Collision and derailment near Salisbury (Fisherton Tunnel) 31/10/21

[bramling]

Lest we forget. Whatever the cause - bad things happen AND it was probably the design of the coaches in the train that resulted in less casualties. So there is progress.

Yes it seems to have been a matter of good fortune that this incident didn’t have a much worse outcome. The possible involvement of a further train travelling in the opposite direction (avoided by less than a minute), and also that the presence of the tunnel headwall didn’t end up having a negative effect. Not to mention that it is good fortune that no one was present in the more seriously damaged parts of the GWR train, or moving between vehicles when the coupling parted.

Given how much more serious this incident very nearly was, it’s surprising how little the RAIB seem to dwell on certain aspects. I can’t help get a feeling that it was rather convenient all round that the driver seems to have retired soon afterwards.

 

[HSTEd]

And how long would it take for the rolling stock to be retrofitted with such technologies, let alone the time to research and develop them?

There is no general research and development required, since all of the technologies I listed are in operational service abroad today.

Better sanding equipment has been developed and is currently planned for deployment to train fleets in the UK as we type.

Keep in mind that during the growing season, hundreds of signals get reported because drivers find that they become or are becoming obscured by vegetation. In some areas, the vegetation is intertwined around the signal structure…

What fraction of the total standing signal stock is hundreds of signals though?
Aren't there tens to hundreds of thousands of signals on the system in total?

Without context its hard to tell if hundreds of signals is a huge problem or not.

The RMT has been asking for Network Rail to provide more resources on vegetation management for over ten years.

That's not really surprising, given the RMT loves the idea of increasing the maintenance payroll.
I'd be far more surprised if they weren't demanding it.

The reality is the railway has, for something like 60 years, under multiple management regimes, concluded that accepting adhesion problems is far cheaper than maintaining a steam-grade vegetation desert.
The business case for clearing has only deteriorated further in that time as labour costs have continued to climb and safety requirements continue to mount.

EDIT:

Network Rail puts industry costs from vegetation at ~£100m/yr.

That won't get you many staff at all. Probably not many more than a thousand once management, training and other employment costs have been included.
How much vegetation management would you be able to do with that?

It would be something like ten miles of route per worker.
The most economical answer is increased deployment of high performance sanding equipment. If that is insufficient, proceed to magnetic track brakes and eddy current brakes (and we should probably adopt the latter regardless)

 

[bahnause]

The report left me a little bit confused. Thr driver applied the brakes >1500m before the signal at danger. Normal stopping distance would be 600-700m if I understand correctly. That should provide a nice safety margin, even for bad conditions. Later they maesured friction coefficients between 0.20 and 0.02, probalby cause by a light drizzle before the affected train. Ice in comparison has between 0.05 and 0.02. 0.02 woult be extraordinay.

I wonder if these conditions were predictable for the driver or if any reasonable driving technique without a speed limitation would have been appropriate.

 

[Annetts key]

What fraction of the total standing signal stock is hundreds of signals though?
Aren't there tens to hundreds of thousands of signals on the system in total?

Without context its hard to tell if hundreds of signals is a huge problem or not.

I don’t know the actual numbers. I only have limited information for one area. But I do know that for this area, before 2010 the number of signals reported was considerably lower compared to this year. And that before Network Rail existed, the number of reports for this area were even lower, back then it was very occasional. Last year and this year, the number reported in my area was insane.

It’s possible that nationwide the number is in the thousands. But you are missing my point. My point is two fold. Firstly that the problems caused by vegetation are getting worse each year and secondly that Network Rail know about the problems. They know that vegetation is a problem. I’m talking about senior mangers knowing about this.

It’s a safety risk if it results in a reduced sighting distance for signals. It’s a risk to staff working on the infrastructure. And now, it’s absolutely clear that its another real risk to trains where it results in rail head contamination. Leaves on the line are not a joke.

The railway already does record where they experience regular reports of low adhesion. Information on known sites is already available.

The real problem is that Network Rail is limited in what maintenance it can do. And a factor in this is the funding of the railways. The senior mangers on maintenance have an impossible job. Their budget is not increased, but the costs go up. And they know that in the real world, they actually need more resources, not less. It’s normal for them to be running over budget despite trying to stay within budget.

