What makes a train safe?

52 Replies to “What makes a train safe?”

1. Well said. You nailed it with the last paragraph. This was a process failure. Many layers of management failed to communicate the dangers properly. The report strongly emphasized this. Yet it is much easier to go out and buy something, just to show you are doing something about safety. It is ironic when that change might not even help, but could make things worse.

   1. What’s status of legal action now? Good comment, but somehow “process failure” describes the crime so weakly it tempts justice to the point where the more you know about passenger transportation, the greater our personal guilt if we don’t at least try to prevent the next one.

      Mark Dublin

2. I agree also, and hopefully expanding: As you aim for higher speeds, the safety strategy *has* to be focused on avoidance because at some point hardening just slows the train without appreciably reducing the harm of failure. We don’t harden airplanes to survive a crash, we put all our eggs in the avoidance category, which is most likely the only strategy appropriate to high speed rail should we ever achieve it.

   The Talgos were brought here due to their tilt technology, intended to be stable through curves at the higher speeds we’ve always aspired to achieving. Unfortunately we’ve never improved the infrastructure to allow those speeds (as evidenced by the slow speed required on this new track). If we still want to reach higher speeds in the future, the focus has to be increasingly on avoidance by designing the infrastructure and managing the operation safely.

3. The cars are looking dated and the awkward ADA access is also an issue not easy to redesign. I could see how the NTSB would throw replacement into the report recommendations.

   Still, there have been many derailing because of trains hitting curves at speeds too high. It’s not just this one instance. PTC is a great remedy — but without proper curves it’s still just a work-around.

   The primary long-term solution: The curve should be changed to allow for faster trains. The design process also needs to have a different way of approving difficult curves.

   Finally, a question: Is there a nearby, readily-available use for the Talgo cars? A supplemental service to Bellingham? A slower cross-Cascade route? Something on the ERC? A ski-rail express? A DMU-style feeder service on an existing underutilized track?

   1. It is practically impossible to fully design out track features that require localized speed restrictions. There will always be curves, switchpoints, lift bridges, grades etc that require some sort of speed restriction.

      That said, I would support rebuilding that curve to higher speed standards. But not as a safety feature, rather as a speed and efficiency improvement. The goal of the line is to get the average speed between Portland and Seattle up, and rebuilding that curve would support that goal.

      However, if the curve is rebuilt it should be rebuilt to allow speeds higher than 79 mph. There is no reason that the Lakewood line can’t operate at speeds of 99 or even 110 mph with proper crossing improvements.

      1. Yeah, there is nothing fundamentally dangerous about this curve. Nor was there anything wrong with the trains. You don’t blame the scalpel when the surgeon removes the wrong organ. It is a system failure. The driver is at fault, but so too is the person in charge of the driver, all the way up the chain. Just as with medicine, you can’t expect people to be perfect. You don’t fix things by buying some whiz-bang technology. You fix things by improving the process (the checks and balances that go into every procedure).

      2. Straightening out this curve to allow faster speeds just moves the speed restriction a quarter mile or so down the line. Trains will have to slow similarly for the junction with the current route.

   2. “Is there a nearby, readily-available use for the Talgo cars?”

      Did the report recommend removing the Talgos from all freight-sharing track, or only from this route with this curve? I thought the argument was that the Talgos didn’t comply with federal regulations and shouldn’t be on any freight-sharing track. That would preclude using them on the Bellingham route or a Spokane route or Sumner – Maple Valley.

      1. The NTSB recommended that these Talgos not be used anywhere.

   3. If they put a hairpin curve on I-5, let’s say in between Marysville and Tulalip, and freeway traffic had to slow from 65 mph, or whatever the speed limit is there, down to 10 mph, then back up to 65 again, and there were more accidents at that location than before they put in the hairpin curve, you would place no blame on the curve for the increased accidents?

      1. The 30 mph curve always existed. What was new was the 79 mph tracks leading into the curve. Maneuvering a train through a 30 mph curve from a 79 mph right-of-way isn’t an extraordinary feat as long as the crews are well-trained and there are proper signage and functioning signals that keep the crew informed about their location.

