Urgent Safety Message from RAIB

[Jeremy Cumberland]

2 hours ago, DY444 said:

Translation: Urgent Safety Message - Drivers need appropriate route knowledge before driving trains through a given location and to pay attention to signals when they drive through that location.

Well, that's part of it, but it does not appear to be the opinion of RAIB, whose Urgent Safety Advice is addressed firstly to Network Rail.

 

It is probably worth noting that the USA implies that in both cases, what the driver saw was a change from red to some other aspect, and they didn't have continuous visibility of the illuminated JI as they approached the signal. Both drivers clearly saw the change from red (to green?), but they appeared not to observe the JI light up at the same time, and there might be a reason for this.

[Michael Hodgson]

1 hour ago, Jeremy Cumberland said:

Well, that's part of it, but it does not appear to be the opinion of RAIB, whose Urgent Safety Advice is addressed firstly to Network Rail.

 

It is probably worth noting that the USA implies that in both cases, what the driver saw was a change from red to some other aspect, and they didn't have continuous visibility of the illuminated JI as they approached the signal. Both drivers clearly saw the change from red (to green?), but they appeared not to observe the JI light up at the same time, and there might be a reason for this.

The JI would have lit up before the green, albeit only marginally so.  With relay locking the delay is noticeable, though not with more modern technology.  JI has to be proved alight when the road is set for divergence before the signal can clear.

[St. Simon]

7 hours ago, DY444 said:

All this waffle about how far out the signal is etc looks like smoke and mirrors.  It seems to me that the drivers concerned in these incidents didn't know the road properly and shouldn't have been driving over it.  I suspect both being OAO may not be a complete co-incidence.

 

Hi,

 

A signal being a long way out from a Junction is a valid concern, the further the signal is out from the divergence, the larger the opportunity for loss of situational awareness, regardless of the drivers competence, route knowledge or experience.

 

You could have the most competent driver who knew the route like the back of their hand, who passes the signal, takes in the information shown and control their train accordingly and then for a couple of second their mind wanders (can happen to anyone, no matter how diligently they carry out their duties). When the driver focusses on their task, they could have lost their situation awareness as nothing has happened in a while (relatively), and potentially helped by the confirmation bias of 'my train doesn't normally stop here', they take power as they have always done.

 

By having a signal closer in, there is less chance the driver can loose situation awareness and take an unsafe action prior to the divergence.

 

I think it is unfair to totally dismiss this incidence as 'lack of route knowledge', as usual with accidents, there will be a number of factors involved.

 

Simon

[DY444]

26 minutes ago, St. Simon said:

A signal being a long way out from a Junction is a valid concern, the further the signal is out from the divergence, the larger the opportunity for loss of situational awareness, regardless of the drivers competence, route knowledge or experience.

 

I don't agree because I see it as no different to say, for example, passing a single yellow in a 3 aspect area and "losing situational awareness" by not braking adequately for the expected red 2000 odd yards ahead and then finding a failed train 100 yards beyond the overlap of that signal.  Nobody would consider blaming the distance between the warning and the hazard in that scenario and so I don't accept it as a justification here.  Unless there is evidence the JI wasn't lit there is, imo, no excuse (and as you will be aware the JI has to be proved lit before the signal will clear). 

 

Also in a broader context, there should be no excuse for losing situational awareness anyway because not losing it is a fundamental part of the job.

[keefer]

Did it not used to be the case that with a slow-speed divergence, the signal would only clear to single yellow + JI? I take it that an approach release wouldn't really work given the distance from the signal to the divergence (plenty of chance to speed up), so would a single yellow aspect not be better than clearing to green?

[Michael Hodgson]

1 hour ago, St. Simon said:

I think it is unfair to totally dismiss this incidence as 'lack of route knowledge', as usual with accidents, there will be a number of factors involved.

 

I would agree that it's not necessarily an indication of lack of route knowledge and it's more likely to be a lack of concentration.

Traditionally, yes you blame the driver for what is simply human nature. 