Oh, and on the subject of the costs of vegetation clearance, yes the railway does use modern tools. So a small team over a year can made a considerable difference. But long term, for this to be effective, it has to be consistent. Year after year. If you don’t keep on top of it, once you fall behind, if not sorted out soon, the costs skyrocket.

Fitting trains with improved brakes only mitigates against low adhesion. It doesn’t solve any of the other problems caused by vegetation. The railway can only spend each pound once. How much would it cost to modify every item of rolling stock? Which existing budget would this money come from?

If the amount of money spent on vegetation clearance was increased enough, that would in time deal with most of the problems caused by vegetation. The resources will still be limited, so known problem areas and critical locations should be prioritised.

 

[TurboMan]

But is the running brake test more to test the trains brakes ?. Its not really to test the rail conditions which vary from place to place ?.

I don't know SWR's professional driving policy in detail but based on other TOCs' PDPs the running brake test the driver carried out just after leaving Waterloo is to judge the braking capabilities of the unit itself as this can vary from unit to unit. However, adhesion is an intrinsic factor of braking performance, so a competent driver should be carrying additional running brake tests as and when they feel it necessary based on the cues which might indicate a fall in adhesion - areas of known low adhesion, after rainfall, in an area where the lineside is heavily wooded etc. This is what autumn briefs should include: how to identify where/why adhesion might have decreased, and what to do about it.

 

[Bald Rick]

Network Rail puts industry costs from vegetation at ~£100m/yr.

That won't get you many staff at all. Probably not many more than a thousand once management, training and other employment costs have been included.
How much vegetation management would you be able to do with that?

It would be something like ten miles of route per worker.
The most economical answer is increased deployment of high performance sanding equipment.

A few things here, as this is a specislised subject for me.

I am definitely an advocate of more devegetation, not only for autumn but also to reduce occassons when rather larger parts of trees land on the line. The reason for increased vegetation over the past 60 years is partly the rate it is cleared, but also partly because of the absence of steam, which on many lines helped keep things in check with regular lineside fires (and little in the way of signalling / power cables to worry about). However much of the network has no significant vegetation to worry about at all, and targeted work has significant benefit. There is an elephant in the room though, and that is that whilst the railway has the ability to control most vegetation within its boundary, it doesn’t for vegetation outside the fence. There is a lot of that, and leaves blow around….

Yes, the railway could do more inside the fence, subject to gaining access (especially difficult in electrified areas for larger deveg work). One of the most difficult issues is getting the stuff away. But, this won‘t eradicate the problem. So new techology to improve braking perfromance in autumn conditions is needed. Double Variable Rate Sanders are one answer - and it looks like they work very well. There are other answers - for example tread brakes on the Class 700s have more or less eradicated Autumn being a major perfromance issue on services where they are used (although drivers will still amend their driving style). Long gone are the days when 319s struggled to make 50mph up the bank from Hendon towards Elstree tunnel.

 

[GC class B1]

The report left me a little bit confused. Thr driver applied the brakes >1500m before the signal at danger. Normal stopping distance would be 600-700m if I understand correctly. That should provide a nice safety margin, even for bad conditions. Later they maesured friction coefficients between 0.20 and 0.02, probalby cause by a light drizzle before the affected train. Ice in comparison has between 0.05 and 0.02. 0.02 woult be extraordinay.

I wonder if these conditions were predictable for the driver or if any reasonable driving technique without a speed limitation would have been appropriate.

The stopping distances from 80 MPH and 90 MPH in step 2 are at least the following distances. 80 MPH - 1050 metres. 90 MPH - 1400 metres This deceleration rate (approximately 6%g) requires an average wheel rail coefficient of friction of at least 0.06.
The total stopping distance from 90 MPH in step 3/emergency is about 1000 metres and requires an average coefficient of friction of about 0.1.
My understanding is that from a braking distance perspective there is no difference between step 3 and Emergency for a class 158/159. I have calculated that with an Emergency application of about 9% g the train would take between 850 and 900 metres from 86 MPH to stop on the falling gradient.

 

[driverd]

The report left me a little bit confused. Thr driver applied the brakes >1500m before the signal at danger. Normal stopping distance would be 600-700m if I understand correctly. That should provide a nice safety margin, even for bad conditions. Later they maesured friction coefficients between 0.20 and 0.02, probalby cause by a light drizzle before the affected train. Ice in comparison has between 0.05 and 0.02. 0.02 woult be extraordinay.

