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

[John R]

But drivers do second guess the signals. That is exactly what happened at Colwich,

Colwich was 35 years ago (and has already been pointed out the cause was not actually caused by second guessing the signal). I would imagine driver training has moved on a bit since then - indeed it's well documented that more defensive driving is encouraged these days.

As for the second example, I'd suggest that is a completely different kettle of fish, as there is absolutely no risk (and indeed with slam door stock, it could be said to reduce risk for the reason given). In more recent years I used to see drivers often crawl at walking pace towards Bedminster station having seen a faster train overtake theirs knowing that the signal is expected to clear in due course. By doing so they avoided having to jump out and call the box, thus delaying the service further, and again there is absolutely no risk.

 

[Nottingham59]

Thank you all for the helpful replies to my earlier questions about signalling.

 

[The Ham]

Of course, this is something air traffic control cracked in its earliest times, telling aircraft pilots by radio "aircraft xxx, minimum approach speed, one to depart ahead", and such like. It doesn't replace fixed instructions, but just adds to them.

Indeed, such instructions may not be there only to allow a signal to clear. For instance it could be to reduce speeds as there's been reports of reduced breaking conditions or (depending on the setup) to keep trains close to the timetable whilst not using more energy than is needed.

 

[Annetts key]

X Jcn ———— SY31 ———— SY29 ————— SY29R
<—> <——183m—-><—-787m—-><—-2336m—->
<—> <-200yds-><—860yds-><—2554yds—>
<————191m—-——>
<———————————978m———————>
<———————————————3314m———————————-—>
————————————— Y ——————- G ——————- G
————————————- R ——————- Y ——————- YY
——————————————————————- R ——————- Y
The distances in the first line are estimated and assume that SY31 is 183m away from the junction (standard overlap distance). X marks the point of the estimated point of collision.

At a line speed of 90MPH, it appears there is insufficient (service) braking distance (for the worse case brakes train type) between SY31 and SY29. That’s why SA29R has been provided with a double yellow aspect. It’s likely this was done to keep the cost of the scheme at a reasonable cost. Otherwise either SY29 would need moving further away from SY31 (not preferred) or an additional SA27 signal would need to be provided between SA29 and the current SA29R, with SA29R being removed and another signal, SA27R being provided to act as a distant to SA27.

At least, that’s my take on it.

 

[bengley]

My take on as a driver is that the driver braked appropriately for the speed and distance. The leaf fall contamination was clearly severe and unexpected, even for the time of year. It's clear that the RHTT needed to run more regularly

 

[Nicholas Lewis]

One of the questions that occurs to me, to which I haven't seen an answer, is when was the previous time a train had to stop at signal SY31? or received a caution (yellow or double yellow) at SY29R? Had it been earlier that day? How had it got on? (Perhaps the specific question is - had a driver had a caution at SY29R, and breathed a sigh of relief at finding SY31 clear?)

Thats a good point and perhaps under this further RAIB objective for the investigation

Continue to establish, as far as practical, the level of wheel/rail adhesion present on the down main line on approach to Salisbury Tunnel Junction

they will have looked at previous movements which would have been recorded by data loggers if they are fitted to the signalling system as well as getting OTMDRs from previous trains.

Have to say hadn't appreciated that the collision speed was so high shows the resilience that BR bought to its carriage design over 40 years ago.

 

[bramling]

My take on as a driver is that the driver braked appropriately for the speed and distance. The leaf fall contamination was clearly severe and unexpected, even for the time of year. It's clear that the RHTT needed to run more regularly

I just can't quite shake off the feeling that braking was left late.

It was a foul night by all accounts, smack bang in the middle of the leaf-fall season, a line with an infrequent service, falling gradient, and a short train. To me there's quite a lot of potential alarm bells there, and I must say I was *very* surprised to read that the driver didn't at least test the water with a light brake application as soon as the double yellow on SY29R was sighted.

Yes the adhesion was particularly bad and the railhead hadn't been treated, but equally hazard anticipation is a part of driving any vehicle, and I can't help but find it a little surprising that the sequence of actions outlined by the RAIB simply doesn't seem to indicate any anticipation. One can only surmise, especially taking into consideration the driver's clearly very extensive experience, is that this location wasn't a location known for issues, but given the characteristics I can't help but find that surprising in itself.