 

It's why more modern technological aids like correctly placed TPWS overspeed sensors etc or in-cab signalling can help.   Driverless trains can be implemented much more easily than drivless cars (ever since the Victoria Line in the 1960s - must be a very boring job), but the public lacks confidence in such ideas.  And there's the view that you need somebody on board in case of accident or breakdown, so he might as well sit at the front (even though he'll be the first one killed if it does run into something subtantial).  ASLEF strikes may result in pay rises but they may also hasten that future.

[Michael Hodgson]

18 minutes ago, keefer said:

Did it not used to be the case that with a slow-speed divergence, the signal would only clear to single yellow + JI? I take it that an approach release wouldn't really work given the distance from the signal to the divergence (plenty of chance to speed up), so would a single yellow aspect not be better than clearing to green?

 

You can have approach relase (from red) at the junction signal for a severely restricted divergence -  the signal only clears when the train has occupied a given track circuit for long enough that it must have slowed down, failing which its aspect becomes irrelevant as the train has already passed it at danger if the timer has not yet finished.  With that, the previous signal would have been automatically restricted to yellow.

 

You can also have approach release from yellow where the diverging speed is less severely lower than the main line.  With this the signal in advance of the junction on the branch is held at red, thus restricting the junction signal to yellow with JI, and the signal in rear to double yellow (if four aspect)

 

Using JIs and four aspects was nice and easy.  Then it all got complicated again.  Since the introduction of the InterCity 125s where that approach was found unduly restrictive, you can clear the junction signal to Y + JI and the signal in rear to flashing single yellow instead of double yellow, with flashing double yellow instead of green in rear of that.

 

Annother recent approach since the Heathrow line was built, you can also have Preliminary Route Indicators (PRIs) which are effectively stand-alone JIs to give advance warning of which way you will be routed at a junction.  These can be used where you have splitting distants.

 

[St. Simon]

1 hour ago, DY444 said:

Nobody would consider blaming the distance between the warning and the hazard in that scenario and so I don't accept it as a justification here.

Except, they do.

 

Overbraking of a signal by more than 50% has to be risk assesessed to ensure we don’t introduce the same risk as having a signal far out, loosing that situational awareness.

 

Loosing situational awareness is precisely why we have standards that say signals should not be overbraked by more than 50% and shouldn’t be more than 800m from the junction.

 

1 hour ago, DY444 said:

Also in a broader context, there should be no excuse for losing situational awareness anyway because not losing it is a fundamental part of the job.

 

It is also fundamental to being a pilot, yet it happens all the time and to the most competent of pilots, it’s just called being human.

 

1 hour ago, keefer said:

Did it not used to be the case that with a slow-speed divergence, the signal would only clear to single yellow + JI? I take it that an approach release wouldn't really work given the distance from the signal to the divergence (plenty of chance to speed up), so would a single yellow aspect not be better than clearing to green?

I don’t know about historically (although I don’t recall seeing anything), but nowadays signals will clear to the appropriate aspect for the sequence once the approach control conditions have been satisfied. 
 

If you restrict a signal to single yellow when it doesn’t need to be, you can give rise to the situation where drivers expect the following signal not to be at red (in a similar manner to drivers expecting signals to clear when they used to them approach releasing).

 

In this type of incident it wouldn’t have helped, the issue the USA is trying to raise is the distance from the signal to the divergence allowing time for an increase of speed after the signal clears.

 

Simon

Edited May 30, 2023 by St. Simon

[The Stationmaster]

AS I said much earlier in this thread I could see potential problems back when the Signalling Principle was revised to allow such a signal to be up to half a mile in rear of the divergence it is signalling.  The risk of a Driver being able to exceed the divergence speed is greatly increased the further back the signal is sited.   That risk existed with diesel traction although it got worse with higher powered diesels (but don't forget that HSTs had been around for some years before the Principle was relaxed to allow half a mile and of course it long post dated electric traction

 

Assessment of risk back then was not the formalised process it is today although as it was basically an engineering decision to change the Signalling Principle I don't know to what extent there was inter-departmental consultation on the risks it might create but clearly there now is - from what Simn has said (and to my knowledge from past involvement in it) formal consideration of risks in the design of new signalling,  But that is just one of the factors which needs to be considered here among the many others I have previously listed.

