Train collision in Bavaria

[bikkuri bahn]

story is still developing

http://www.dw.com/en/police-fatalities-in-bavarian-train-crash/a-19034690

Edited February 9, 2016 by bikkuri bahn

[Welshbloke]

Police now reporting at least four dead and many injured. Something has evidently gone horribly wrong, given how many layers of safety equipment modern rail networks have.

[TheNicofabi]

Its 9 dead now and about 50 injured, currently watching the news on it, I gonna try to keep you guys updated!

Edited February 9, 2016 by TheNicofabi

[kvp]

The line is essentially a single track branch line operated by an independent train operating company. (http://de.wikipedia.org/wiki/Mangfalltalbahn) The rules for these lines are very relaxed (even more than for 3rd sector railways in Japan), including permission to use station to station signalling with no safety equipment. (these branchline rules were actually made for ancient 2 axle railbuses) Also, on this railcar type it's possible to pass a signal at danger if the driver manually cancels the automatic stop system after it brings the train to a complete stop. At least one of them had to go really fast as telescoping is not something that is usually seen with these articulated railcars and they have explicit equipment to prevent this. (btw. their top speed is around 160 km/h)

 

Edit: It looks like the line has PZB90 (indusi) installed with an allowed top speed of 120 km/h, but i couldn't manage to find info on the presence of track occupancy detection and control method (centralised or local), but this system still has a manual release button that fully disables it.

Edited February 9, 2016 by kvp

[TheNicofabi]

The line is essentially a single track branch line operated by an independent train operating company. (http://de.wikipedia.org/wiki/Mangfalltalbahn) The rules for these lines are very relaxed (even more than for 3rd sector railways in Japan), including permission to use station to station signalling with no safety equipment. (these branchline rules were actually made for ancient 2 axle railbuses) Also, on this railcar type it's possible to pass a signal at danger if the driver manually cancels the automatic stop system after it brings the train to a complete stop. At least one of them had to go really fast as telescoping is not something that is usually seen with these articulated railcars and they have explicit equipment to prevent this. (btw. their top speed is around 160 km/h)

The part where the accident happend is 120km/h max but someone seriously fucked up... or something... Still waiting for the news to say what caused it.

[bikkuri bahn]

Apparently the signal system in use on this line is the pzb90.

https://en.m.wikipedia.org/wiki/Punktförmige_Zugbeeinflussung#PZB90

[TheNicofabi]

Apparently the signal system in use on this line is the pzb90.

https://en.m.wikipedia.org/wiki/Punktförmige_Zugbeeinflussung#PZB90

as everywhere in German, yes. From what was said the system at this line just got checked a week ago, so they doubt it is the reason.

[railsquid]

The line is essentially a single track branch line operated by an independent train operating company. (http://de.wikipedia.org/wiki/Mangfalltalbahn)

 

 

To be precise, the operating company is the Bayerische Oberlandbahn.

 

he rules for these lines are very relaxed (even more than for 3rd sector railways in Japan), including permission to use station to station signalling with no safety equipment. (these branchline rules were actually made for ancient 2 axle railbuses) Also, on this railcar type it's possible to pass a signal at danger if the driver manually cancels the automatic stop system after it brings the train to a complete stop. At least one of them had to go really fast as telescoping is not something that is usually seen with these articulated railcars and they have explicit equipment to prevent this. (btw. their top speed is around 160 km/h)

 

Quellenangaben?

[Welshbloke]

While not wishing to speculate, this sounds remarkably like the infamous Abermule train crash in 1921. In that case the token/tablet system fell down when multiple people failed to check that a train crew had the right tablet for the section, leading to a head-on collision between an innocent express and a stopping service whose crew had been handed a pouch containing a tablet for a completely different section.