      2. Even if I-5 always had a hairpin curve, of course, I think the blame would primarily be the curve in that situation. — especially if the same problem occurred with half of interstate auto crashes around the country. Thanks for understanding my point, Sam!

      3. The Renton S-curves, for instance? They were always there and had many accidents, and finally after decades the state straightened them out.

      4. Guy, well strictly speaking, not “always”. The kink has been there since about 1950 to 1960, when I-5 was built through Fort Lewis. Before that the line was a roughly straight diagonal down the hill toward the junction just north of the Nisqually River.

   4. They could certainly speed up a Seattle-Tri Cities line over Stampede. But WSDOT is not going to touch them after they’ve withdrawn them from service on the Cascades.

   5. My guess is that if WashDOT sells the Talgos they will probably be bought by North Carolina DOT. The cars they are using on the Piedmont service are from the 1950s and meet fewer standards than the Talgos do, and the Talgos have ADA access from cars 2 and 3 into the bistro. I don’t think anything in the Piedmont and Carolinian service currently has that type of access.

      For a while it was the fastest growing state provided service in the Amtrak system and at some point they will need additional cars. The rebuild work on the 1950s cars to get them into Amtrak service runs about $2 million each if I remember right, and it is hard to fathom WashDOT selling the Talgos at a higher price.

      They could also wind up in one of the eastern Canadian routes., or countless other places. A buyer will be found, I’m certain.

4. Talgo’s submission to the NTSB claims that the Talgo VI cars likely performed better in the accident than US designed equipment would have performed because the Talgo VI cars are lighter and are designed with a lower center-of-gravity. The advantages are that lighter weight equipment can stop sooner and a lower CG will keep the cars on the tracks and prevent toppling over. The Talgo cars did remain upright except for the car that fell off the bridge. Talgo claims that the VIs would have been able to remain on the tracks at 78 mph if they hadn’t been pulled off the tracks by the locomotive.

   Talgo also compares the DuPont accident with several other accidents and suggests that the US designed rail cars are more likely to tip over, de-couple and separate from their wheel assemblies. Talgo compares the DuPont crash to a 2013 Metro North derailment where a train entered a 30 mph curve at 82 mph and claims that the Talgos performed better than the MNRR cars.

   Nevertheless, there were 2 points of coupling failure in the Talgo trainset at DuPont and the FRA needs to look into those failures and determine if the newer Talgo trainset couplings would be likely to fail in a similar accident.

   The bottom line in the DuPont accident is that there were many other points of failure in the timeline that led to the crash of 501. The main focus of the effort to prevent another crash should be on crew training, signal systems and management of the Cascades operations rather than the trains themselves.

   Talgo’s submission to the NTSB: https://dms.ntsb.gov/public/62000-62499/62431/624988.pdf

   1. The main focus of the effort to prevent another crash should be on crew training, signal systems and management of the Cascades operations rather than the trains themselves.

      Exactly.

5. There was an airplane accident somewhere in Asia (I can’t remember the actual flight) where an airplane flew into a mountain in limited visibility conditions. Everyone died. The last words on the CVR were, “What does ‘Pull up’ mean?” Basically the airplane systems were doing their job and warning the cockpit crew that they were about to fly into terrain, but the cockpit crew spoke English as a second language and didn’t understand the warning.

   In this case the engineer was in an unfamiliar locomotive type. The locomotive systems were doing their job and warning the engineer that the train was over speed. The overspeed alarm was going off, but the engineer spent 20 secs trying to figure out what it meant. Instead of reacting to the alarm and slowing down, he did nothing. If he had understood and reacted the train would not have derailed.

   So how can that failure be laid at the feet of the rolling stock?

   Additionally, it was the locomotive that derailed and pulled the cars off the track, not the other way around. Amtrak puts standard American locomotives on its Talgos. If the had locomotives with low CG’s that were designed to Talgo standards, then again this accident might not have happened.