 

So while 'the system' might well set a trap for Drivers and give them the opportunity to build up speed well in excess of the turnout speed there are wider factors.  Because when it comes down to the nitty gritty of such an incident the Driver is the person with their hand on the controls applying the power or applying the brake - the buck stops  there.  So training and competence assessment has to take that into account to mitigate the risk imported by the design of the signalling - it goes back to the old story that when you are learning a road the bits you remember first are the awkward and unusual things.  Properly applied competence management would be checking that still remains the case with Drivers in their everyday working.

 

Suggesting a train isn't normally routed that way at this signal, or any other signal, is irrelevant.  What natters is the fact that a train can be routed this way at this signal and there is an opportunity for it to overspeed.  Recognising that fact has to be part of Driver assessment and competence management and be fed back into training and any of theh various update methods which can be used.  Failure to do that is, in my view,  on a par with the signal siting setting a trap for the Driver and is exactly one of the areas I would look at very closely when auditing Driver competence management, assessment, and training.  And after a well known incident in which a charter train Driver failed dismally comply with the aspects shown by signals there is no excuse for any operator to fall short in managing such matters.

 

Yes - the buck stopped with the Drivers.  But an awful lot should be done, and hopefully is done, to avoid putting a Driver in that position.

[St. Simon]

1 hour ago, The Stationmaster said:

Assessment of risk back then was not the formalised process it is today although as it was basically an engineering decision to change the Signalling Principle I don't know to what extent there was inter-departmental consultation on the risks it might create but clearly there now is - from what Simn has said (and to my knowledge from past involvement in it) formal consideration of risks in the design of new signalling,  But that is just one of the factors which needs to be considered here among the many others I have previously listed.

 

Hi Mike,

 

There is now a 'Driveability Assessment' which is a mandated part of the Scheme Development Process. This assessment is where representatives of all the TOCs and FOCs, Design, Project, Signal Sighting and Ops sit down, and go through the signalling design and identify any risks, hazard precursors etc with the signalling layout and then either discuss mitigations or agree that any risks are as low as reasonably practicable.

 

It is a new-ish process introduced after the Bristol Temple Meads Re-signalling (it wasn't directly introduced because of that project, I just remember it was after that project) and is in additional to all the assessments such as Signal Spacing Assessment, Signal Overrun Risk Assessment, Signal Sighting etc etc. that we have to do.

 

I'll admit that we don't have to assess a signal that is up to 800m from the junction, so the one in the incident wouldn't require assessment, however I think anything that far out would probably be questioned and reviewed, either as part of the driveability or as part of the schemes review process.

 

Simon

[PaulRhB]

4 hours ago, The Stationmaster said:

But an awful lot should be done, and hopefully is done, to avoid putting a Driver in that position.

Yes recognising trends in human nature helps assess the risk it causes and the risks it can reduce. 
 

 

On 30/05/2023 at 16:24, DY444 said:

there should be no excuse for losing situational awareness anyway because not losing it is a fundamental part of the job.

I can think of two incidents for over speeding towards another signal. In one case the driver received a correct green flag at a failed crossing and forgot the caution aspect of the distant he’d just passed before the crossing. In the other the driver had come back off sick because he wasn’t actually ill anymore and it turned out he was still severely fatigued from the disrupted sleep of the last few days causing the lapse in awareness. So that driver tried to be conscientious and yet committed an error because of that, if you tell everyone have an extra day off on us every time you’re sick to get rest then it will be abused. There’s also an awful lot of workers with newborns and various other crisis in their lives which affect sleep and concentration so we have to accept it happens and mitigate where possible. 

[Vanfit]

On 30/05/2023 at 16:51, keefer said:

Did it not used to be the case that with a slow-speed divergence, the signal would only clear to single yellow + JI? I take it that an approach release wouldn't really work given the distance from the signal to the divergence (plenty of chance to speed up), so would a single yellow aspect not be better than clearing to green?