 

[Nicholas Lewis]

I just can't quite shake off the feeling that braking was left late.

It was a foul night by all accounts, smack bang in the middle of the leaf-fall season, a line with an infrequent service, falling gradient, and a short train. To me there's quite a lot of potential alarm bells there, and I must say I was *very* surprised to read that the driver didn't at least test the water with a light brake application as soon as the double yellow on SY29R was sighted.

Yes the adhesion was particularly bad and the railhead hadn't been treated, but equally hazard anticipation is a part of driving any vehicle, and I can't help but find it a little surprising that the sequence of actions outlined by the RAIB simply doesn't seem to indicate any anticipation. One can only surmise, especially taking into consideration the driver's clearly very extensive experience, is that this location wasn't a location known for issues, but given the characteristics I can't help but find that surprising in itself.

Guess it's quite possible to be come conditioned to not braking at a YY at SY29R given its significantly overbraked and the line speed doesn't change for some distance either. So in the absence of any advice to the driver over railhead conditions which is probably as previous trains haven't reported anything no external trigger to change behaviour.

Be interesting to see whether this jcn should be double blocked under autumn precautions.

 

[Grumpy Git]

I dont want to gar OT as there is already a thread on Network vegetation but-----

1) RAIB report - low adhesion - leaves - trees by line bad

2) The latest RAIL magazine has a great article published anonymously by a FOC driver. Trees by line bad

3) Latest bad weather where trees disrupt everything badly --- Trees by line bad.

I really hope and pray that attention is finally paid.

Yes, but according to some on here it's not all due to the trees beside the line!

Some folk need to wake up. The biggest force on leaves is gravity (not wind)!

I know people have found all sorts of unknown detritus in their gardens this week, but autumn comes every year and trees don't move.

 

[Nottingham59]

My take on as a driver is that the driver braked appropriately for the speed and distance.

I agree. The RAIB report says that "the driver made a step 2 brake application. This was in accordance with his usual practice and was done with the intention of being able to stop the train at signal SY31, which was at this point around 1,500 metres away."

I just can't quite shake off the feeling that braking was left late.

I agree with that too! Perhaps it would be more constructive to ask should the braking on that particular day have been earlier? I think we need to wait to see what the final report has to say.

I suspect it's going to be one of those cases where all the holes in the cheese lined up - the RHTT didn't run; that fact wasn't communicated to the driver; other trains hadn't had to stop at that particular signal; the low adhesion hadn't been noticed; there is no approach control to this junction; signal SY29R is a long way before the usual braking point; there was no early Step 1 brake application to test conditions; etc. etc. etc.

 

[GB]

Guess it's quite possible to be come conditioned to not braking at a YY at SY29R given its significantly overbraked and the line speed doesn't change for some distance either. So in the absence of any advice to the driver over railhead conditions which is probably as previous trains haven't reported anything no external trigger to change behaviour.

Be interesting to see whether this jcn should be double blocked under autumn precautions.

What happened to defensive driving?

 

[theking]

All well and good saying the driver should have braked 10 miles away and predicted low adhesion that in these days is pretty much unprecedented.

But no interrogation or investigation of the signallers actions.

If they didn't route a train that was 18 late infront of the SWR then this all could have been avoided.

If the driver is supposed to be super human and know the exact conditions of the rail and expect to slide over a mile and only lose 30 mph then why is the signaller stopping trains at 90mph and routing late trains over a junction that's 20mph when the SWR line had a 50mph and would have went across and been past it with no danger.

 

[norbitonflyer]

why is the signaller stopping trains at 90mph and routing late trains over a junction that's 20mph when the SWR line had a 50mph and would have went across and been past it with no danger.

There were three trains approaching the junction. He chose to delay 1F53 so that both the others had a clear run.

 

[Domh245]

All well and good saying the driver should have braked 10 miles away and predicted low adhesion that in these days is pretty much unprecedented.

But no interrogation or investigation of the signallers actions.