   And once derailed, the Talgo rolling stock performed the same, or better, than traditional American rolling stock would have performed. Yes, the nylon retaining straps for the wheel assemblies were degraded, but it is unclear what impact that had.

   That said though, amtrak was planning to retire the VI’s anyhow in about 10 years. This just moves up the timeline a bit.

   1. Pilots learn a specific subset of English to communicate with air traffic controllers to avoid the problem of non-native speakers misunderstanding something, so I’d think aircraft manufacturers would use that same subset of English for cockpit indicators.

      1. If I’m thinking of the right crash, part of the issue was a divergence in safety protocol cultures as well; the crew was Russian, and Russian proximity alarms include audio cues, while Western/Anglo ones are visual-only, and they hadn’t had much training or experience on Western-style cockpits.

      2. The Russian metros are also more audio-based. In 90s most US subways I encountered had electronic signs telling the current or next station and I mainly used those. In Moscow and St Petersburg the trains had audio announcements instead: when the train approaches a station am automated voice says. “Prospekt Mira [station], doors right. Doors open themselves.” When it’s leaving the station it says, “Doors close themselves. Next station — Rizhskaya.” So I had to get used to not having the signs and having to remember what the last announcement was. And foreign movies and TV shows are dubbed rather than subtitled. I assumed it was due to a lower expectation of literacy. Russians in the late 20th century were highly educated but maybe in the early 20th century they weren’t, and many procedures had remained unchanged for decades.

      3. @Ness,

         Ground prox warnings are aural.

         The flight was Chinese with a Chinese crew. The system gave them 10 secs of warning, but they didn’t know what the words “Pull Up!” meant.

         The last words on the CVR were in Chinese. One crew member was asking the other “What does ‘Pull Up!’ mean?”

         This accident was similar. The engineer wasn’t familiar with the locomotive type and spent 20 secs trying to determine what the alarm meant. While he was trying to figure out the alarm he was speeding towards the curve.

      4. If any kind of alarn goes off, shouldn’t the driver’s instinct be to stop first and figure out what the alarm means later? If I heard a strange alarm while driving a car I’d never seen before, that’s what I’d do.

   2. But the overspeed alarm was that he was over 79mph, which is the max allowable speed for passenger operations on the corridor.

      The overspeed alarm was not warning him that he was overspeed for the curve he was about to enter.

      If he had reduced from 82mph to 79mph, the overspeed alarm would have shut off. And the train would have still derailed.

      1. But the overspeed alarm was a distraction from normal operations. If the engineer hadn’t been trying to figure out what the alarm was about he might have noticed the upcoming curve and been able to reduce the train’s speed before the curve.

   3. The overspeed alarm had nothing to do with the speed limit at that particular location. It was telling the engineer that the locomotive was going over its preset maximum speed (79 mph).

6. The reason the train crashed was human error. That’s it, it derailed because the engineer was going almost 3X the posted speed. Was the training crap? Yes. Would adequate training (the training that occurred obviously wasn’t) possibly have prevented this? That seems highly likely. Would PTC have prevented the accident? Almost assuredly. Now, who’s responsible for continually pushing out the “deadline” for mandatory PTC?

   Finally, did the trainset in any way contribute to the crash. The Talgo cars certainly did not. The locomotive could possibly be implicated although again it’s more of a training issue and… The fix is mandating PTC. Why isn’t PTC in place? The most common excuse is lack of funds. So, if there’s a “throw money at the problem” mentality isn’t it insane not to put all of that money into PTC?

   1. Exactly. I’d go a second mile and put a custom radar trip trackside like subway cars have.

      It’s an antique but effective method of enforcing signal observance. There’s a pivoting arm hanging down from the car that if flipped upward sets the brakes. At the trackside a similar arm rises up when the signal being protected is red, and if a train trespasses into the protected block the trackside arm flips the train arm up, big holing the train.