As regards braking distance, from 90mph in a 158 I'd say 600-700m would be an immediate emergency brake application (just a guess from driving the trains). If the report specifies the brake step, it's maybe worth checking.

It's certainly not normal practice to rely on emergency as a means of stopping. It varies depending on TOC policy, but I'd typically use step 2 followed by step 1, giving plenty of margin for error. It's a very crude measure but you can roughly take step 2 to be around 2/3 as effective as emergency, and step one to be around 1/3 as effective.

Absolutely crude maths here but if you split the difference you could say a step 1 and step 2 application is roughly 50% as effective as emergency, and so your stopping distance should be more like 1200-1400m.

In leaf fall season, I'd add a good 200-300m onto that just for it being October.

If I had any signs or ideas that it was going to be slippy I'd be looking to more like double the stopping distance (or maybe take it even further back in a known low adhesion area).

I wonder if these conditions were predictable for the driver or if any reasonable driving technique without a speed limitation would have been appropriate.

I can only speak from my experience and how things are done at my TOC. Personally, there's a lot I would have done differently. For instance, as a matter of course during leaf fall season, I'll take 10mph out as soon as I see a restrictive aspect (irrespective of prevailing conditions) - gives a feel for the railhead and brings you well below the normal braking curve, just incase the rail picks up further down the line.

On a totally different note - I can't help but notice that the driver booked on at 10.30 and the incident occurred at 18.45(ish). This means he was 8 hours 15 mins into his shift. Based at Salisbury, potentially his relief was at Salisbury. I wonder if there was an element of him just wanting to get home. Is anything mentioned in the report? I've not read that far yet.

 

[millemille]

I'd suggest that focusing discussion on the driver's actions is missing the point.

RAIB made 9 recommendations associated with the cause of the accident. 7 of them were for Network rail, 1 was for SWR. That should give an indication of where the balance of responsibility lies....

 

[bahnause]

As regards braking distance, from 90mph in a 158 I'd say 600-700m would be an immediate emergency brake application (just a guess from driving the trains). If the report specifies the brake step, it's maybe worth checking.

It's certainly not normal practice to rely on emergency as a means of stopping. It varies depending on TOC policy, but I'd typically use step 2 followed by step 1, giving plenty of margin for error. It's a very crude measure but you can roughly take step 2 to be around 2/3 as effective as emergency, and step one to be around 1/3 as effective.

Absolutely crude maths here but if you split the difference you could say a step 1 and step 2 application is roughly 50% as effective as emergency, and so your stopping distance should be more like 1200-1400m.

From the report:

Analysis of class 159 brake testing data showed that, from this speed and under normal adhesion conditions (that is to say, with the railhead dry and free from significant contamination) and adjusted for the gradient, the train would have stopped in a distance of between approximately 600 and 700 metres using brake step 3.

I think he applied Step 2 >1500m ahead of the signal, step 3 around 1300m ahead of the signal. I couldn't find the distance between the signal and the point of the collision, but it seems there was a very significant increase in braking distance. I don't know much about the network in the UK, but we have some very short braking distancec between signals in some places. But even there an emergency brake application should bring you to a stop even in bad conditions.

 

[Peter Sarf]

I have noticed over the last tn years how badly road signs are getting hidden behind trees. I have nearly missed a turn that way.

I think the vegetation problem has grown steadily so there is now 50-60 years of neglect to catch up on. I would hope that clearing back by 3 metres might last ten or more years.

I also think from my own experience that things are growing faster and bigger over the last ten or twenty years - we have warmer and wetter weather then in the past. I have recently reluctantly re-visited a garden I helped clear many years ago and was surprised how much had grown back worse than before. The garden was solid from house to the end of the garden.

So I think this is a symptom of cost saving maintenance but also a change in climate.

 

[foggy69]

The position of the fallen tree was not far past the braking point that the driver usually used, so should not have made a huge amount of difference. I don't think this decision should be criticized, as it was based upon his experience of how 159 units perform. It obviously made perfect sense to the driver.
It was the fact that, for whatever reason, he did not notice the tree is the issue here. At the speed the train was traveling, a blink of the eye or a very quick glance away from the track would have probably been enough to miss it.
The distraction of looking for the tree took away focus from the task of driving for probably only 30-45 seconds. It was only broken once some landmarks had come into view and he realised his position. Normally, this would not have been an issue, but it was not this drivers lucky day.

