It seems more rail crashes in 2026 .Sad day in Spain, 64+ dead in high speed rail crash.
https://www.reuters.com/world/two-high-speed-trains-rear-end-spain-rail-operator-says-2026-10-28/
Seems 2026 not good year
started with
2x in the same year!
I’m curious if it was a human error or not. Surely there’s some system built in to automatically detect incoming collision? Idk about trains but it seems like this type of thing should never happen if they are on set tracks.
It’s not like planes where they could be anywhere.
Trains can onlt be on a train track. Surely not having 2 trains hit each other on the same track would be the first thing you get right with multiple points of fail safe that can withstand even some points of failure happening.
I guess the details are still emerging but I did read a report (Reuters) that the traffic control system failed but for inexplicable reason , breaking protocol, a human controller gave a go-ahead to one train.
link: How a high-speed train crash unfolded in central Spain - https://www.reuters.com/world/europe/how-high-speed-train-crash-unfolded-central-spain-2026-10-28/
Since its 1964 launch, Japan’s Shinkansen has carried over 6.5 billion passengers with zero fatalities.
The Shinkansen’s legendary operation and safety record | High Speed Rail Alliance
I understand that human error is behind this awful incident.
I also understand that the two trains are operated by different companies? That’s something we’d never see on Japan or Taiwan’s HSR systems (though it does appear on regular rail in Japan).
Guy
And that is why Taiwan’s HSR decided to switch halfway through designing the entire thing based on the European system. There was a huge ICE accident at the time, and they decided it’s safer to go with the Japanese system, despite all the contract violation penalties.
I had always assumed that Lee Teng-hui used the convulsive impacts of the 921 earthquake as a way to pivot to get what he wanted: Japan’s shinkansen trains, which are of course engineered to deal with similar earthquakes in Japan.
Did I get this wrong?
Guy
Well, Taiwan HSR picked the European system in September 1997. Then the Eschede train disaster happened in June of 1998. The Japanese Shinkansen began promising that their system won’t allow similar accidents, and that they would give Taiwan HSR low interest loans. The European system offered to match Japan’s loan rate, but then 921 happened on Sep 21 1999, and the switch to the Japanese system finalized on Dec 28 1999.
So while the 921 earthquake definitely played a part in the final decision, the pivot was already put in motion a year earlier, otherwise I doubt they’d be ready to make the switch 3 months after 921.
Since that time, Deutsche Bahn has seem get worse over time (and seldom on time nowdays). Used be at top of Euro trains, now not so much.
I recall in one of the NVIDIA GTC I attended remotely, Deutsche Bahn was saying how they used AI to improve scheduling. Guess it didn’t really work in the real world?
Maybe this is better to DB
Since the focus is on signaling, I’ll include this link to describe the (probable) system involved.
The section on ERTMS implementation strategies describes the way it works and gives an example of the mixed operations on the Spanish high speed line to Malaga
While maintaining cleanliness, profitability, and generally being on time.
It’s very impressive.
I don’t think scheduling is the main issue, but hardware malfunction, never-ending construction and lack of digitalization. I think Belgium’s system is fully digitalized while the Deutsche Bahn often still relies on systems from the dark ages.
Madrid-Sevilla line is the oldest in Spain and still not fully in ERTMS. The equivalent system (LZB) went offline and several trains automatically stopped. Short after, for some reason a train was verbally ordered (from central command) to restart at normal speed while another train was stopped several km ahead. Despite various calls, including several from the conductor of the stopped train saying she couldn’t restart, and a conductor of a third train saying that “normal speed” was against protocol and it should be “visual speed” (40 km\h), nobody tried to stop the incoming train.
Not the AI thread. But this is one reason AI might be more useful in the future. Autonomous driving is already proven to be safer with less human errors. Even those videos you see of those autonomous taxi’s doing weird stuff, humans do that all the time like drive the wrong way.
Is this because still doing upgrades from the old Eastern systems, or has it all be upgraded? A few of the Eastern Euro nations still has hardware from the USSR times (though a lot has updated)
That is quite sad, as it looks like human error (s), hopefully they fix it and spend a more time (and money if needed) to be 100% safe (as is Japan & Taiwan)
But the system is designed to be human error proof. So either a machine or another human should detect and correct a human error. And vice versa. The question is why this crash still happened. Specially just 9 months after another terrible one.
It should be at least.
As mentioned in my first post. How did they fuck this up so badly? It’s bizzare.