      Something similar could be devised using ordinary traffic radar as a “backup” to the PTC if it fails or loses its GPS location. If an approaching train is overspeed, the trackside arm raises, flips the train arm and stops the train.

      A “kludge”? Yes, sort of, but a proven “fall-back” to the all-electronic PTC.

      1. The overspeed alarm was going off in the locomotive cab for 20 secs before the train arrived at the curve. If the engineer had understood what the alarm meant, then he would have slowed down and the train would not have derailed.

         Before installing more automated systems, it would make more mandate that engineers be checked out in a locomotive type before operating that locomotive type in revenue service.

         Additionally, standardized crew alerting systems would help.

      2. There’s no reason additional training and a back-up system to PTC can’t both be implemented. It’s impossible to eliminate human error. It’s also impossible to build a 100% fail safe system. Accidents are almost always a convergence of failures so doing both would be prudent.

      3. I agree with both of you. Let’s have more training, more standardization of alerting systems, AND independent back-ups at dangerous places. PTC won’t fail very often; engineers aren’t “new” to territory for very long; crews become familiar with new types of equipment in a short time.

         It’s rare that two of those failure points occur simultaneously, and that three occurred at Dupont is a Black Swan. But Black Swans do occasionally happen so there’s a place for crude, “shut it down at all costs” systems at places where “value engineering” creates hazards.

      4. As I noted in another comment – the overspeed was because he was exceeding the 79mph maximum speed limit – not because he was going too fast for a 30mph curve.

         If he slowed from 82 to 79, the alarm would shut off.

      5. K H, Iwasn’t writing about the over speed indicator. You’re right that the train was “barely” speeding and that 79 would have derailed the train also.

         What the original comment addressed was the opportunity for a radar actuated trackside “trip”.

         Yes, it’s “old technology” but a backup for PTC which is bespoke software used nowhere except railroads.

   2. PTC is in place in that area now, and it was installed before the crash but it wasn’t active yet because it was still in testing. And did I hear it they finished installing it in the entire north-south corridor in Washington a year or so ago? Other railroads have been slower about installing PTC, because it’s a large capital cost that they can’t recoup in short-term profits.

      1. The railroads will continue to stall until they are forced to comply. It really hurts short term profits if the government safety regulations are enforced and operations ceased instead of perpetuating the culture of endless extensions to the “deadline”.

      2. The Class I railroads supported the 2008 legislation that mandated PTC. But the law didn’t provide any money to help with development or technical assistance which left the railroads having to do all the development and implementation on their own dime and time. There wasn’t an off-the-shelf system ready to plug and play. Full implementation of PTC is going to cost over $10 billion.

         PTC isn’t a moon shot, but it is a major technical feat that ran into more than a few roadblocks, including, delays caused by the FCC refusing to allocate sufficient radio bandwidth for the system.

      3. Of course, if it threatens trips by Congress members, then money will be found in an instant.

         Every time the Air Traffic Control system was threatened by government shutdowns, tax money got allocated to keep that system running.

      4. The FAA and Congress have been kicking the can down the road for decades about upgrading the ATC system. But Congress hasn’t delegated the responsibility for funding and implementing the new system to the airlines.

   3. It is simply astounding that an engineer would not remember that 10 minutes out of Tacoma there is a 30 mph curve.

      1. Steve Lee is the primary locomotive engineer in the Union Pacific steam program. He has operated the 844 into Salt Lake City perhaps 100 times. He is still not considered qualified for the route and has to have a pilot crew member with him.

         To operate a route completely from memory really requires more than several trips.

7. The $100 million would be better spent on a new bridge over I-5 in Dupont.

   More stupid Federal rail policies. Meanwhile the FAA is bed with the airlines while with rail the Feds do everything they can to severely handicap with BS regulations.

