Report of collision Bedford (Elstow) - 19/06/26

[Haywain]

That gives us a chronology, but very little in the way of explanation

That's because it's an interim report.

 

[Dave W]

One wonders if what the RAIB have written is a little unclear on this. I’d be surprised if no train crossed over with the signal initially displaying red. What they presumably mean is that on all occasions it had stepped up to a proceed aspect by the time the approaching train reached it.

This would be the only time it *remained* at danger, which is quite an important distinction.

This feels a little pedantic - surely by "encountered" they mean at the point a driver would respond to the signal.

 

[AlterEgo]

It’s notable there is pretty much nothing on human factors relating to the 360 driver.

This will probably come out in the final report. RAIB colocate with AAIB who are human factors experts. We can only hope there is a detailed deep dive into the human factors for this incident as it seems abundantly clear that was the proximal cause of the accident.

 

[LYuen]

It was a link in a chain that caused the collision, the right side failure had it been acknowledged, said train carries on, no collision. That isnt blaming anyone its just a fact, same as had that same train(1B67) not been late, again the same event, no collision. Investigators won't just look at one thing and say thats the reason, they look at all possible links in the chain including those major and make recommendations based on those to prevent the next as every link can hold the opportunity to deal with an issue that may cause the next unfortunate event.

I agree that areas of investigation should not be dropped at this stage, but it is also important the investigation not just focusing on immediate chain of technical events, but also the hidden link in systematic and human factors

Eastern Air Lines 401 [1] is a good example. During the descend, when pilot extended the landing gears, the landing gear indicators did not show locked. The pilot went on investigating the landing gear, got distracted and crashed.
It turned out to be a burnt light bulb at the cockpit, but the conclusion of the investigation wasn't "a burnt light bulb took down the plane". It was in fact cited pilot errors, and the underlying causes were poor crew resource management and company procedures.

Same for here, surely the reason for the AWS error and how the class 810 reacted to it warrant further investigation, it must be noted that all protection designed for this fault case was in effect, and why such protection hasn't actually prevented the accident should be discussed. The issue with class 810 must not be closed, but I expect to see investigation on driver training on AWS acknowledgement and interviewing drivers from different TOCs on this procedures.

[1] https://en.wikipedia.org/wiki/Eastern_Air_Lines_Flight_401. There is a good Mayday episode too.

 

[edwin_m]

One wonders if what the RAIB have written is a little unclear on this. I’d be surprised if no train crossed over with the signal initially displaying red. What they presumably mean is that on all occasions it had stepped up to a proceed aspect by the time the approaching train reached it.

I can't see how they could know that. I doubt they can do any more than analyse data logs to see what the signal was showing at the time its berth track circuit went occupied, and with no pointwork on the approach this track circuit probably goes back to signal 343 or at least to the end of its overlap. Maybe they could include a time lag, but that would involve estimating the time from occupying the track circuit to sighting the signal, which in turn depends on how fast the train was approaching and therefore indirectly on the signal aspect, which sounds a bit like using the answer to prove the answer. A signal leading out of a complex area into a long stretch of plain line is unlikely to be approached at red unless there's some out of course event beyond it as happened here.

It’s notable there is pretty much nothing on human factors relating to the 360 driver.

They note that why the driver passed the signal is the big question here, and they aren't going to present part of the answer until they have all the relevant information. But as someone mentioned above, it's unlikely we will ever know for certain.

 

[sharpener]

Are systems available where the recording server constructs the file from a camera live stream and can neatly conclude it at any time if the feed is interrupted?

You would hope so. My 15 y/o Topfield domestic PVR can do this with broadcast-standard TV so >2Mbits/s. A Ring doorbell can probably do it too. I would have assumed it was in the spec.

<CCTV generally records in 5 or 10 minute blocks, the missing video will simply be due to the file not having been completely written to disk.> Seems unnecessarily cumbersome for a professional grade system.

Vulnerability of GSM-R to external power failure is pretty worrying too. And the different and very loud warning from above driver's head in 810 cab.

I can see RAIB making a lot of further recommendations, many relating to MMI/Human Factors.

 

[MarkyT]

One wonders if what the RAIB have written is a little unclear on this. I’d be surprised if no train crossed over with the signal initially displaying red. What they presumably mean is that on all occasions it had stepped up to a proceed aspect by the time the approaching train reached it.

This would be the only time it *remained* at danger, which is quite an important distinction.

Very important. Stepping up from red may be an occasional experience for trains joining the up fast closely following an express, but an unrestricted run on the fast is likely to be well ahead as the second train snakes across the junction at less than half its speed. A cautionary proceed aspect typically seen at WH154 is more likely. This is why I'm so concerned about the possible reading through of WH152, as that showing a green may have been very persuasive reinforcement of the mistaken notion that the forward train is racing on ahead, not stopped immediately beneath the signal.