[kvp]

Quellenangaben?

http://hu.wikipedia.org/wiki/Stadler_FLIRT

http://de.wikipedia.org/wiki/Stadler_Flirt

http://de.wikipedia.org/wiki/Mangfalltalbahn

http://de.wikipedia.org/wiki/Eisenbahnunfall_von_Bad_Aibling

http://en.wikipedia.org/wiki/Punktförmige_Zugbeeinflussung

also please google SpDrS60 for more information on the control system (it's a bit old and allows undetected aka. dark blocks, but i can't find info if this line had any)

the PZB90 override button is labeled "pzb frei" and should only be used with care (this may or may not have anything to do with the accident)

[bikkuri bahn]

good article about the pzb system, and *possibly* what went wrong:

http://www.dw.com/en/there-are-technical-problems-in-railway-operations/a-19036367

[HantuBlauLOL]

max 120km/h with those rules? seriously..

[kvp]

 http://de.wikipedia.org/wiki/Eisenbahnunfall_von_Bad_Aibling

Betrieb

Die Zugkreuzung der beiden Züge findet fahrplanmäßig im Bahnhof Kolbermoor statt, wobei Zug 79506 einen Aufenthalt von 5 Minuten (6:40 bis 6:45 Uhr) hat, während Zug 79505 nach Ankunft um 6:44 Uhr fahrplanmäßig ohne weiteren betriebsbedingten Aufenthalt weiterfahren soll. Zug 79505 hatte jedoch 4 Minuten Verspätung. Zug 79506 fuhr laut Reisendeninformationssystem pünktlich in Kolbermoor ab und damit in den eingleisigen Streckenabschnitt ein.

 

So one of the trains seem to have left either against the exit signal or got a green one into an occupied block. The limit in the curve was only 100 km/h, but 100 km/h x 2 is still fast. A continous signalling system (like the ancient hungarian variable frequency based EÉVB or the american pulse code based PRR-PCCS) would show any oncoming train or lone car (actually anything and anyone shorting the track circuit) in the same block as an emergency stop signal. The german signal wire based LZB (mandatory only for lines above 160 km/h) is also usable, but doesn't detect track occupancy and rail breaks. Last year, hungary had two near accidents that were prevented by the trains seeing each other's occupancy signal and stopping before a head on collision, the initial cause was manual exit signalling due to track reconstructions. The last actual deadly accident many years ago was caused by sabotaged track circuits miswired by cable thiefs to avoid theft detection. In the german PZB90 system, there is no way to stop the trains after they cleared the exit signals, until they pass over one of the balises, usually placed next to signals, so any extra safety on the lines between signals are provided by the train radios and the alertness of any signalling crew present on the line.

[velotrain]

 So one of the trains seem to have left either against the exit signal or got a green one into an occupied block.

 

What I'm not getting is that these were commuter trains, so they had a daily routine, so the crews should have been used to passing each other at a given location each morning.  I'd like to think that one or both crews would have said / done something when the routine was changed - unless somehow the fasching holiday played a role in terms of changed schedules.

[marknewton]

G'day Charles,

 

The trains themselves have a daily routine, but the crews may not. That would depend on how they are rostered. I don't know what the current German practice is, but I know that in the past it was common for crews to be rostered to work a different job on every day of a ten-day fortnight. Depending on the number of crews on the roster and the length and arrangement of the roster cycle they could go for weeks between doing the same job twice. So it's possible that one or both of these trains were being worked by crews who weren't used to the routine.

 

Another aspect of this incident to consider is this - according to one news report that I've read;

 

"A spokesman for the train operator, TransDev, a French company, confirmed that a driving instructor together with an apprentice train driver had been travelling in the traction unit of one of the trains."

 

I don't know whether this means one of the trains was being driven by a trainee under instruction, or they were doing road knowledge instruction or traction qualification. I don't know anything about how driver training is currently conducted by this operator, but either way I think it's a significant bit of information.

 

Cheers,

 

Mark.