   1. I don’t think the FRA is anti-passenger-rail as much as it has the “not infented here” syndrome Frank mentioned. Some congresscritters and the president are anti-passenger-rail but that’s a different thing and it has to do with taxes and car/plane privilege. The US has a very different economic model for railroads than European or Asian countries. In Europe most rail is government-owned and passenger-priority, and there isn’t as much freight because it goes on trucks instead. In the US the railroads are land-grant corporations and focused on freight profits, which are higher than passenger profits. The railroads have found a niche in moving commodity freight at a low cost, so they;re like the Wal-Mart of distribution. As such they’re not interested much in speeding up the trains or maintaining the track well because that doesn’t increase profits. A distributor in Florida doesn’t care whether the container of iPhones arriving today was shipped from Seattle two days ago or three days ago or seven days ago as log as a container of identical iPhones arrives today.

      1. there isn’t as much freight because it goes on trucks instead.

         No, and no. There isn’t as much freight because there’s less coal to haul around long distances overland, which is traditionally the American Class I’s biggest strength. Freight in Europe writ large goes by sea more than in the US, and in countries with large volumes of overland heavy freight, i.e. Switzerland and Sweden, it goes by rail at the same rate as in the US.

         And then there’s China, which as of a few years ago has a higher density of rail freight than the US and a higher density of passenger rail than any European country, and does both on the same network (as does Switzerland), rather than just passengers like Britain or France or just freight like the US.

      2. As such they’re not interested much in speeding up the trains or maintaining the track well because that doesn’t increase profits.

         I don’t agree about “maintaining the track”. Just go look at any Class 1 main line and you’ll see precision of a very high degree. There are two-and-a-half to three feet of ballast supporting concrete-tied 1/4 mile welded rails of 156 pounds per yard and signals which can be seen for four to five miles if a straight-line path is available.

         Track maintenance is why there are two western, two eastern and two Canadian railroads, plus a couple of tiddlers. MofW is enormously expensive, and the line that can mechanize and replicate it best is going to “win” in the end. Better track means fewer derailments which can be ruinous. Better track means higher speeds and fewer set-outs for damaged equipment.

         It’s the right way to run a railroad.

8. They need a crash test course for trains. Sure, 50 million a pop, but could you imagine how much popcorn you could sell. I would go pay to watch..

9. Interesting some conflicting or incomplete elements:

   Alon Levy: NTSB ” demands that Amtrak withdraw the passenger cars used.”

   WSDOT: “The NTSB doesn’t have the power to regulate railroad operators, so the state is free to continue using them, she said.”

   Alon Levy: “As far as factors that are controllable by the coach builder go, the Talgos performed well.”

   NTSB: “Unlike on typical trains, whose wheel and axle frames are horizontal under the car, the Talgo railcars in the DuPont crash had wheels at the bottom of large vertical posts that the NTSB said broke loose on impact. Nylon straps that secured wheelposts to the railcars were past their useful life, a finding said.”
   “NTSB staff said ….while a third (casualty) was struck by a flying wheel assembly”

   TALGO: “…maintains that the critical damage to the car containing passenger fatalities came from the concrete embankment, not from the projectile truck.”

   NTSB: ” When the articulated connections failed, it resulted in a secondary collision that caused railcar AMTK 7424 (8) to roll over onto its roof and collapse its structure which dislodged the windows and allowed passengers to be ejected.”

   1. Talgos: *stay upright, resulting in fewer deaths than at Spuyten Duyvil and Frankford Junction*

      NTSB: *recommends withdrawal of equipment from service, which it didn’t at Spuyten Duyvil or Frankford Junction*

      The NTSB is in full nobody-gets-fired-for-buying-IBM mode, with one important exception: Amfleets and Shoreliners are not actually IBM – they’re more like Soviet computers, developed without regard for actual standards in the parts of the world where the technology works.

10. Impact loading of structures is quite different than static loads. It’s worth pointing out that the USA concentrated on static load bearing. The car structure is put in a large compression machine and loaded up over a day to the required point. Until they actually collide with something, nobody really knows how well collision survival will work.

    It’s expensive to collide railroad equipment, so they do the tests with stuff that is retired.