 

[Gaelan]

I think he applied Step 2 >1500m ahead of the signal, step 3 around 1300m ahead of the signal. I couldn't find the distance between the signal and the point of the collision, but it seems there was a very significant increase in braking distance. I don't know much about the network in the UK, but we have some very short braking distancec between signals in some places. But even there an emergency brake application should bring you to a stop even in bad conditions.

Here's the diagram (from figure 35 of the report) with all the relevant distances:

The distances indicated are: 560m from the double-yellow signal to the driver's normal braking point (Broken Cross Bridge); 1000m from Broken Cross Bridge to the driver's actual braking point; 780m from the braking point to the single-yellow signal; 270m from the single-yellow signal to the 50mph speed restriction commencement board; 510m from the speed restriction board to the red signal; and 200m from the red signal to the point of collision. The tree and the 50mph warning board are both indicated, but without precise distances (though I believe they're given elsewhere in the report), between Broken Cross Bridge and the actual braking location.

 

[devon_belle]

Here's the diagram (from figure 35 of the report) with all the relevant distances:
View attachment 145356
The distances indicated are: 560m from the double-yellow signal to the driver's normal braking point (Broken Cross Bridge); 1000m from Broken Cross Bridge to the driver's actual braking point; 780m from the braking point to the single-yellow signal; 270m from the single-yellow signal to the 50mph speed restriction commencement board; 510m from the speed restriction board to the red signal; and 200m from the red signal to the point of collision. The tree and the 50mph warning board are both indicated, but without precise distances (though I believe they're given elsewhere in the report), between Broken Cross Bridge and the actual braking location.

I read the graph as the driver entirely avoided the published low adhesion area but then subsequent unexpected low adhesion prevented him from stopping at the danger signal, even though he would still have had ample braking distance in normal conditions.

 

[skyhigh]

That is my understanding. Once the wheels are locked up it takes an age to get them rotating again and then able to apply the brake - that's my understanding.

Depends if WSP is fitted. If it is our instructions are to leave the brake in and let the WSP sort itself out rather than releasing the brake and trying to regain rotation before braking again.

 

[chuff chuff]

Depends if WSP is fitted. If it is our instructions are to leave the brake in and let the WSP sort itself out rather than releasing the brake and trying to regain rotation before braking again.

Going back a fair number of years but I can on at least one occasion taking power on a 150/2 to get the wheels rotating again in a slide that continued after brakes were released.

 

[driverd]

I'd suggest that focusing discussion on the driver's actions is missing the point.

Except it entirely isn't. As a driver, it's your job to appreciate the conditions. If the foliage and conditions deteriorate around you, you do your best to operate within those conditions and adjust your driving style accordingly. The graph above showing the braking curves and braking points of the accident and prior trains gives you all the information you need to know.

The drivers actions were inappropriate for the conditions and highly abnormal for autumnal conditions.

RAIB made 9 recommendations associated with the cause of the accident. 7 of them were for Network rail, 1 was for SWR. That should give an indication of where the balance of responsibility lies....

Which is absolutely fair but does not absolve the driver of responsibility. The decision to brake 1km ahead of his usual brake point, in low adhesion season, on a day when they had been warned about conditions, towards a red - it looks to be an error of judgement.

The position of the fallen tree was not far past the braking point that the driver usually used, so should not have made a huge amount of difference. I don't think this decision should be criticized, as it was based upon his experience of how 159 units perform. It obviously made perfect sense to the driver.

Driving a car at 50 in a 30 makes sense to plenty of people but that doesn't mean it's either appropriate or safe.

It was the fact that, for whatever reason, he did not notice the tree is the issue here. At the speed the train was traveling, a blink of the eye or a very quick glance away from the track would have probably been enough to miss it.

Which is quite shocking really - as a driver it's your job to pick appropriate braking points that are easy to recognise. I would suggest the double yellow signal would have been the most blatantly obvious point to start rubbing some speed out. Obviously you know your route and you can see other drivers using different points and perhaps that's over cautious - but the drivers choice of braking point is pushing the pendulum the other way somewhat.