[westfalen]

 http://de.wikipedia.org/wiki/Eisenbahnunfall_von_Bad_Aibling

 

So one of the trains seem to have left either against the exit signal or got a green one into an occupied block. The limit in the curve was only 100 km/h, but 100 km/h x 2 is still fast. A continous signalling system (like the ancient hungarian variable frequency based EÉVB or the american pulse code based PRR-PCCS) would show any oncoming train or lone car (actually anything and anyone shorting the track circuit) in the same block as an emergency stop signal. The german signal wire based LZB (mandatory only for lines above 160 km/h) is also usable, but doesn't detect track occupancy and rail breaks. Last year, hungary had two near accidents that were prevented by the trains seeing each other's occupancy signal and stopping before a head on collision, the initial cause was manual exit signalling due to track reconstructions. The last actual deadly accident many years ago was caused by sabotaged track circuits miswired by cable thiefs to avoid theft detection. In the german PZB90 system, there is no way to stop the trains after they cleared the exit signals, until they pass over one of the balises, usually placed next to signals, so any extra safety on the lines between signals are provided by the train radios and the alertness of any signalling crew present on the line.

 

 

"A spokesman for the train operator, TransDev, a French company, confirmed that a driving instructor together with an apprentice train driver had been travelling in the traction unit of one of the trains."

 

I don't know whether this means one of the trains was being driven by a trainee under instruction, or they were doing road knowledge instruction or traction qualification. I don't know anything about how driver training is currently conducted by this operator, but either way I think it's a significant bit of information.

 

Cheers,

 

Mark.

Those are both new bits of information to me.  I'm surprised the signalling system does not show block occupancy though if the other train had already passed the opposing signal it wouldn't have mattered.

 

I notice from the diagram on Wikipedia that the collision occured about half way between the two crossing loops, around three kilometres from each, even at 120kph that is plenty of time for the controller/dispatcher/signalman to send out an emergency radio message for both trains to stop after one of the trains passed a red signal, on QR if a train passes a red its driver and that of every other train knows about it in seconds.  This leads to a couple of questions, does the German dispatcher get a indication that a train has spadded and was there two way radio in use on the line?  If both answers are yes then I'm leaning toward a wrong side signal failure giving both trains a green.  The only definite answer would be eyewitnesses who saw both trains pass their respective signals and could testify as to the aspects.  I wonder if the trains had driver's view video cameras or if the stations had cameras with a view of the signals.

Edited February 11, 2016 by westfalen

[kvp]

 

This leads to a couple of questions, does the German dispatcher get a indication that a train has spadded and was there two way radio in use on the line?

This depends on the number of blocks on the line. If there is only a single detection block between the two opposing exit signals, then if a train is already in the block, there is no way for the signaller to know if a second one has also entered into the same block. But this assumes at least one intentional spad or a signalling system failiure.

 

Third and unlikely option is signal delay. There is a small chance that the signal was green for one of the trains when it started moving, passed the signal, then after this point its exit signal was dropped back to red and given green on the other end and if the detection blocks are sparse, then there is a chance that the first train didn't enter the detected area before this happened. This can be caused by having dark spots across the turnout districts, so there is some time after a train leaves an exit signal before it is seen on the line. So if the first train leaves the station before its signal is dropped back to red but enters the detected block after the other train gets a green, then the pzb system can't catch the first train anymore, but the signaller will think that the 2nd train triggered the block occupancy and only sees the free station track of the 1st train some time later when the last axle has left it. This can only be triggered by setting up a route, then cancelling it manually, which opens up this time window that depends on the speed of the first train and the length of the undetected turnout district. This is very unlikely. (a simple spad override by the driver sounds more realistic) Anyway a good defense against this scenario is having at least 3 blocks between two exit signals, the 2 turnout districts and one open line, with bidirectional stop emitters somewhere around the entry signals. (at the border of the station turnout district and the open line) Unfortunately for most scenarios, this signalling setup is considered overkill. Also if the signaller has to ask for position confirmation from a driver before cancelling an already set up route, then this case can't really happen, even without extra detection and stop equipment.

