It is actually bonkers that in this day and age the NYC subway cannot install accurate speed gauges on their trains. You can get an accurate bike speedometer for $20 on amazon, but the NYC subway cannot provide an accurate speedometer for train carrying a couple of hundred people. It is true that trains require more rigorous standards, but it is not a difficult problem. For example you can put a sensor on the electrical motor that will accurately measure motor position. This will allow you to derive speed, as subway motors are directly geared to the wheels at a single unchanging gear ratio.
Same thing for stationary speed sensors in the signaling system. It is maddening that they have to time a train over a distance of 50 feet and use an electro-mechanical timing mechanism. Which has to be regularly maintained and it tends to give wrong readings when not properly maintained, etc. Why not get an instantaneous speed reading with lasers? Those have worked for police officers for 40 years now. Or infrared. Or if you have to time a train over a distance time it with a purely electronic system over a distance of a couple of inches. There are definitely electronics fast enough to do that nowadays. And you can get a purely electronic system that requires zero or almost zero maintenance.
While i agree that it is surprising that MTA continues to use archaic technology, I don't the solution is nearly as simple as you pose it. Hardware engineering is hard, and in-service engineering of complex systems desiring near 100% uptime is challenging.
The sensors must not simply survive a dirty, dusty environment, they must work perfectly with no glitches and for long periods between maintenance. And if they do swap out for a different sensor system and it fails, there's no hardware equivalent of git reset --hard (the favored way for this hardware designer to undo my soft mistakes). You have to take a train out of service, or put people in the tracks during a maintenance window.
What they have there already obviously also requires maintenance, but its performance limitations and failure modes are well understood. It takes time to cycle new things in, and old out.
Nonetheless I was also fairly shocked that their system is quite as archaic as it is. I assume it's a budget limitation driving slow progress.
My 1955 Citroen (designed in 1933) has the original speedometer and speed sender cable. It's accurate (GPS tracked) to within 2-3km/h. This is with nearly 70k miles on the clock. This is the same mechanism Citroen used in their 2CV from 1946 until 1992. Most Model A Fords from 1929 have their original speedometer (which works fine, might need greasing every few decades though). With tens of thousands of miles on them.
We are way, WAY beyond "hardware engineering is hard", this is "this was a solved problem a century ago, using archaic means". I am happy to hand-wave away all sorts of problems but speedometers were 100% a solved problem many many years ago and no allowances or leeway should be given for this specific problem. Zero.
NYCT was, until this year, operating R32 trainsets still limping along from the 1960s that were repeatedly life-extended as newer (1970s) train types experienced structural failures & delivery of their planned replacement trains was repeatedly delayed. It's obviously possible to maintain old trains, but generally expensive & challenging, especially when they're still doing 100,000 miles a year, and things clearly slipped.
Especially because the trains were operating well past their planned lifespan because their replacements were ordered but not delivered, so the major overhaul that would that would normally be done to life-extend a train kept running for decades didn't happen.
That's fair. And also, the environment is probably different, as in more dust in the tunnels, etc.
But to stick with French technology, the Paris metro, for all its issues, does have working speed gauges. Some lines still use rolling stock from the 60s and 70s.
But the problem described in the article seems rather unique to NYC and one has to ask how other subway systems manage just fine without artificial slowdowns.
I'm pretty sure the person you're replying to was saying that the original speedometer mechanism in their 1955 car is accurate to within 2-3 km/h of the speed reported by GPS, and thus suggesting that this pre-GPS mechanism ought suffice for the subway / be better than whatever they currently use.
You can simulate the GPS if you wanted to, drop cables down to pipe through the real signal, or simply do very basic positioning with custom radios underground.
OP is correct. We have significantly better techniques.
They're slowly switching each line to a CBCT based signaling, the lines that do have it are vastly improved. Aside from all the budget and construction corruption issues of the MTA, it's difficult to do because the NYC subway is expected to run 24/7. And so upgrades need to be scheduled around limited times during which lines can be shut down, usually just a few hours overnight. And also keep the old system running while upgrades happen which tend to take years for a full line upgrade.
Anecdotally, CBTC is a greeat improvement in many scenarios. On long and straight stretches between express stops, trains now fly along (probably at least 40mph; never got to see inside the TO’s cab). And upon approaching an occupied station, they slow down to a crawl but keep moving until about 25-50 feet before the station, instead of being stuck several fixed blocks back before the station. Thus there is much greater track capacity = more (potential) trains per hour, and generally more reliable trips.
Unfortunately, it's a pretty tired HN comment to posit something as being more simple than it actually is.
Even during the pandemic there were widespread complaints about the signal work on the L line that has now transformed it from one of the worst lines to one of the best. There's no such thing as "simple work" on a system that millions of people depend on for consistent uptime.
If you count in the financial constrait that service is working under, sure. But meassuring the speed/position of a rail vehicle on a piece of track reliably is not an unsolvable feat of hardware engineering. It is done elsewhere and it is done elsewhere where similar constraints for uptime exist.
Theoretically by that logic we could argue that trains are hard because making the motors for them is non-trivial. It is non-trivial, and depending on your standards it might be even hard. But it is essentially a solved problem. You want a motor? You get one from the big companies, let them design one or use one from an existing similar train. Same thing goes for measuring speed. You want it? Create a team researching which ones to get.
Like in many places NY infrastructure has it's best days long behind itself and it is a wonder it still works. That infrastructure is in dire need of modernization and it has been for a while. The reason this is not done is not because it is hard or impossible to do. It is just expensive.