    So, we could never know how well the updated standards work until 40 or so years go by, when equipment built to that standard is retired and actually run through collision tests.

11. The NTSB report has two important sections. The part about the total breakdown in safety culture and the administration of the Amtrak Cascades operation – something that crossed three different agencies and a private freight company – is very well written and is a total indictment of these agencies.

    The section about the Talgo cars themselves is absurd and as noted previously, the conclusion that the fatalities were caused by the concrete embankment and not the cars themselves makes you wonder what the NTSB was thinking, writing that section.

    Now back to the first part…

    It is mind-boggling to see the layers upon layers upon layers of failures in these agencies that led to this crash. From inadequate engineering to poor training, there is blame to go around. And while some of these issues were diffused due to the number of agencies involved, making it hard for anyone to see the “big picture” of what was going on – at the end of the day any one of them could have said – stop – this isn’t right and we need to fix this before we proceed.

    Not a single one did. All those managers and administrators – smart, educated, well paid individuals, just let the proverbial train keep running down the tracks – right into disaster. Is this an issue of the individuals at their desks not doing their jobs – or the leadership that sets the culture and pushes these employees towards excellence (or in this case, not)?

    And finally, to the engineer and the conductor in the training in the cab.

    Literally WTF were they doing?

    In operations like this – a big concern is complacency. You have an engineer (or a bus driver, or whatever) that does the same route – every day – for years. They know the track route like the back of their hand. And an accident occurs because they’ve become complacent – they have relaxed their awareness because they are comfortable – and they go barreling into a turn at a high speed because they just aren’t paying close enough attention to what they are doing.

    In this case – even though the engineer’s training was suspect (at night only) this was a new engine, on a new track, on the first day. That curve was specifically covered. He knew exactly when he diverted off of the old track – that he was very familiar with – and onto the new track. He should have had heightened awareness to the upcoming hazards. And if he didn’t know where he was – but knew a sharp, slow curve, was ahead, he should have preemptively slowed down until he got his bearings.

    And the conductor in training? What was he doing, watching Sportscenter on this phone? The entire reason he was there was to learn the track, the mileposts, and the hazards along the territory. He should have been following along, closely, with his timetable, and communicating with the engineer about the hazards. He should have been a second set of eyes watching the milemarkers – and with nothing else to do but internalize what he was seeing against the timetable – should have been able to identify the need to slow down even when the engineer – busy with the controls – did not.

    The fact that he did none of this shows how broken Amtrak’s safety culture really is.

    What were the expectations for his qualification on the territory?

    How often do conductors training actually meet the qualifications – or is this just thought of as “something to do to make a few more hours.”

    We lost two great individuals – transit advocates – train advocates – guys who talked the talk and walked the walk. They supported and fought for Amtrak and specifically this line. The negligence of their agencies they believed in so strongly – that they spent so much time advocating for – ended up causing their deaths. A year and a half later, it’s still hard to comprehend.

    1. I think much could be learned from having amateur train and bus driver camps. I can’t drive one — and not having that experience does limit my ability to fathom some decisions and choices. It’s bad enough that we have administrators, board members and stakeholders making choices without frequent experiencing riding first hand; almost none of these people actually drive a vehicle.

       I can only imagine how many stupid transit design decisions get made just because these influencers don’t fully understand how to drive these vehicles! A “transit driving camp” of a few days for all administrators, planners, designers, board members and maybe major stakeholders would pay off in spades! I think that it’s too bad that no agency or university popularizes this idea for these groups.

    2. If I were to zero in on one factor that I would say was an overall driver of how decisions were made or not made was that the process was – Deadline Driven.

       It should have been Checklist Driven.

       Each move forward should have only happened after the success of the previous checkpoint.

       It should have been “The service will start after All Engineers and Conductors have qualified with at least 8 ‘at the controls’ runs”

       Rather than claim incompetence, it might have come down to an unfortunate misplaced sense of pride. As in “It’s hard, but I Can Do This”.

Comments are closed.