If they didn't route a train that was 18 late infront of the SWR then this all could have been avoided.

..

38 Evidence indicates that the signalling system was operating as designed before the accident. Signal SY31 was at danger as train 1L53 approached it, and the routing of trains across Salisbury Tunnel Junction on the day of the accident was in accordance with normal signalling practice.

Looks like a fairly sensible case of regulation - 3 trains, none of which are on-time, but routing 1F27 and 1F30 across the junction doesn't add delay minutes to either, where routing 1L53 through would have added delay to both. It's no doubt a factor, but signallers shouldn't be constrained in leaf-fall season by assuming that every train approaching from high speed will fail to stop

 

[GB]

All well and good saying the driver should have braked 10 miles away and predicted low adhesion that in these days is pretty much unprecedented.

But no interrogation or investigation of the signallers actions.

If they didn't route a train that was 18 late infront of the SWR then this all could have been avoided.

If the driver is supposed to be super human and know the exact conditions of the rail and expect to slide over a mile and only lose 30 mph then why is the signaller stopping trains at 90mph and routing late trains over a junction that's 20mph when the SWR line had a 50mph and would have went across and been past it with no danger.

No one suggested he should have braked 10 miles away, or that he should have predicted low adhesion or that he's super human and should know the exact conditions at xx point. Signallers aren't there to second guess whether trains will stop at signals or not. They set routes within the regulations on the express understanding that trains will stop otherwise the network would soon grind to a halt. Signallers often are not aware of geographical features or the speed at which the trains are actually travelling at a given point.

 

[theking]

..



Looks like a fairly sensible case of regulation - 3 trains, none of which are on-time, but routing 1F27 and 1F30 across the junction doesn't add delay minutes to either, where routing 1L53 through would have added delay to both. It's no doubt a factor, but signallers shouldn't be constrained in leaf-fall season by assuming that every train approaching from high speed will fail to stop

No but there should be safety mitigation in place

Stopping a train from 90mph so two can cross at 20mph is asking for trouble.

The GWR that was hit according to that report was already 18 late the SWR would have been past at 50mph adding 1 or 2 minutes at most.

Signallers should absolutely be aware that stopping a train in leaf fall season from 90mph cam result in something like this and make decisions accordingly

 

[O L Leigh]

Yes, but according to some on here it's not all due to the trees beside the line!

Some folk need to wake up. The biggest force on leaves is gravity (not wind)!

I know people have found all sorts of unknown detritus in their gardens this week, but autumn comes every year and trees don't move.

I can only assume that you must live in a very sheltered spot, because leaves really do move around a lot once they’re down. I have no trees overhanging my garden by alway seem to end up with leaves coming in from somewhere.

 

[Nicholas Lewis]

No but there should be safety mitigation in place

Stopping a train from 90mph so two can cross at 20mph is asking for trouble.

The GWR that was hit according to that report was already 18 late the SWR would have been past at 50mph adding 1 or 2 minutes at most.

Signallers should absolutely be aware that stopping a train in leaf fall season from 90mph cam result in something like this and make decisions accordingly

Its not asking for trouble you can question the regulating policy if you want but the signalling is there to be compiled with by drivers. Under normal adhesion conditions there is plenty of distance to stop the train but clearly not when railhead conditions are poor. The signaller can't second guess that but whether there are specific additional provisions like double blocking here we will have to wait for RAIB to issue final report.

 

[bengley]

I don't think the blame lies with any of the operational staff involved.

We can't mitigate against everything* and in this case even double blocking wouldn't have helped much because it seems the slide probably would have still taken the train through the junction (though, it could, of course, have delayed the SWR sufficiently to prevent this particular collision).

Even if the driver had tried the brake at the YY (which in my opinion is excessively early, some 3.3km from the red) there was no guarantee that his train's speed would have reduced enough to prevent a SPAD

* Apart from by perhaps running the RHTT...

 

[AlexNL]

Would it be useful or a distraction to have something like OTT maps visible in the cab so drivers can see the situation ahead in rather more detail than the signals alone can?

Dutch train drivers have something like that at their disposal, in a system called 'TimTim':

Picture by Driver Stefan on Twitter.