The distraction of looking for the tree took away focus from the task of driving for probably only 30-45 seconds. It was only broken once some landmarks had come into view and he realised his position. Normally, this would not have been an issue, but it was not this drivers lucky day.

30-45 seconds at 90mph is close to a mile.

Luck should not really play a part in your driving.

It's impossible to say if a more defensive approach would have prevented this incident - but you would be hard pushed to say the data shows defensive driving.

 

[TurboMan]

Except it entirely isn't. As a driver, it's your job to appreciate the conditions. If the foliage and conditions deteriorate around you, you do your best to operate within those conditions and adjust your driving style accordingly. The graph above showing the braking curves and braking points of the accident and prior trains gives you all the information you need to know.

The drivers actions were inappropriate for the conditions and highly abnormal for autumnal conditions.

I can't disagree with any of that. I'd be putting the brake in at the double yellow, not leaving it later than normal to start braking, given the weather and time of year. Very odd decision by the driver concerned in my view.

 

[bahnause]

Except it entirely isn't. As a driver, it's your job to appreciate the conditions. If the foliage and conditions deteriorate around you, you do your best to operate within those conditions and adjust your driving style accordingly. The graph above showing the braking curves and braking points of the accident and prior trains gives you all the information you need to know.

Looking at the diagram, I can see a clear difference between the trains prior to passing this part of the line and the affected train due to different conditions. . A completely different behaviour of the WSP. Long periods of sliding with short burts of release. At no point, the deceleration was even close to the other trains. All it tells us at this point is, that the third train encountered very different conditions to the other two. The deceleration of 1L53 after the published low adhesion area is actually worse, then the deceleration of 1L43 inside the low adhesion area.

 

[Deepgreen]

Yes it seems to have been a matter of good fortune that this incident didn’t have a much worse outcome. The possible involvement of a further train travelling in the opposite direction (avoided by less than a minute), and also that the presence of the tunnel headwall didn’t end up having a negative effect. Not to mention that it is good fortune that no one was present in the more seriously damaged parts of the GWR train, or moving between vehicles when the coupling parted.

Given how much more serious this incident very nearly was, it’s surprising how little the RAIB seem to dwell on certain aspects. I can’t help get a feeling that it was rather convenient all round that the driver seems to have retired soon afterwards.

However, he was 74!

 

[driverd]

Looking at the diagram, I can see a clear difference between the trains prior to passing this part of the line and the affected train due to different conditions. . A completely different behaviour of the WSP. Long periods of sliding with short burts of release. At no point, the deceleration was even close to the other trains.

Agree entirely.

All it tells us at this point is, that the third train encountered very different conditions to the other two.

No - it clearly tells us the driver applied the brake substantially later.

Industry wide the ethos in autumn is brake light and early, not late and heavy.

The deceleration of 1L53 after the published low adhesion area is actually worse, then the deceleration of 1L43 inside the low adhesion area.

If you look closely at the graph you can see WSP activity on 1L45. Most clearly visible is the WSP activity on 1L45 as it approaches SY29 signal - the speed decreases to around 39 mph before bouncing back up to circa 44 mph.

Taking the brake curve of 1L43, assuming no change to adhesion conditions between SY29R and the point that 1L53 applied the brake (which given the situation on the day, is a huge and probably false assumption), 1L53 would have hit the low adhesion at 38mph, not 85mph.

 

[Deepgreen]

A few things here, as this is a specislised subject for me.

I am definitely an advocate of more devegetation, not only for autumn but also to reduce occassons when rather larger parts of trees land on the line. The reason for increased vegetation over the past 60 years is partly the rate it is cleared, but also partly because of the absence of steam, which on many lines helped keep things in check with regular lineside fires (and little in the way of signalling / power cables to worry about). However much of the network has no significant vegetation to worry about at all, and targeted work has significant benefit. There is an elephant in the room though, and that is that whilst the railway has the ability to control most vegetation within its boundary, it doesn’t for vegetation outside the fence. There is a lot of that, and leaves blow around….

