New Signalling as part of the sub-surface lines modernization (updated)
On Monday 28 March, I spent a little while at the western end of London’s Hammersmith & City Line watching the new automatic trains doing their business: this was the first day the system was actually in passenger service. Travel west of Latimer Road (the present boundary of the first stage of automatic operation) was noticeably different from performance further east, in particular the trains accelerated more quickly. Coming into Hammersmith was interesting. There is only a 3 metre or so space between the approved stopping point and the buffer (behind which is the concrete mass of the station concourse) and special measures have been taken to make sure there is no possibility of a train not following the stopping profile or unexpectedly powering up.
This is the first stage of the resignalling programme for the whole of the sub-surface network. You will recall the various contractual failures that have much-delayed the new signalling; this has hopelessly disrupted the expectations set at the start of the century that we would have had the improved train services across the whole of the sub-surface lines by now. Although we still have not had most of these improvements, nor will we see most of them for a couple of years or so, the apparently successful introduction of new signalling and automatic operation is at least an encouraging sign it is on its way.
However, even the present signalling contract has hit snags and this modest section of automation (Hammersmith to Latimer Road) should have been in service last summer. After some problems were identified during trials, the decision was made to hold off until all worked perfectly. I dare say this was prudent. In any case it was rather the point of doing this small section in advance. I am led to understand that many of the challenges were software related; this seems to be the way of things at the moment in the railway (and apparently the airline) industry.
Assuming this section now performs OK, it is hoped to make up lost time elsewhere as equipment installation has been carrying on in the meantime. I believe the plan is to introduce the rest of the west end of the H&C and the whole of the north side of the circle (and Met to Finchley Road) in one go when the existing section has proved itself. I’m looking forward to seeing how well Baker Street and Edgware Road cope with this.
Some observed benefits
One of the more visible changes, and a useful one at that, is the display of the train ready-to-start signal in the form of white indicator lamps on the train exterior. Regulars will probably work out quite quickly that this means the doors are likely to close very soon and for those in the know it is easier to spot the bright white lights than the old starting signals (all the old signals have been taken out of use).
At Hammersmith, I was surprised to hear the happy sound of the quick-acting air-operated points that have been retained for the time being, and I hope the new electric point machines that will doubtless turn up in due course are not too much slower. I understand the old (1951) signal box closed on Friday evening (25th) after traffic, leaving only four lever-operated signal boxes left on the system, at Edgware Road, Whitechapel, Harrow and Rickmansworth.
Coming back into town during the evening peak I was reminded how heavily used this line is, especially east of Paddington. Also observable is the enormous amount of development along the line which will at some point be further adding to train loadings. At Edgware Road passengers waiting near the front already had difficulty getting on. The present 5-minute train service really does struggle to handle the demand east of Paddington. As the new signalling is introduced, and as I understand the train service proposals to be, intervals will be improved to 4-minute intervals, and finally 3¾-minute intervals. This is obviously better than now and I look forward to it. Of course the arrival of Crossrail ought to ease the loadings at Paddington H&C, when it finally arrives, which I suppose might be next year (or the year after).
But is performance any better?
Returning to the new automatic section at the Hammersmith end, I think it is worth asking what success looks like. From a strictly technical point of view I have no doubt ‘success’ will be suggested by very high reliability, nothing untoward being discovered, and perhaps train performance falling broadly within expected parameters. It is this last area that interests me. As a railway operator I am less interested in the technical wizardry than the outputs that are required to run the service. For example, can the service match the traffic presenting itself to appropriate levels of performance and reliability? By performance, I mean that bit of magic that minimizes journey time by maximizing train performance and eradicating unwanted and unexpected delays. We have been promised much during the endless process of getting modern signalling, but during the six visits I have now made to this fully-automated section I have been left wondering what has changed.
In a nutshell, I have measured wheel start to wheel stop times between stations and find that on the automated section the start-stop timings are virtually unchanged from pre-automatic days, and to the extent there is any difference, timings are now a tad longer. As one who is used to the brisk performance on the Victoria Line this is slightly disappointing. Analysis of the run times suggests that top speeds are unchanged, acceleration rates have improved so that top speed is reached ten seconds faster than hitherto, but braking rates have much reduced, increasing braking time by about ten seconds. The slightly longer run at top speed gives some advantage in automatic, and although the net result would be a slightly quicker overall run we find that the automatic system introduces an extra delay before the operator can open the doors. I measured the average door open delay as 3.5 seconds on an automatic train compared with no more than 1 second in manual driving. This converted a slight overall speed improvement into a slightly slower run instead. (By door open delay, I mean the interval between wheels stopping and the doors beginning to open.)
The braking I found rather odd. Apart from its lack of vigour, the trains slow down to about 1 mph while the equipment appeared to be seeking the exact inch within which the train had to stop, which it then did abruptly. This, typically, added a couple of seconds more braking time than would otherwise have been necessary (and this does not happen on the other four automatic lines). It looks to me as though some tweaking is necessary, both with the braking arrangements and the door-open delay. This alone would claw back five seconds or so per stop.
