A Perfect Storm: The Story Behind The London Bridge Delays
We really had been trying to not keep writing about London Bridge. Then, at the beginning of the year, a lot of problems emerged rendering it an apt topic for discussion again. The trouble was that whilst pictures, headlines and hyperbole were easy to find, verified facts were far harder to come by. With a recent GLA Transport Committee Meeting having concentrated solely on London Bridge, however, a steady trickle of information from other sources finally beginning to surface and enough original suppositions confirmed, it is finally time to give some explanation as to the causes of the recent chaos at the station.
Network Rail – not quite open
This article really should not be needed. At the recent GLA Transport Committee Meeting on March 27th (you can watch it for yourself here) the Network Rail delegation tried to assure the committee that they were not trying to hide anything. Network Rail have also shown how open and honest they can be with their own report on what went wrong at King’s Cross last Christmas. That report was a model of openness and one felt that nothing was left out or suppressed.
One cannot so far say quite the same of Network Rail and the Train Operating Companies (TOCs) when it comes to the problems at London Bridge and the reason for not getting the full story in one digestible portion is unclear. The older members of the team here at LR Towers have been reminded of visits to the doctor in the 1970s – it was hard to shake off the impression that a detailed explanation was being avoided because it was felt it might confuse the patient.
It is not that we are accusing Network Rail (or the TOCs) of malice. Their representatives at the Transport Committee meeting clearly had a hard time, despite what one felt was a genuine commitment by the managers involved to get on top of problems and make life better for the passenger. It is just that somehow they don’t quite seem able to get the message across about what went wrong. Nothing significant is being hidden now, but it is noticeable how many different sources one has to go to in order to try and build up a picture of what happened. It is also noticeable how one or two minor details simply are not emerging.
What follows is what we believe happened. In the absence of any further information from Network Rail or the TOCs all the available evidence points to this. Ultimately, it seems that four fundamental problems occurred at London Bridge following the Christmas blockade and we explore all these below. We’ll then look to the future to see if we can expect any improvements.
Problem 1: An evening peak timetable that would never have worked
Incredible as it would seem, what has emerged as a clear fundamental problem is that the planned timetable for January 2015 of trains from London Bridge was unsound and was just unworkable. It is difficult to believe that this could be the case but it appears that a combination of circumstances exposed a fundamental modelling and train operating flaw that until now never got a mention.
To understand how this happened we have to look at what the proposed plan was for London Bridge, morning and evening. Three years ago London Bridge terminating platforms used to handle 30tph in its nine platforms in peak hours. Before January 2015, the maximum service that could be run was 24tph using the remaining six platforms. The 2tph South London Line service (Victoria – London Bridge via Peckham Rye) finally ceased running in December 2012 so the 2014 timetable had to remove an extra 4tph from the peak timetable. This could be done in a relatively pain-free way but it was known that beyond that difficult (and unpopular) decisions would have to be made.
Given the extreme inconvenience to passengers of reducing the service further from January 2015, a very detailed modelling exercise was undertaken which simulated what would happen from January 2015 on the revised track layout. The objective was to run as many trains as possible and the results seemed to suggest that 22tph was difficult but achievable. This was from a software package that had been reliable in the past and was known to closely match reality.
The model is not quite right
Unfortunately it appears that, unrecognised at the time, there was a major problem. We get the strong impression that the model, with the appropriate parameters input, had never been implemented in a situation where the critical factor was train movements over a station throat (the part of the station where tracks divide to enter platforms) with no margin whatsoever for making up for any delay. This is most likely to happen at a station where the number of available tracks for trains leaving the station is limited – with one being the worst case. London Bridge manages slightly better than this with one and a bit (a “down” line and a reversible line), but it is about as close to that theoretical worst case as you are likely to find in a major metropolitan station.
The lack of experience of using the modelling in the situation described and an inappropriate parameter would soon turn out to expose a major weakness. For what was quickly discovered when the timetable came into force was that although the new timetable could conceivably work in the morning peak, it just could not do so in the evening.
The problem appears to be that there is a slight delay between a train being signalled that it can leave the platform and it beginning to move out. This is because, when the proceed aspect appears, the station staff have to satisfy themselves that the doors can be closed and then tell the driver or guard to “Close the Doors” (in practice done by pressing a button that causes the letters CD to light up next to the signal). The driver or guard then closes the doors and, once the platform staff have confirmed from the outside that the doors are properly closed and the train is safe to depart, the platform staff give the “Right Away” authority (once again in practice by pressing a button that this time causes RA to be displayed next to the signal).