The 'ribbon' shows the path the train will take, with each block labelled. The driver who made the picture is awaiting the right time departure of his train (running under number 2621), and it shows him that late running train 56106 will be routed in front of his further ahead in section DGRW-72. It also shows him that train 1518 is behind him.

Drivers have this app on their Android tablet, which they can use while they're driving (the devices are otherwise locked down, you can't watch YouTube on them whilst moving). In newer trains, the TimTim functionality has been integrated into the cab.

 

[GC class B1]

My layman’s understanding is that as the driver would have known that the double yellow at SY29R meant that SY31 was at red he would be braking at the appropriate time to stop at this signal. Also he would be aware that SY31 was at the junction he would know that there could be a conflicting move signalled. Knowing from some other source that there was likely to be a conflicting move happening at the junction would not therefore have added to his understanding. Am I missing something?

 

[bramling]

I don't think the blame lies with any of the operational staff involved.

We can't mitigate against everything* and in this case even double blocking wouldn't have helped much because it seems the slide probably would have still taken the train through the junction (though, it could, of course, have delayed the SWR sufficiently to prevent this particular collision).

Even if the driver had tried the brake at the YY (which in my opinion is excessively early, some 3.3km from the red) there was no guarantee that his train's speed would have reduced enough to prevent a SPAD

* Apart from by perhaps running the RHTT...

I'm not saying it would have necessarily prevented it, just I find it surprising there seems to have been little if any defensive driving here, when there were a few "red flags" which one might have expected to prompt it.

 

[Need2]

My take on as a driver is that the driver braked appropriately for the speed and distance. The leaf fall contamination was clearly severe and unexpected, even for the time of year. It's clear that the RHTT needed to run more regularly

A common sense post at last!

 

[Watershed]

All well and good saying the driver should have braked 10 miles away and predicted low adhesion that in these days is pretty much unprecedented.

But no interrogation or investigation of the signallers actions.

If they didn't route a train that was 18 late infront of the SWR then this all could have been avoided.

If the driver is supposed to be super human and know the exact conditions of the rail and expect to slide over a mile and only lose 30 mph then why is the signaller stopping trains at 90mph and routing late trains over a junction that's 20mph when the SWR line had a 50mph and would have went across and been past it with no danger.

There is already a margin of safety afforded by the existence of the overlap. This, together with the advent of TPWS, better performing units, and increasing levels of training and professional standards means that incidents such as these are now exceptionally rare.

It signallers were instructed to act in the expectation that trains may pass signals at danger, it would cause delays all over the shop. It would be utterly disproportionate and completely unworkable, even if it did prevent a 1 in 10/20 year incident like this.

My layman’s understanding is that as the driver would have known that the double yellow at SY29R meant that SY31 was at red he would be braking at the appropriate time to stop at this signal. Also he would be aware that SY31 was at the junction he would know that there could be a conflicting move signalled. Knowing from some other source that there was likely to be a conflicting move happening at the junction would not therefore have added to his understanding. Am I missing something?

There is no way of knowing that a conflicting move was signalled - it could simply have been another train ahead, or an obstruction, or any one of a number of other reasons.

 

[O L Leigh]

I just can't quite shake off the feeling that braking was left late.

It was a foul night by all accounts, smack bang in the middle of the leaf-fall season, a line with an infrequent service, falling gradient, and a short train. To me there's quite a lot of potential alarm bells there, and I must say I was *very* surprised to read that the driver didn't at least test the water with a light brake application as soon as the double yellow on SY29R was sighted.

Yes the adhesion was particularly bad and the railhead hadn't been treated, but equally hazard anticipation is a part of driving any vehicle, and I can't help but find it a little surprising that the sequence of actions outlined by the RAIB simply doesn't seem to indicate any anticipation. One can only surmise, especially taking into consideration the driver's clearly very extensive experience, is that this location wasn't a location known for issues, but given the characteristics I can't help but find that surprising in itself.