 

[norbitonflyer]

I don't believe they are - I can't recall ever seeing a lock on the door of the box containing a signal post phone.

I would expect a lock if the signal post and its telephone were in an area accessible to the public, such as a station platform, but not otherwise.

== Doublepost prevention - post automatically merged: Thursday at 14:11 ==

This is why I'm so concerned about the possible reading through of WH152, as that showing a green may have been very persuasive reinforcement of the mistaken notion that the forward train is racing on ahead, not stopped immediately beneath the signal.

From how far away can WH152 be seen? Would a driver passing WH154 be able to see it, even when a train is standing underneath it?

 

[Haywain]

I would expect a lock if the signal post and its telephone were in an area accessible to the public, such as a station platform, but not otherwise.

== Doublepost prevention - post automatically merged: Thursday at 14:11 ==


From how far away can WH152 be seen? Would a driver passing WH154 be able to see it, even when a train is standing underneath it?

WH152 is round a curve from 1H46's direction of approach, and I doubt it can be seen from WH154.

 

[IrishDave]

From how far away can WH152 be seen? Would a driver passing WH154 be able to see it, even when a train is standing underneath it?

The way I read @MarkyT's point was that, having passed WH154 at danger, the driver of 1H46 may have seen WH152 at green before seeing 1B67 stopped, and seeing that green may have reinforced the notion that 1B67 was racing ahead.

 

[Harpo]

Train Manager AND driver. And I think "heroic" overplays it a touch.

I’d agree. Just very good railway work and a task that isn’t needed that often.

TC clips though, wouldn’t have any effect on other locations with axle counter sections which adds to the concerns about the loss of GSMR.

 

[MarkyT]

From how far away can WH152 be seen? Would a driver passing WH154 be able to see it, even when a train is standing underneath it?

I reckon at around 700m on approach to WH152, based on the video I linked above, and that is ~600m after passing WH154.

I suggest the signal, high up on the gantry, is far more visible than a short dark-hued train stopped below it. The report states the train becomes visible in FFCCTV around the signal sighting point, but that doesn't mean the driver noticed it.

 

[bramling]

The way I read @MarkyT's point was that, having passed WH154 at danger, the driver of 1H46 may have seen WH152 at green before seeing 1B67 stopped, and seeing that green may have reinforced the notion that 1B67 was racing ahead.

I’d say this is absolutely a reasonable interpretation of the sequence of events, given that it has clearly taken the driver a few seconds to process what he was seeing. This doesn’t explain what was happening at the point when the red aspect was passed though.

There are essentially two related factors here. Firstly the original single yellow aspect, and then secondly the complete disregard of the red aspect. One could certainly understand forgetting the significance of the single yellow and coming up too hard against the red, but completely disregarding the red is quite perplexing.

 

[PZ 08895]

I would expect a lock if the signal post and its telephone were in an area accessible to the public, such as a station platform, but not otherwise.

The Network Rail Standard for the Installation of Lineside Telephones says:

"In areas of high vandalism equipment and telephones may require suitable protection or provision of vandal resistant types".

That normally means putting them in a lockable enclosure with a drivers No 1 key lock. In reality that depends on where they are being installed.

As someone who has previously involved in managing SPT installations, trackside telephones not on platforms maybe in locked enclosures as they can be on station platforms, but not always, depending on the location and the majority are not locked as appears to be the case at Elstow. Indeed I did once use a SPT on a station platform myself to make an emergency call to the signaller (some drunk members of the public were on the track between the platforms) without needing a key.

 

[Haywain]

forgetting the significance of the single yellow

The significance might depend on whether a single yellow with feathers is the expected signal for a 1Hxx train at that signal.

 

[saismee]

<CCTV generally records in 5 or 10 minute blocks, the missing video will simply be due to the file not having been completely written to disk.> Seems unnecessarily cumbersome for a professional grade system.

There are plenty of video formats that can be played back without "gracefully" ending the recording. I would assume it's to do with filesize but it does seem quite silly to make the most important parts of a recording be so fragile.

 

[edwin_m]

The significance might depend on whether a single yellow with feathers is the expected signal for a 1Hxx train at that signal.

I haven't seen anything about what approach control conditions apply to this signal (WH356), which might keep it at single yellow for any diverging route.