 

ps: We have to wait and see what happened, as there were data recorders on board both trains and there should be one somewhere in the signalling system.

[velotrain]

ps: We have to wait and see what happened, as there were data recorders on board both trains and there should be one somewhere in the signalling system.

 

Will it take a year as NTSB plans to with the Amtrak derailment in Philly, or are the European authorities better organized?

[kvp]

They can't really move the wreck until it's done and this year the line was scheduled to be used as a spare route for mainline traffic while the high speed lines get upgraded. I guess they say someting fast and then do a more in depth analysis later.

[velotrain]

I really question why NTSB takes so long.  Are they really working all day for those 365 days (minus weekends and holidays)?  Just what will they discover in those 11 months beyond the first?  To the extent that whatever is learned could help prevent future problems, you'd think they would want to reach a conclusion much more quickly. 

I'm not suggesting a rush to judgment, but just don't understand the calculated delay.

[bikkuri bahn]

Going a bit OT wrt to the NTSB, but here is my (*cynical version*) take: One year not only allows plent of time for investigation, it allows plenty of time for the media to forget about the accident.  Thus when the results turn up 365 days later, the general public and legislators have forgotten about the whole thing, and the railroads can delay installing safety equipment for another three, or four, or ~...years.

[westfalen]

This depends on the number of blocks on the line. If there is only a single detection block between the two opposing exit signals, then if a train is already in the block, there is no way for the signaller to know if a second one has also entered into the same block. But this assumes at least one intentional spad or a signalling system failiure.

 

Third and unlikely option is signal delay. There is a small chance that the signal was green for one of the trains when it started moving, passed the signal, then after this point its exit signal was dropped back to red and given green on the other end and if the detection blocks are sparse, then there is a chance that the first train didn't enter the detected area before this happened. This can be caused by having dark spots across the turnout districts, so there is some time after a train leaves an exit signal before it is seen on the line. So if the first train leaves the station before its signal is dropped back to red but enters the detected block after the other train gets a green, then the pzb system can't catch the first train anymore, but the signaller will think that the 2nd train triggered the block occupancy and only sees the free station track of the 1st train some time later when the last axle has left it. This can only be triggered by setting up a route, then cancelling it manually, which opens up this time window that depends on the speed of the first train and the length of the undetected turnout district. This is very unlikely. (a simple spad override by the driver sounds more realistic) Anyway a good defense against this scenario is having at least 3 blocks between two exit signals, the 2 turnout districts and one open line, with bidirectional stop emitters somewhere around the entry signals. (at the border of the station turnout district and the open line) Unfortunately for most scenarios, this signalling setup is considered overkill. Also if the signaller has to ask for position confirmation from a driver before cancelling an already set up route, then this case can't really happen, even without extra detection and stop equipment.

 

ps: We have to wait and see what happened, as there were data recorders on board both trains and there should be one somewhere in the signalling system.

None of that seems like a very 'safe' safeworking system to me, there are so many holes in it that you couldn't come close to calling it failsafe.

[velotrain]

Exactly my theory also.

[kvp]

The Amtrak investigation took a week and then they added speed restriction emitters to all mainline curves used by sprinters. The rest is paper pushing.

[Suica]

German media outlets report that the dispatcher confessed. 

He confessed that he gave the delayed train ZS1 permission to enter the section beyond the red signal. He did realise his error but his emergency call didn't reach the trains. 

He's being charges with involuntary manslaughter, physical injury resulting from negligence and dangerous manipulation of rail traffic. 

 

However, when a train receives a ZS1 permission, it's allowed to only proceed with max. 40 km/h. Earlier reports stated that both trains collided at nearly 100 km/h. 

 

ZS1 = three white dots on red signal

 

Source: http://www.welt.de/vermischtes/article152293809/Fahrdienstleiter-gesteht-Ursache-ist-menschliches-Versagen.html

Edited February 16, 2016 by Suica