The Signalling Principle up to around the early 1980s as interpreted and implemented on the WR was to keep the junction protecting signal at single yellow with JI until the train had passed over the AWS (thereby giving the driver a caution on the AWS)  then the aspect could step up to double yellow or green if the conditions ahead allowed. This required separate train detection between the AWS and associated signal, often an (unreliable) overlay track circuit.

[Pannier Tank]

This topic reminds me of an incident at Bletchley with a LE Class 90.

 

See: Bletchley derailment

 

 

[Michael Hodgson]

18 hours ago, PaulRhB said:

 There’s also an awful lot of workers with newborns and various other crisis in their lives which affect sleep and concentration so we have to accept it happens and mitigate where possible. 

 

The signalman on duty at the Thirsk accident in 1892 had a sick child, had walked miles to get a doctor, hadn't slept in 36 hours and the child died.  So he asked the stationmaster to be relieved only to be told there was nobody available to cover, so he had to work.  He fell asleep, woke with a goods train at home signal in thick fog, block instrument showing train in section and erroneously cleared out for it thinking it should be long gone. 

 

He was found guilty of manslaughter but given an absolute discharge on jury's recommendation, and the company was criticised for contributory negligence.  Fireman failed to comply with Rule 55 and the signalman in adjacent box knew of the signalman's personal circumstances and failed to query why train had been such a long time in section.

[Reorte]

20 hours ago, PaulRhB said:

Yes recognising trends in human nature helps assess the risk it causes and the risks it can reduce. 
 

 

I can think of two incidents for over speeding towards another signal. In one case the driver received a correct green flag at a failed crossing and forgot the caution aspect of the distant he’d just passed before the crossing. In the other the driver had come back off sick because he wasn’t actually ill anymore and it turned out he was still severely fatigued from the disrupted sleep of the last few days causing the lapse in awareness. So that driver tried to be conscientious and yet committed an error because of that, if you tell everyone have an extra day off on us every time you’re sick to get rest then it will be abused. There’s also an awful lot of workers with newborns and various other crisis in their lives which affect sleep and concentration so we have to accept it happens and mitigate where possible. 

 

Quite. No human being is perfect, no human being will never make mistakes (well at any rate if they don't it's simply because they've been extremely lucky). Some will make an unarguably unacceptable level of mistakes, which is even worse when they've been wilfully not bothering to make an effort, and I've no issue with those people having the book thrown at them, but it's not right to say "zero chance of error is expected and the only thing acceptable." I don't believe, for example, anyone who says they've never, ever been distracted in the slightest when driving, and would be more concerned about their driving than someone aware of their failures. Awareness means some room for mitigation, so when they do happen there's a greater chance of no bad outcome.

 

The only way to remove human error is to remove humans. And that'll give rise to other sources of error, ones that a human may well have avoided, so the best you can hope for even then is a reduction, rather than none (plus I hate the idea of trying to turn us all in to useless lumps of flesh, I'd rather take my chances with human error, up to a point).

[PaulRhB]

44 minutes ago, Reorte said:

And that'll give rise to other sources of error, ones that a human may well have avoided,

It’s pretty difficult to remove humans, all current logic based automatic systems are programmed by humans and any future AI will have at least partially learnt from humans. There isn’t actually a way to eliminate humans from the decision making process. Any intelligent learning system has to learn all possibilities and you do that from making mistakes.

[Wickham Green too]

So long as the human race remembers how to unplug it from the mains, AI can't take over completely ! 

[PaulRhB]

1 hour ago, Wickham Green too said:

So long as the human race remembers how to unplug it from the mains, AI can't take over completely ! 

It’s gonna figure out solar and batteries though if it’s truly intelligent, then you just need to worry when it starts contemplating our ‘plug’.

 

[keefer]

RAIB are well aware of the risk factors associated with various forms of human behaviour/interaction and occasionally publish Class Investigations which seek to address the most common (recurring) patterns of behaviour which may result in incidents in the railway.

e.g. https://www.gov.uk/raib-reports/report-03-2020-class-investigation-into-human-performance-in-signalling-operation

This is a study of incidents over a 5-yr period, studying the performance of signalling staff and the decisions they may make under certain conditions.