This delay is inevitably a fraction longer in the evening peak than in the morning peak as the process of dispatching a train being is much more challenging when there are a lot of people on the train, others still trying to board and possibly additional people on the platform block staff sight lines on the platform. It is that few seconds extra that appears to be causing the problem. Unfortunately the value specified for this parameter during the modelling was in hindsight too optimistic.
Whilst the issue seems obvious once pointed out, it seems never to have been a problem before. Although there are circumstances where the evening peak is less intense than the morning peak (SouthEastern being a well-known one) the reason for this has never been given as the fact that it takes longer to dispatch trains in the evening from a London terminus than it does in the morning. It may well be that, until now, no-one bothered or thought it necessary to distinguish between the morning and evening peak and thus when modelling the same value has been used for both.
Southern Railway, Network Rail and Govia Thameslink have published a joint document with their improvement plan. In it they state matter-of-factly:
In addition, we are still working to stabilise the evening peak period as we have less capacity in the evening than the morning at London Bridge. This is because it takes longer to dispatch full trains than empty ones, and while all trains are full in the evening, over 30% of the morning trains are not in service when they leave the station.
Now if this was known to be a problem when writing the timetable and the timetable aimed to run the maximum number of trains possible then why on earth did they attempt to run the same number of trains in the evening peak as the morning peak? Instead it seems clear that this problem, or at least the extent of it, was not realised when the timetable was written.
Timing is everything
If we have read the signals correctly and identified the true cause of the timetable problem we then have a obvious question to ask: How come this problem wasn’t identified in advance by simply using a stopwatch to get an appropriate value to feed into the model? It does seem a bit of an oversight, so it initially suggests that there may be more to this than we have surmised.
After preparing and rejecting a long list of plausible explanations to explain this discrepancy we keep coming back to the notion that maybe it was simply never realised that the figure obtained varied by a small but significant amount depending on what time of day the timings were taken.
Getting out of this mess
Once the problem had been identified in the evening peak then next thing to do seemed obvious – remove two trains an hour from the evening peak departures. The West Croydon trains were chosen for two reasons. The first reason was because it was considered that there was an alternative route available (Jubilee Line to Canada Water then London Overground), although whether or not passengers could actually get on a Jubilee Line or London Overground train is another matter. The second reason for choosing the West Croydon trains was that they were fairly easy to remove from the timetable without knock-on effects or additional complications.
Those wondering why there were multiple minor timing announcements in the weeks following the initial disruption will find their answer here – that extra few seconds meant there needed to be some minor retimings which were worked out during this period. This also applied to the morning peak timetable, which also needed a bit of tweaking, but no removal of trains.
Problem 2: Operational teething and implementation troubles with the timetable and signalling
It is almost inevitable that when a timetable is significantly changed it takes a few days for it to settle down. This is because staff and passengers take time to adjust. It may be trivial things like a drivers checking which stations he is due to call at or passengers hesitating for a second or two before boarding trains.
In the case of the new January timetable there were a lot of other factors that would take time to “bed in”. Drivers were using a new track layout that they had never actually driven on before, signals were located in different places and two of the platforms at London Bridge were completely new.
Of much more significance, trains which had been controlled from the old London Bridge Signal Box (which dated from the early 1970s) were now controlled from a state-of-the-art Regional Operations Centre at Three Bridges in Sussex. The timing of the move would have seemed utterly logical to a signal engineer but operationally it meant people who were perhaps unfamiliar with the approaches to London Bridge and the services run on it were using equipment that was new and possibly alien to them to run trains to a timetable that had just been introduced. Worse, in the first week Charing Cross trains on SouthEastern were still calling at London Bridge. This meant that, because one change had not yet kicked in, more people were going to be using London Bridge than in subsequent weeks.
It does seem a bit surprising that the plan was not to stop Charing Cross trains calling at London Bridge from the beginning of January – even if they were still capable of doing so – in order to ensure that the worst first week was not even busier than the others on the Southern “low level” side.
In this “Big Bang” approach in January the gateline position was changed as was the location of the departure screens. This was quickly identified as causing a problem (and, it has to be said, quickly rectified) as people waited at inappropriate places on the concourse in order to be able to see the screens. Another minor problem adding to the whole.
As a result of the experience at London Bridge, South Western Alliance has vowed not to move the signalling control to the new Regional Operations Centre at the same time as other changes at Waterloo. It will be interesting to see if these good intentions will be feasible or whether the reality is that it is just not practical (or incredibly expensive) to adopt a multi-stage process to this.