As a driver, I am generally cautious about commenting on another driver's technique, even where the route and traction are familiar to me. Therefore I would feel unqualified to comment on another person's driving style where the route and/or traction is unfamiliar, and especially uncomfortable about trying to unpick the decision-making process that this particular driver went through on that specific day in those prevailing conditions.

It does strike me that it's easy to work through the chronology of this incident backwards and arrive at the conclusion that the driver should have applied the brake earlier. The problem with this approach is that it implies a degree of foresight that could only be described as superhuman. Had the driver known that he was going to lose control of the train I'm certain that he would have done as you expected, but the point is that he didn't know.

But if we are to say that the driver should have braked sooner, the question is how much sooner? My own feeling is that the braking technique used was probably about right. A little under a mile is plenty of space to stop a unit from 86mph using a Step 2 application, even on a modest falling gradient under indifferent railhead conditions. But given the exceptionally poor railhead conditions, how much extra room should the driver have allowed? The driver followed the correct technique and received assistance from the unit's WSP and (presumably) sanders, and yet still lost control of the train. Is it not possible that, no matter where the driver had applied the brake, control would still have been lost and a collision occurred? Bear in mind that the train travelled for more than a mile from the point where the brake was initially applied to where the two trains collided but it's speed was only reduced by 30-34mph.

As for carrying out a running brake test (RBT), I am unsure how much difference this would have made. RBTs are useful for getting a feel for a train's braking performance and can give an idea about rail conditions, but they can only do so at the time and place where you carry one out. The nature of railhead conditions being what they are, the results of a RBT do not necessarily tell you anything about what might happen further on when you have to brake for a station or signal. Poor rail conditions can be extremely localised, to the extent that all can be fine, then you get 20 feet of Torvill and Dean, and then it's back to being hunky-dory again. Perhaps a RBT would have been helpful in this situation, or maybe it would have been misleading or done at the wrong time to have made any difference and the collision still happened.

I understand what you say about hazard perception but, as with road users, it's hard to know precisely what the hazard is until you're neck deep in whatever situation you find yourself in. I would like to hope that I am sufficiently cautious in my driving whatever the conditions, but it is very hard to foresee the truly exceptional and to be prepared for it. The consequence of such preparedness would be greatly increased journey times as drivers creep around at greatly reduced speeds and braking incredibly early. I think it would be better to allow drivers to exercise their own judgement based on their experience and training as we do now, learn the lessons from this incident and then take things forward from there.

I would go back and remind folk that, in the immediate aftermath of the collision, SWR publicly went on record to praise their driver's exemplary conduct. Notwithstanding what the RAIB's final report might say on the matter, I do not think that SWR would have been so quick to have said as much had there been anything on the OTMR download or in the driver's debrief to suggest that he'd done anything other than precisely what his employer would have expected him to do. It may be that the advice we get for driving in low adhesion changes as a result of this incident, but as things stand at present it would appear that the driver followed the correct procedures as they are now.

Consequent to all of this, I do not think that now is the right time to question what the driver did or did not do, especially when we have neither the full facts, the expertise and knowledge, nor the findings from the official investigation.

 

[The Ham]

Bear in mind that the train travelled for more than a mile from the point where the brake was initially applied to where the two trains collided but it's speed was only reduced by 30-34mph.

One thing I would point out is that reducing speed under braking isn't liner, this means that the faster you are going the less speed you lose for the same distance of braking compared to lower speeds.

As such if (and I'm not suggesting that the driver would be able to guess to brake sooner, but if there was a system which advised that they slowed a little sooner so that is they maintained a suitable speed that they should arrive at the signal with it being clear) they had slowed down a bit more before the point that the brakes were actually applied then even if the rest of events carried on as before then they would have been slower.

Not just (say) 20mph slower (if they had slowed by 20mph before breaking where they did) but the speed that the brakes would have reduced the speed by more than the speed that they did, as such it could have been that the train slowed ~45mph due to breaking (say 32mph slower than the actual impact).

Even if breaking conditions had been bad then the extra distance would have slowed the train by a little more, if the distance was (say) 10% more it may well have slowed the train by (say) 15% more. That would have put the reduction in speed in the 35mph to 39mph range, which may not sound a lot slower, but it could have reduced the severity of the crash.

