Worn Train Rails Get Judged By Laser

[Calango] is a railway technician, and for a school final project created the Rail Wear Surveillance Trolley (RWST) which is a delightfully designed device made mainly from PVC conduit with one job: travel down a segment of train track while shining a green laser onto the rail, and capture camera images. The trolley holds both the laser and the camera at just the right angles for the camera to capture a profile of the rail’s curved surface. The images are sent via Bluetooth to a smartphone for later analysis. Rail wear can be judged by checking how well the profile of the rail conforms to the ideal profile of an unworn segment. The trolley is manually pushed by an operator, but [Calango] says that ideally, it would be self-propelled and able to inspect a length of the track then return on its own.

The project was made on a tight budget, which led to some clever solutions like using a rotary encoder attached to a wheel as a makeshift distance sensor. If things get desperate enough, it’s even possible to roll your own rotary encoder with a 3D printer and two microswitches.

32 thoughts on “Worn Train Rails Get Judged By Laser

      1. I’m guessing that you could build such a system. There are industries where you get insane speeds and have to optically measure in real time a surface going at high speed, with great accuracy (~0.05mm) such as rubber profile extrusion for tire manufacturing. A train is faster than that, but it should be possible.

      2. I dunoooo lasers are pretty fast. There should be hardware out there that can refresh at a high enough rate to keep up with that. I mean there are 8k monitors which refresh at 140hz, surely this is doable.

        1. After thinking about it, since you don’t need evaluation in real time, this problem boils down to a high speed camera looking at a rail track that’s illuminated by a laser, recording a black and white image during the part of the ride that is of interest. The evaluation of the video can be done at any later time.

          1. DrainBramage is just off by a factor of 1000. 1ns/ft would be roughly Mach 888,630. Or perhaps more palatable, about 1.01C. At 300mph, you would have about 2.27 milliseconds per foot. If you used the same methodology and captured a very narrow slice at a time you would need a pretty high capture rate (~134,000 frames per second for a 1mm wide capture). A cheaper/simpler approach would be to capture a wider space at a time. If you captured 1cm at at time, that capture rate (~13,400 fps) would be well within the capabilities of affordable high speed cameras like the Chronos at the low resolution you would need.

          2. “By my (admittedly unreliable) math, that would be a train going about Mach 900.”
            Actually Mach 900000, surprisingly close to the speed of light.

          3. In reply to x88x, you don’t need to use a camera. One reason a camera is so slow is it has to take a snapshot of a large grid of sensors (Megapixels) in the space of a frame. If you’re just looking at a scanning laser, I would think you would be able to keep up with the track more easily

      3. Wear monitoring would probably be quite do-able, as, if many or all locomotive have a system and collect data, the data can be combined to build a higher resolution picture over time. Fault identification might also be reasonable, since typically a critical fault (such as out of gauge, buckling, or a rolling rail) may be large enough to detect with a single pass at speed before it becomes an incident. Probably would not be able to detect things like a cracked rail or separated joint, but there are other systems for that (such as monitoring the sound as a wheel rolls on the rail).

        1. why does it have to be video. why not a snapshot at 1 second Intervals. can be a long(relative)exposure just strobe the Laser. the rails aren’t going to change profile every 5mm. I imagine the rails will change profile gradually over meters at a time

          1. It doesn’t need to be video to build a history, but video is cheap and off the shelf. Given that the wear issue does require a profile every few mm along the rails, to catch things like wheel burn and shelling, video would make sense. Each pass with a loco at 100Km/h (about 28m/sec), grabbing images at 60Hz, would get a profile every 463mm. This implies that roughly 200 passes would be required to get a good profile every 10mm (statistics requires more than 46 passes). In addition, this is a short enough distance (under many conditions) to allow registration of successive images and find other key locating features/details (such as unique features of a crosstie or rail clip/spike head) so as to allow the data to be aligned with different run under different conditions.

            At one second intervals, a shot would be every 27700mm, requiring of the order of 15000 passes to get 10mm resolution (if I did the arithmetic correctly), and registration would be impractical.

            Video and processing, at this time, are cheap and off the shelf, so the much of processing could be done on the fly, minimizing storage requirements.

      4. I work for a railroad maintenance company and we use a similar system. We can do real time analysis, typically grabbing 4 pictures (both sides of both rails) every 5 feet at 20 mph. The system can go faster but that is all we need for the maintenance we do.

    1. The rails are not the only things that need inspecting. The railroads have to monitor the railroad continuously for damaged ties, dislodged spikes, breaks in the fences, properly operating signals and grade crossings, etc. Typically this is done daily with a dedicated vehicle. It takes a long time for steel rails to degrade, you don’t have to test them continuously.

  1. Nice project – shows what’s now within the cost and ability of hobbyists rather than the realm of industry. It’s certainly cheaper than the UK Network Rail’s measurement train which does broadly the same thing. That does a lot more though as it travels about with multiple lasers, cameras, accelerometers and so on. Including causing people to report alien invasions when it first started using an upward pointing blue laser to measure the distance of overhead cables and obstructions.

    1. No intention to diminish the achievement of Calango. It’a a good tool as a result of a college project. It could even work good enough for regular field use by the creator.
      But it’s no commercial solution, which has a guaranteed reliability and accuracy as well as usually support, service and easy operation. And because salaries are not everywhere cheap, it would be good to measure, check and/or record other rail and track properties at the same time. So dependng on all this you can add up to 2 Zeros to your guess, because a complete mesurement rail car can cost a few million USD/EUR/GBP.

        1. In my experience reliability and affordability go hand in hand. Do you want 99%, 99.9% or 99.99% reliability. You’ll generally be adding about one to two zeros for each extra 9 in there. If your expensive tool doesn’t have the reliability to get the job done when it needs to, it’s an expensive paperweight.

  2. I’ve got to wonder how this differs from what we called a Geometry Car that rolled slowly along the tracks in the wee hours when service was down at a Bay Area commuter rail system.

    1. BART is most interesting in that all of its rolling stock and tracks are non-standard. The Bay Area will be well and truly f’d when the trains start to wear out, they will either have to replace the whole thing all at once with standard equipment, or else they will have to buy custom made- rolling stock (at greatly inflated prices, no doubt) to fit their no-standard rails.

  3. Great project, I love the low cost approach.
    Practically though, the speed at which you can check the track isn’t governed by the measurement apparatus, it’s the permissible line speed on the track.
    Network Rail have a measurement train that has both engineers and automated AI trained checking that will run at 125mph

    However, those of us in the UK will likely agree that walking pace is a speed at which we can only dream that our trains will attain…. ;)


  4. The camera could also capture the rotary encoder image (e.g. a patterned disk or strip). That way the track image and rotary encoder signal would always be in sync.

  5. I was working for a company (JRB Engineering/MRX Technologies) that made this system pretty much exactly.
    They used high speed cameras and had them attached to their road-rail vehicles, which had encoders on the wheels, and used the location beacons to determine where they were on the track.

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.