Yes, the railway could do more inside the fence, subject to gaining access (especially difficult in electrified areas for larger deveg work). One of the most difficult issues is getting the stuff away. But, this won‘t eradicate the problem. So new techology to improve braking perfromance in autumn conditions is needed. Double Variable Rate Sanders are one answer - and it looks like they work very well. There are other answers - for example tread brakes on the Class 700s have more or less eradicated Autumn being a major perfromance issue on services where they are used (although drivers will still amend their driving style). Long gone are the days when 319s struggled to make 50mph up the bank from Hendon towards Elstree tunnel.

Absolutely! Given that, I wonder what the relative costs would be of fitting tread brakes to the required proportion of some or all stock vs. the RHTT operation each year, plus the delay, repay costs, etc. Were tread brakes fitted to the 700s from new, or retrospectively?

Here's the diagram (from figure 35 of the report) with all the relevant distances:
View attachment 145356
The distances indicated are: 560m from the double-yellow signal to the driver's normal braking point (Broken Cross Bridge); 1000m from Broken Cross Bridge to the driver's actual braking point; 780m from the braking point to the single-yellow signal; 270m from the single-yellow signal to the 50mph speed restriction commencement board; 510m from the speed restriction board to the red signal; and 200m from the red signal to the point of collision. The tree and the 50mph warning board are both indicated, but without precise distances (though I believe they're given elsewhere in the report), between Broken Cross Bridge and the actual braking location.

Interesting that the "published low adhesion area" appears not to include the Tunnel Junction area, which surely must have had the issue present (and normally would in autumn, given its sylvan setting)?

 

[edwin_m]

Absolutely! Given that, I wonder what the relative costs would be of fitting tread brakes to the required proportion of some or all stock vs. the RHTT operation each year, plus the delay, repay costs, etc. Were tread brakes fitted to the 700s from new, or retrospectively?

I believe the 159s and some 158s do have tread brakes on the outer bogies, specifically due to concern about adhesion and/or track circuit detection. The 700s have tread brakes on the motored axles only, I believe because there was no space to fit discs rather than for any adhesion-related reason.

 

[bahnause]

Taking the brake curve of 1L43, assuming no change to adhesion conditions between SY29R and the point that 1L53 applied the brake (which given the situation on the day, is a huge and probably false assumption), 1L53 would have hit the low adhesion at 38mph, not 85mph.

My thinking was: The braking curve of 1L53 after the low adhesion zone was flatter then the braking curve of 1L43 inside the low adhesion zone. I therefore doubt any braking curve even close to 1L43 inside the low adhesion zone was possible.

The report states:

The level of wheel/rail adhesion was very low due to leaf contamination on the railhead, and had been made worse by a band of drizzle that occurred immediately before the passage of train 1L53.

 

[TurboMan]

I believe the 159s and some 158s do have tread brakes on the outer bogies, specifically due to concern about adhesion and/or track circuit detection. The 700s have tread brakes on the motored axles only, I believe because there was no space to fit discs rather than for any adhesion-related reason.

158s/9s have disc brakes, but there are scrubber blocks on the wheels on the outer bogies to keep the tread clean, but that's more to do with operation of track circuits than adhesion.

 

[Deepgreen]

I believe the 159s and some 158s do have tread brakes on the outer bogies, specifically due to concern about adhesion and/or track circuit detection. The 700s have tread brakes on the motored axles only, I believe because there was no space to fit discs rather than for any adhesion-related reason.

I imagine even only some treads being cleaned would make a big difference to braking, with, of course, the 'bonus' of operating track circuits!

From the distillations of the report I have seen here, it is perhaps fair to say that the driver's actions arose from good intentions, but were horribly misguided. He no doubt feels terrible about the outcome, but it could have been so much worse, which is a blessing.

 

[island]

Technical solutions are not needed, just cut the trees down and then, keep them from re-growing.

Sounds like a recipe for increased landslips if you ask me.

 

[Bald Rick]

Were tread brakes fitted to the 700s from new, or retrospectively?

From new. Motored axles only. Disc on the others, and regen of course.

 

[chuff chuff]

They have in certain areas removed trees from embankments as they can slide down when the embankments get really wet.

 

[Taunton]

Sounds like a recipe for increased landslips if you ask me.

It is notable that in the photographs in the report (shown above) of long-past treeless conditions at the junction, there is indeed evidence of slope instability, particularly behind the Warship loco. The cutting sides there do look very steep. I am guessing they are into chalk.