Assuming this isn't the case, as the train approaches this signal it will show a junction indicator with an aspect that depends on that of the next signal. In this situation where a non-stop train has just passed through, the aspect will depend on how far that train has got beyond signal WH154, which in turn depends on the speed of the trains and the relative time between the non-stop train passing Bedford and the stopping train starting to move. I'd say therefore that a single yellow was entirely possible but it would be very unusual, as noted in the report, for WH154 to be red on approach.

 

[Annetts key]

My thoughts so far, after reading the RAIB interim report.

83. On 9 July 2026, another class 810 unit approaching signal WH154 incorrectlyreceived an AWS warning indication while this signal was displaying a green aspect. In this case, the driver acknowledged the incorrect warning indication within the 2.7 second response window. They reported the matter to the signaller immediately, and the train was able to continue its journey.

84. Further investigation following this second incident found that the electromagnet within the AWS track equipment (which is energised only when a green aspect is shown, and interfaces with the AWS equipment on the train to produce the AWS clear indication) was positioned lower than the associated permanent magnet. It was also positioned 16 mm below railhead level, outside the required tolerance of ±12 mm. Work is continuing to understand the interface between the AWS electromagnet at signal WH154 and the AWS equipment on train 1B67, which led to the driver receiving a warning indication with the signal displaying a green aspect. RAIB has found no evidence that the AWS electromagnet was not correctly energised.

I'm concerned that the electro-magnet for WH154 AWS was not mounted at the correct height with respect to the rail level.

This is part of the tests and tests that are done when AWS track equipment is installed, renewed, replaced and should also be done during maintenance.

Here's some extracts from a Network Rail maintenance document {NR/L3/SIG/10663 Signal Maintenance Specifications, NR/SMS/PartC/AW11, Automatic Warning System (AWS)).
It says:

SERVICE B

2. Full Inspection and Test

...

2.5 Gauge the top of the AWS magnet with respect to rail head (±12mm from rail level).
This should be done by use of a line / bob weights and associated plastic gauge or rule.

2.6 Carry out NR/SMS/PartB/Test/024 (AWS Tests).

However it also gives two alternative options, one of which is:

SERVICE RE
Fix on failure.

Note that NR/SMS/PartB/Test/024 is the electrical and magnetic AWS Tests.

As the rails wear, the AWS height does not change with respect to the sleepers, but the distance between the top of the rail and the top of the magnets will reduce, not increase.

Unless the tops of the sleepers are wearing very badly, there is normally nothing else that wears that will change the AWS height.

However, if the rails are renewed (say old badly worn replaced with brand new), that makes a big difference. Especially if a different specification of rail is installed (RT60 / UIC60 replacing BS113A).

58 In the immediate aftermath of the collision, the driver of train 1B67 realised that their train had derailed and that this presented a potential risk of collision with a train passing on an adjacent line. As the train had lost power in the collision,the driver could not make a GSM-R REC. They therefore worked with the train manager to place track circuit operating clips on the adjacent lines (these clips are emergency equipment which simulates the presence of a train, causing signals on the approach to change to red).

The failure of any GSM-R to fail to operate in an emergency is a concern. For it not to work on both trains is extremely concerning.

IMHO all cab GSM-R should have their own internal battery so that there is no dependence on power from the trains power system.

In the meantime, the aspiration of the railway to remove fixed lineside telephones like SPTs should be put on permanent hold until all trains have a far more reliable and dependable means of contacting the signaller.

It's good that the crew used track circuit operating clips on the adjacent lines to provide protection from another train potentially hitting the crashed trains or passengers. However, track circuit operating clips are useless on lines that use axle counters. Where axle counters are used, there is nothing that train crew or passengers can do to directly put the signals back to red on adjacent lines. They have to contact the signaller.

 

[norbitonflyer]

I reckon at around 700m on approach to WH152, based on the video I linked above, and that is ~600m after passing WH154.

I suggest the signal, high up on the gantry, is far more visible than a short dark-hued train stopped below it. The report states the train becomes visible in FFCCTV around the signal sighting point, but that doesn't mean the driver noticed it.

Thanks. Could you identify which signal is which on that video, as i'm finding it hard to follow (the signal number plates are not clear)

 

[baz962]

WH152 is round a curve from 1H46's direction of approach, and I doubt it can be seen from WH154.

It can’t be seen from WH154 , correct.

 

[MarkyT]

Thanks. Could you identify which signal is which on that video, as i'm finding it hard to follow (the signal number plates are not clear)

I've set the video link to start roughly at the sighting point for WH152. If you pause and step back and forth using L/R arrow keys, the previous signal WH154 can be found, and you can step all the way back through Bedford station to view the whole approach. If you set playback speed to 0.25, arrow steps are ~1s rather than default 5s. (Edit: note the finer stepping only works at YouTube, not using the embedded player here, where the step is always 10s regardless of playback speed setting).