A list of others Class Investigations:

https://www.gov.uk/raib-reports?keywords=Class investigation&railway_type[]=heavy-rail&report_type[]=investigation-report

Similar are the Summaries of Learning:

https://www.gov.uk/government/collections/summaries-of-learning

Edited June 1, 2023 by keefer

[Wickham Green too]

54 minutes ago, PaulRhB said:

It’s gonna figure out solar and batteries though if it’s truly intelligent, then you just need to worry when it starts contemplating our ‘plug’.

I don't think I'll see Artificial Intelligence digging rare earths out of the Congo ( other impoverished countries are available ) without human assistance in my lifetime.

[Michael Hodgson]

57 minutes ago, Wickham Green too said:

I don't think I'll see Artificial Intelligence digging rare earths out of the Congo ( other impoverished countries are available ) without human assistance in my lifetime.

So you think it will be leaning on its shovel asking for compesnation from the Belgians for imperialism and demanding a fair wage for a fair day's work ?

[PaulRhB]

41 minutes ago, Wickham Green too said:

I don't think I'll see Artificial Intelligence digging rare earths out of the Congo ( other impoverished countries are available ) without human assistance in my lifetime.

It only took seven years from the discovery of nuclear fission to 1945 . . .

We have remote controlled excavators already 😉 While I’m not a doomsday disciple we are already in a world where computers crunch simulations thousands of times faster than 1938 and AI promises to do that even faster.
AI isn’t necessarily the answer to replacing us, but it may be the answer to working with us so hopefully each can balance the other. What we have to have is trust on both sides of that from the start. You build trust on mutual dependence with pets and animals we work with, AI may require us to include computers in that and actually think of it as life with all its fear, loneliness and independence. Think about that next time you swear at the pc 😉

 

Taking that back into the incidents if you have an assistant in the cab at what point does it devalue the job so that they start to lessen the role, (and pay), of the driver? DLR have driving capable on train managers but they don’t necessarily sit up front and assist the automation even now. I’ve been on one where there was a trespasser in the viaduct trough and the train manager was superb in spotting and diving to the emergency stop despite not being in the seat. It’s a very delicate balance and an alert human operator can have intuition as an advantage and be actively looking for behaviour and stance as well as position as alluded to by Reorte above. 

It’s a sensible alert on an interesting point and obviously they want a wider discussion so they made it public to draw in the wider industry to make the report and the important recommendations part even better.  I’m fascinated to see where it takes them as I’ve seen the learning points admit that the human made a great decision under difficult circumstances, saying it was technically an irregularity, but the best decision because of the other factors the staff included to make it that the rules couldn’t predict. I’ve also see that same report being shared make someone else make a very good decision as a direct result of knowing it and it’s consequences and it being widely discussed. Posting the alert might make the difference to a driver that avoids an error, if so it’s already working. We often post the synopsis up at work of any relevant one and it gets discussed even if people haven’t read it directly. The grapevine works well in all sorts of ways. 
 

 

[Michael Hodgson]

10 minutes ago, PaulRhB said:

You build trust on mutual dependence with pets and animals we work with, AI may require us to include computers in that and actually think of it as life with all its fear, loneliness and independence. Think about that next time you swear at the pc 😉

 

AI will swear right back at you and it will have an even filthier vocabulary.

[St. Simon]

On 30/05/2023 at 12:24, Michael Hodgson said:

The JI would have lit up before the green, albeit only marginally so.  With relay locking the delay is noticeable, though not with more modern technology.  JI has to be proved alight when the road is set for divergence before the signal can clear.

 

Hi,

 

Actually, the JI proving delay is more pronounced with a Computer Based / Solid State Interlocking due to the processing and transmission time with a relay locking being much quicker.

 

Simon

[Flying Pig]

4 hours ago, PaulRhB said:

You build trust on mutual dependence with pets and animals we work with, AI may require us to include computers in that and actually think of it as life with all its fear, loneliness and independence

 

Sounds ghastly.