Problem 3: Signalling Equipment Failures
There is a well known traditional “bathtub curve” of equipment failure. In the modern world, with quality assured testing, it is hoped that the expected initially high failure rate is a thing of the past. Unfortunately with the new signalling at London Bridge this definitely wasn’t in this case.
Following on from the fracas with the overrunning engineering works at Christmas at King’s Cross, the recent Transport Committee naturally homed in on this and wanted to know exactly what the problem was. The answers given were slightly evasive – which didn’t help – but the problems do appear to be genuinely unforeseen problems using equipment that was previously thought to be reliable.
One problem was a particular set of points at New Cross Gate. It appeared that this failed intermittently multiple times, but that the cause could not initially be determined. The impression given and rumoured was it that it was down to the contractor and supplier of the equipment and “personnel have now undergone retraining” to ensure this is not repeated. What has never been clear is whether the error was an installation error or one that occurred in the factory where the points were assembled. There also appears to be a reluctance to blame the contractor as this goes against modern attitudes and working practices where the approach is “we are all in this together”.
A further problem that has previously never publicly come to light was an earthing fault. Earth faults can be extremely difficult to track down. Sometimes it may not even be obvious that they are the cause of a particular problem. Croydon Tramlink’s opening was delayed for months whilst earthing issues were sorted out and anyone who has an electrical circuit in their house that occasionally trips for no apparent reason will appreciate how difficult these can be to identify and fix.
It appears that the earthing fault here was in one of the new signal rooms. Again, despite promises of not wishing to hide anything, Network Rail haven’t explained the exact circumstances. This problem occurred despite using equipment that has worked successfully and reliably elsewhere. This earthing fault turned out to be the ultimate cause of a number of seemingly unrelated failures and since being fixed we are assured that the equipment is working reliably.
The problem is not now at London Bridge
What has also emerged at the Transport Committee is that most of the problems affecting London Bridge now do not originate there. Typically they are signal failures further down the Brighton Line. The trouble is that, with everything at London Bridge having to be exactly sequenced with no margin for error, it is at London Bridge that the problem truly manifests.
It was notable that even before the work at London Bridge the Brighton Line was starting to suffer a regular stream of signal failures, which is hardly surprising given the age of much of the signalling equipment. The problem appears to be that now it is having even more impact. In the past few months there have been only a few closures due to Network Rail working at London Bridge on the Southern Side. Meanwhile there has been other work going on further down the line as Network Rail tries to either belatedly catch up with the backlog of maintenance and renewal elsewhere or at least keep the equipment in good enough repair.
Problem 4: Driver Shortage
Driver Shortage is a problem that simply shouldn’t have happened. It is also one that cannot be blamed on Network Rail – the TOCs being entirely responsible for providing drivers.
Again this is an area where the full picture has yet to be made fully clear. Unlike the other problems it also predated the Christmas track layout change at London Bridge. There appear to be a number of issues.
The first problem appears to be that when Govia Thameslink Railway (GTR) took over from First Capital Connect (FCC) they found fewer drivers than expected. FCC seemed to have just enough to run a basic service but that was all – and even that assumed voluntary overtime. This does seem to be a fundamental weakness of the franchise process and it may be that in future the DfT needs to ensure that TOCs do not run down their trainee driver numbers towards the end of a franchise.
This initial driver problem was then made worse by an unusually high level of winter sickness that swept the country towards Christmas. On top of that, nearer Christmas drivers were naturally more reluctant to work overtime and, at the same time, it was necessary to give drivers two days of training to cover the new layout that would come into operation at London Bridge after Christmas.
Things got so bad that Southern resorted to withdrawing all its Milton Keynes services in the week before Christmas just to ensure they could run the rest of the service fairly reliably – though there were cancellations there as well.
Unfortunately, although Southern and Thameslink have now recruited new drivers it takes over a year to train them. Charles Horton, who heads GTR, claims that they have enough drivers now for normal day to day service provided everything else runs smoothly. The trouble is, of course, other things are not running smoothly and trains continue to be cancelled due to staff shortage.
One reason for the need to recruit so many drivers is to cover future training. Not only will drivers have to be retrained on the new class 387 entering service, they will also need to be eventually retrained further on the Siemens class 700 Thameslink trains. On top of that there will be training for signalling under ERTMS prior to 2018.
It can work
In short summary of the above, the main causes of the problems that occurred at London Bridge at the beginning of the year are now solved or largely solved.