Here's a still from the video above showing the approach to WH154 from the trailing crossover where a move from up slow joins the up fast.

The signal, here at green, is clearly visible from this point just under 1km away. It is the tiny point light source I have circled in yellow. That's excellent sighting in my experience.

 

[westcoaster]

I've set the video link to start roughly at the sighting point for WH152. If you pause and step back and forth using L/R arrow keys, the previous signal WH154 can be found, and you can step all the way back through Bedford station to view the whole approach. If you set playback speed to 0.25, arrow steps are ~1s rather than default 5s. (Edit: note the finer stepping only works at YouTube, not using the embedded player here, where the step is always 10s regardless of playback speed setting).

Here's a still from the video above showing the approach to WH154 from the trailing crossover where a move from up slow joins the up fast.

The signal, here at green, is clearly visible from this point just under 1km away. It is the tiny point light source I have circled in yellow. That's excellent sighting in my experience.
View attachment 212529

I have not read the interim report yet, but one thing i have not seen mentioned is the location of the class 360 that was on the down slow line in the are of the yellow dot.

 

[35B]

That's because it's an interim report.

Precisely my point

 

[Falcon1200]

I'm concerned that the electro-magnet for WH154 AWS was not mounted at the correct height with respect to the rail level.

The RAIB report does not state whether any train prior to 1B67 had reported the AWS Fault (Code 2) at WH154 signal, although there is mention of a subsequent report. Which presumably indicates that although the magnet was too low, its effectiveness was only impaired for a tiny fraction of the number of trains approaching the signal displaying green, which makes the accident an even more tragic event.

 

[LordSarnoc]

That's because it's an interim report

That’s interesting, because the Carmont (the last but one) interim report gave clear causes. As did the previous Margam report. Maybe the policy changed between 2021 and 2025.

 

[sharpener]

I've set the video link to start roughly at the sighting point for WH152.

Thanks for taking the trouble to set this up for us. What are the two repeaters at 1.01.06 and why is one on and one off?

And in the approach to various signals you can hear pips before the AWS bell, are they TPWS or something else?

 

[norbitonflyer]

I've set the video link to start roughly at the sighting point for WH152. If you pause and step back and forth using L/R arrow keys, the previous signal WH154 can be found, and you can step all the way back through Bedford station to view the whole approach. If you set playback speed to 0.25, arrow steps are ~1s rather than default 5s. (Edit: note the finer stepping only works at YouTube, not using the embedded player here, where the step is always 10s regardless of playback speed setting).

Here's a still from the video above showing the approach to WH154 from the trailing crossover where a move from up slow joins the up fast.

The signal, here at green, is clearly visible from this point just under 1km away. It is the tiny point light source I have circled in yellow. That's excellent sighting in my experience.

I understand now, it was the sighting of signal 152 I was interested in. That is the one first seen at about 1:01:54 and passed at 1:02:00.
signal 154 is passed at 1:01:50 and first seen at about 1:01:19 (maybe earlier if there hadn't been a train coming the other way with a bright headlight)

Because of the curves there is no way that 152 could be "read through" from the preceding signal 154, but the reverse curves could make it difficult to realise that there was a train under the signal (with no yellow nose, and no tail lights either if the reboot knocjked them out as well as GSM-R)

 

[Vexed]

That’s interesting, because the Carmont (the last but one) interim report gave clear causes. As did the previous Margam report. Maybe the policy changed between 2021 and 2025.

Maybe it's because of human factors? They can't say something like "1H46 received a wrong-side AWS bell vice horn at WH154 and then passed it at red" because it's not that simple.

 

[sharpener]

I understand now, it was the sighting of signal 152 I was interested in. That is the one first seen at about 1:01:54 and passed at 1:02:00.
signal 154 is passed at 1:01:50 and first seen at about 1:01:19 (maybe earlier if there hadn't been a train coming the other way with a bright headlight)

Because of the curves there is no way that 152 could be "read through" from the preceding signal 154, but the reverse curves could make it difficult to realise that there was a train under the signal (with no yellow nose, and no tail lights either if the reboot knocjked them out as well as GSM-R)

It's also notable that because of the right hand bend the corresponding signal on the Up Slow is visible before WH152, I don't know if this had any bearing if it was green at the time.

 

[anothertyke]

It's also notable that the information in this report is all attributable to what's been electronically logged. That gives us a chronology, but very little in the way of explanation

The holes in the swiss cheese are described in this interim report, three months out from the event. That's a good, timely job, but I suspect it's the easy bit. Much harder will be the whys and wherefores and what should actually be done.