each axle. ”
    Are trains in the UK designed that much differently from the US? Here the wheels are on the outside with bearings mounted on the shaft just inboard of the wheels. This makes one think a heat sensor (or 2) mounted inside of the tracks would have a better view of hot bearing housings far sooner than outside mounted ones, which would seem to be somewhat shielded by the wheels. I have a shaft with wheels and bearings sitting in my steel pile and it seems to me that by the time the housing gets hot enough to heat the shaft (2-9/16″ /~65mm dia iirc) which then has to heat the wheels….that bearing has probably been long gone?

    1. Look at the line drawing of the truck way up on top. Typical North American design with bearings on the outside. Much easier to oil when they were sleeve, kinda stuck with the design

        1. You could theoretically have the bearings inside the wheels, but I don’t think this was ever done one normal-gauge mainline equipment except for some locomotives (especially steam and early diesels) and for very, very early railcars (think pre-1850). Even in the UK which had a relatively narrow loading gauge, outboard bearings were the norm.

  7. I agree with the reason of the nitpick, but that’s not a solution.
    600 g is a mass, not a force, but when something says “force of 100 g” than what’s probably meant is the gravitational force due to a mass of 100 g on earth’s surface – an amount of force that’s probably deep in the noise when it comes to problems for a train.

    OTOH, 600 times earth surface acceleration is not a force but an acceleration, with the force being entirely up to the mass that’s being accelerated – which is what – the detector’s mass?

    I’ve only ever seen the earth surface acceleration with a lowercase g. Uppercase G would be gravitational constant, which is very definitely not a force (m^3/s^2/kg) – the one that’s in F = Gm_1m_2/r^2

  8. A few years ago we took the Amtrack from Spokane, WA the Chicago. I programmed all of the 162 MHz railroad frequencies into my Ham Radio FT-60 and put it in scan mode. We could hear the reports(voice announcements) when we passed over the detectors. We heard the # of axles and “no problems”.

  9. I wonder whether we could build a citizen network of train defect monitors using Raspberry Pis with IR cameras, an SDR for listening to those voice messages, and a LTE SIM, powered with a solar cell. These gadgets would “phone home” using SMS or MMS messages upon seeing or hearing trouble.

    Machine vision is getting pretty good, so this stuff could be placed adjacent to the right-of-way, not on it.

    The hobbyist-run ADS-B stuff that monitors airplane movements is my model for this. It’s a little simpler, because all it needs is a Pi, power, an SDR, and an internet connection. But it would be the same idea.

    Imagine. “Hey, ChatGPT, which trains are in the worst shape and how long til they pass through my town?”

    Railroad companies’ front offices’ lack of attention to safety-related maintenance cries out for independent defect information gathering and dissemination.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.