There is cause for hope. Unfortunately at the Transport Committee this glimmer of light was given at the nadir of the Committee meeting when frustration was at its highest. A combination of a less than savvy remark and committee members by now interpreting everything in a negative light meant that the point being made was lost.
It was explained to the Transport Committee, quite reasonably, that things were getting better and that on the previous Thursday and Friday a reliable punctual service had run all day. Unfortunately the significance of this was not previously explained, so a by now frustrated Committee treated this comment, quite unjustly, with derision. The obvious comment was made about treating a good service on two days as no cause for congratulation and other comments were made in similar vein.
The point, however, that the transport committee seemed unable to grasp was that running a good service on two consecutive days demonstrates that the current timetable is basically sound and can be made to work. This is highly significant because it shows that the way forward is to build on that reliability, whilst also not falling into the trap of thinking that the problem is fixed. As well as explaining why running a good service for two days was perhaps cause for hope, it might have been a good idea to draw an analogy with the Jubilee Line resignalling – which went through considerable teething troubles but now works very well indeed. So, the timetable has been more or less sorted and it is now time to move on and deal with other issues that are causing problems.
Inevitably, time was wasted in the Transport Committee Meeting with members demanding to know when things were going to run properly and the response the were getting was that there will be steady improvement, but that Network Rail can’t give a date by when the service will reach a particular standard. Unsurprisingly this did not improve the atmosphere.
May Timetable improvements
Now that everyone is confident that they have a timetable that can work, the outlook is more positive and the next stage is to make that timetable as effective and robust as possible. One important thing to do is to reallocate the drivers’ diagrams to the services that are running in a more optimal way so drivers aren’t sitting around doing nothing useful because their train was removed from the timetable for one of its journeys.
The advantage of this reallocation of diagrams is that it will free up drivers, which will leave more spare and therefore reduce the possibility of trains being cancelled due to staff shortage. Inevitably this will involve minor alterations to the timetable and this revised timetable is due to be implemented as part of the May timetable change on Sunday 17th May.
Perhaps unsurprisingly the Committee wanted this implemented straightaway, but is unlikely to have been finalised yet and there are undoubtedly other complications. The new timetable and drivers diagrams are going to be in operation until the December 2015 timetable which itself will probably take into account changes made to the current timetable. It makes sense not to rush it and to get it right.
Not mentioned by Network Rail was the fact that by the time the new timetable is introduced one would hope that platform 10 at London Bridge will be completed and in use to its full operational length. One would hope this would lead to a little bit more flexibility. Currently care has to be taken to only timetable and use it for trains of 10-car maximum length. Having this as a full length platform should also make it possible for more trains to consist of 12 cars.
A warning for the future
Whilst a steady improvement can be hoped for, there is a bit of a metaphorical dark cloud that may appear this summer in the form of sunny days. Not all the trains that run into London Bridge are air conditioned and there is the potentially serious problem of passengers being held outside London Bridge in packed non-air-conditioned trains. Clearly in such circumstances these trains have to be given priority as far as possible which is going to cause considerable problems.
A further problem, exacerbated by this, is that by far the most logical and sensible thing to do after disruption is normally to get trains dispatched out the station even if passengers have not had time to board. Understandably this tends to upset passengers who can’t see the bigger picture – or don’t care and are understandably fixated on their own personal needs.
During disruption one normally has trains waiting outside the platforms. There is an issue that people on some of these trains, if they have been delayed, may be hot and close to fainting or in desperate need of the toilet (if there isn’t one on the train). In any case one wants to clear the backlog as soon as possible and the way to do this is to get rid of the trains in the platforms as soon as one can. One reason for this is that there may not currently be any non-air-conditioned toiletless trains waiting outside the station but if you don’t clear the queue of trains there will be at some point in the future.
Apart from helping prevent a queue of trains building up, or being sustained, outside the station, getting trains to depart from London Bridge as soon as possible after disruption can also do a lot to relieve important stations down the line, most notably East Croydon, which probably by then has a build up of passengers waiting for a train. To put it another way, in such circumstances a train might leave London Bridge half empty but it may well be full anyway by the time it leaves East Croydon.
Things can only get better
In the long term things will improve dramatically as tracks into London Bridge are restored once the Bermondsey diveunder is completed. In the short term it does seem fair to say that an unfortunate and unlucky combination of events has produced a lot of problems. With these largely solved it is probably fair to say that things can only get better. The question really is how quickly and by how much? Only time will tell.