Now I know that trains and cars have different breaking distances and profiles, however to illustrate the point a car traveling at 80mph in a 60 limit will still be traveling at 53mph when a car doing 60mph had stopped and require 65% more stopping length (122m for a car at 80mph be 73m for a car at 60mph).

Yes it's likely that there would have still been a SPAD, it's also likely that there could have still been a crash, however the impact speed would have been reduced. The other thing to note, if the impact had been later by even a matter of (say) 15 seconds then it may well have been that it would have been into the rear of the leading unit and at a speed which may have resulted in fairly limited damage to either vehicle and potentially no derailment.

All of the above isn't critical of what the people involved did, they were working with the information to hand, rather it's highlighting that with the technology available (and not taking away the need for skilled drivers as it's not automating the trains) it could have been easily possible to reduce the severity of the crash, even if that was only enough mean that both trains lived to run again.

Whilst I understand that we are never going to design out all problems without only driving trains around at 5mph and with 10% of services running (and even then I'm not sure we'd stop everything), however there's still steps which can be taken to reduce the risk of harm without materially impacting the operating of the timetable.

Whilst it's right that we should ensure that everyone thinks of SPADS as a bad thing, there also should be an understanding that unexceptional circumstances that they will continue to happen.

The question should be "Is there anything which could be changed to further limit the risk of harm at an acceptable cost?"

Costs may include time costs as well as financial costs. Likewise limit doesn't mean remove entirely (even though the ideal would be that is the net result).

 

[millemille]

The question should be "Is there anything which could be changed to further limit the risk of harm at an acceptable cost?"

Costs may include time costs as well as financial costs. Likewise limit doesn't mean remove entirely (even though the ideal would be that is the net result).

The solution exists already, Dual Variable Rate sanders (DVRS - see RSSB research project T1107), and is being subject to a full fleet service trial on the Class 323's in leaf fall of 2023.

The testing at Old Dalby and on the two WMT 323's fitted with the pilot equipment proved that 6%g, or better, stopping could be achieved no matter what the railhead conditions. The concept and equipment now needs to be subjected to normal service conditions with drivers from the normal driving pool to ensure that the results read over.

 

[norbitonflyer]

There is no way of knowing that a conflicting move was signalled - it could simply have been another train ahead, or an obstruction, or any one of a number of other reasons.

Any one of those could clear before the driver reached the red signal (OK, maybe not an obstruction). What it wouldn't have been at that location is an approach control - using the signals to enforce a speed restriction. In such a situation the line is usually clear ahead and so the signal does usually clear on approach. As I have read elsewhere, if you are expecting to stop, you will brake in such a way that you will come smoothly to a stand at the signal. If you are expecting it to clear, you will slow to the required speed and then maintain it in order to keep momentum - whilst of course being prepared to stop if the signal does not clear. Neither approach is dangerous, but the latter will, on average, cost less time and less power.

 

[Surreytraveller]

No but there should be safety mitigation in place

Stopping a train from 90mph so two can cross at 20mph is asking for trouble.

The GWR that was hit according to that report was already 18 late the SWR would have been past at 50mph adding 1 or 2 minutes at most.

Signallers should absolutely be aware that stopping a train in leaf fall season from 90mph cam result in something like this and make decisions accordingly

Signallers have every right to expect a train to stop at a red signal. And adding additional delay into another train may well have additional risks down the line. You cannot run a railway like that

 

[edwin_m]

The solution exists already, Dual Variable Rate sanders (DVRS - see RSSB research project T1107), and is being subject to a full fleet service trial on the Class 323's in leaf fall of 2023.

The testing at Old Dalby and on the two WMT 323's fitted with the pilot equipment proved that 6%g, or better, stopping could be achieved no matter what the railhead conditions. The concept and equipment now needs to be subjected to normal service conditions with drivers from the normal driving pool to ensure that the results read over.

The 159 fleet was the first to be fitted with emergency sanders in about 1995, and I think the then Wales and West 158 fleet was fitted soon after. I imagine these have since been replaced by something more modern, but I'd be surprised if they were removed without replacement. So I'm also surprised the report doesn't mention them.