Don't forget WET environment.
Due to the closeness to the surface and the era most of the tunneling was done, there's not just high humidity but actual water in the tunnels and stations quite often.
Is there a term for using multiple redundancy and statistics to solve issues, for example using 5x 99% reliable sensors rather than a single 99.99% sensor that costs 100x more?
My understanding from TFA is that this isn’t the problem at all: most (all?) of the modern trainsets do have speedometers that the conductor can read. The problem is not measuring the speed; it’s that the city’s signaling scheme no longer matches the handling conditions of the trains that run over it.
In other words: the conductor might be operating the train at a perfectly reasonable speed according to their accurate instruments and the track signals, but the track signals are no longer calibrated for their trainset or the upgrades done to it over the years. This has made conductors excessively cautious, precipitating the recent slowdown crisis on the subway.
Unfortunately, thanks to a combination of poor design choices and poor upkeep, many of these speedometers have proven to be wildly unreliable. So, as train crews navigated the increasingly speed-controlled subway of the aughts, they not only had to build uncertainty around signal design, but equally about the accuracy of their on-train equipment.
In other words the problem is both inaccurate signals and inaccurate speedometers. The combination requiring operators to be much more careful, and making them lose faith in their data. Exacerbating this is that signal trips caused by faulty timers were still blamed on operators. Making them also lose faith in management. Presumably, this wasn't good for the relationships between management and unions, hence probably contributing to strikes.
My guesses:
- Lack of incentives for executives to improve performance
-Lack of people with good ideas to plan improvements
-Lack of skills to execute an improvement plan (due to relatively low salaries)
-Pointless regulation
-Deferred maintenance to ‘save money’
Indeed, I can see this being a bureaucractic challenge more than a technical one.
There are quite clear paths forward and no shortage of prior art for the engineering portion, unless there is something we aren't privvy to it must be lack of incentives and bureaucracy.
My sense is that "we need 99.9% uptime so solutions that require maintenance downtime are not acceptable" even though the lack of solutions is a bigger drag on uptime. But if your mandate is uptime, voluntary downtime is not at all an easy choice. You get blamed for voluntary downtime whilst accidental downtime is not directly going to be blamed on anyone.
I believe it is actually a hard problem but there is a good solution being explored on the UK.
Read the rails like a card magstrip.
Trains first traverse the track scanning the entire length in detail with careful speed and location tracking.
Then trains can subsequently pattern match to the rail for location and speed.
The rails can also be coded with info like patterns although that's not generally needed. Trains get loaded with info about the track they'll be on and then can monitor progress themselves.
I just came up with silly calibration strategy for speedometer. When the train is on a curve of known radious you can get true speed from measured lateral acceleration.
The approach you describe to measure speed from the motor position is essentially what subway cars use. Ignoring wheel slip, which is not a rare phenomenon, this approach works very well. However, when you defer maintenance, things operated in harsh environments _will_ eventually break down.
(Within the context of the issues mentioned by the linked 2018 NY Post article) Speedometers were (bizarrely) judged as non-critical parts (i.e., the car can still be used in service with it broken) because, after all, the signaling system will catch any over-speed, thus the repair, and more importantly the maintenance, of speedometers was not prioritized. Thankfully most of Cuomo's goons and bean counters have been pushed out.
As for the wayside speed enforcement, the author only briefly touched on the solutions to the problem described in the article, but it's known as Communications Based Train Control (CBTC)[1]. It's a moving block system (compared to current fixed block signals) that used train speed, track geometry, and the location of other trains to determine maximum safe operating speed.
I would argue that it's not "maddening" to control subway speed with electro-mechanical timing mechanisms, control lengths, etc. This was cutting edge in the 1920s & 1930s, and indeed some of the oldest signaling in the system is from that era (though thankfully, the amount is decreasing).
It is however maddening to decide in 1995, given other existing speed control solutions at the time (coded track circuits, CBTC, axle counters) to expand the use of these timers. But as the saying goes if you have a hammer, everything is a nail.
Even more maddening is how slow the subway's transition to CBTC has been. NYCT was an early leader, with the Canarsie line being one of the first brown-field re-signaling jobs (not to mention a 24/7 railway), and then the program just seemed to languish under management that didn't see CBTC's value or the need for modernization (could write pages on this). Thankfully the new cadre of people at 2 Broadway has put the CBTC program into high gear, with 4 (5?) lines under various stages resignaling at the movement.
As a bonus tidbit: the wheel slip issue mentioned above is fixed in CBTC operations with the inclusion of a free axle, equipped with no motor or breaks, thus never experiencing a lack of adhesion. Passive RFID balise's placed at known intervals (i.e. loaded into the train) allow the train to then audit (while in operation) how far its estimated position and speed have deviated from where it truly is. Some CBTC systems also have car-brone backups based on accelerometers or rail-facing doppler radars.
Same thing for stationary speed sensors in the signaling system. It is maddening that they have to time a train over a distance of 50 feet and use an electro-mechanical timing mechanism. Which has to be regularly maintained and it tends to give wrong readings when not properly maintained, etc. Why not get an instantaneous speed reading with lasers? Those have worked for police officers for 40 years now. Or infrared. Or if you have to time a train over a distance time it with a purely electronic system over a distance of a couple of inches. There are definitely electronics fast enough to do that nowadays. And you can get a purely electronic system that requires zero or almost zero maintenance.