Returning to the ‘moderate’ brake rate, I have found out that despite automatic trains having been around for decades there are still some unresolved challenges. These relate to the theoretical possibility of drift in knowing precise position (for example because of missing a track beacon and relying on a possibly poorly-calibrated wheel-speed sensor). There are also plausible opportunities for wheelslip, and therefore in predicting the actual braking rate of the train for any given demand for braking. Moreover wheelslip protection (provided by rolling stock manufacturer) is quite slow in operation, while position correction (and consequential adjustment in brake demand) is provided as part of the train protection and ATO systems and reacts faster; the two responses can attempt to counteract each other. Although this is unlikely to be dangerous, it might precipitate an emergency stop which is undesirable anyway on a passenger train and is more so if there is actually a wheel slide in progress. Although engineers have attempted to mitigate risk of temporarily losing position approaching a station by installing additional track beacons, the reality is that uncertain adhesion in the open air is thought most easily mitigated by reducing braking rate to a level where it is unlikely to occur. Hence my expectation of a brisk Victoria Line style stop isn’t going to be seen on the open sections of the Hammersmith & City any time soon.
I should add that although it is tremendously easy to reduce the braking rate compared with fixing the adhesion and positional uncertainties, it doesn’t come free. S Stock has an emergency brake rate of 1.4 metres/second/second (m/s/s), which is too much for passenger comfort and normal service braking is designed to fall in the range 0.2 – 1.15 m/s/s. These are nominal rates and at the higher end cannot be depended on for the reasons already stated. Perversely, it is expected that on manually-driven trains the operators will detect wheel slide issues and correct for them whilst automatic systems are apparently not trusted. Because high brake rates might be untrustworthy, a notional brake rate (the ‘guaranteed’ brake rate) is defined, and this is considered dependable; I believe the S Stock guaranteed rate is set at 70% of the maximum, which would be about 0.8 metres/second. However, at the moment, rates I actually measured averaged about 0.6 m/s/s (except at Hammersmith where it is lower). In fact for most of the deceleration period the rate was a tad higher (maybe as much as 0.7 m/s/s) and the average was brought down by the faffing about near the stop mark where the train was trying to find its mark.
Lest anyone thinks this doesn’t matter, if we take the Victoria Line as a fine example of what ATO can achieve, the nervousness about open air brake rates on other lines means a reduction of service brake rate of about 40 per cent. This adds 10 seconds to the running time between stations, at every station. This may not sound much, but on a run between Hammersmith and Barking that is roughly an extra ten minutes a round trip. It is not negligible. Looking at the whole of the subsurface system we are talking of perhaps half a dozen more trains than we would otherwise need simply to mitigate the effect of the constrained braking. And if you reckon a train as £10 million each, that is a lot of money tied up which you could argue might be better spent on finding a solution to the adhesion risk. This is all before we consider cost of extra journey time to passengers who are expecting service improvement. It is particularly galling to see (live) train operators accurately braking more efficiently than the new ATO system because they can look out of the window and assess wheel-slide risk, and have their experience to fall back on. I measured the stopping time at Ladbroke Grove (outside the ATO area) and the operator decelerated the same train more briskly and saved six seconds!
It is interesting how introducing automatic operation has allowed us a higher acceleration rate but a reduced rate of braking. I should add that the higher acceleration arises simply because the trains were previously limited to a performance similar to the old trains because of the traditional signalling and the need to keep within the safety envelope. With the old signalling decommissioned the full rate of acceleration could be provided.
I imagine that the braking rate in automatic operation will be significantly increased in the central London tunnel areas when the system is expanded, so we can look forward to some livelier performance. However, since the majority of the sub-surface lines are actually in the open air, there is a concern that most journey times will not be altered very much. It is true that automation and the overlaid train management system would be expected to improve regularity. It is also true that new signalling will reduce delays as the existing system is a bit crude, a matter not helped by the removal of signals and lengthening of signal overlaps to address perceived safety issues a few years ago. Some top speeds might be slightly higher. Reliability should also be much-improved. This is all to the good, but for the enormous cost of the new system, now a decade late, some livelier train performance was hoped for.
Train Working at Hammersmith
Irrespective of automatic train operation, to achieve the improved frequencies referred to earlier we will need smarter train working at Hammersmith than I saw whilst watching it and I wonder if we will get it with the present track layout, which has not altered since 1906 (this has the important crossovers a long way away from the platform ends, much increasing clearance times). The crossover leading to western platform is especially distant from it and holds trains unnecessarily far away from platform (it is located to allow convenient access to a goods yard closed 50 years ago). I watched a train unaccountably held at this very place during one visit, even though that platform had been empty for many minutes before the train eventually came in, and it caused at least a minute’s delay. I suppose the staff at the new signalling centre (next to which it had stopped) are still getting used to the new kit.
Actually, Hammersmith is not a very satisfactory terminus. It is tremendously busy these days but very large numbers seem keen to interchange with the District and Piccadilly station or the bus station to complete their onwards journey. It is a rotten interchange too, and the need for very large numbers of people to cross two busy roads seems to overwhelm the crossings. Can we not do better than this?
UPDATE 28 JULY
Just out of interest I have been riding the automatic section of the H&C between Hammersmith and Latimer Road regularly, and usually time the journey to see how the tweaks (if there are any) are getting along. My last visit was 24th July.
I have tried to persuade myself that the time delay caused by the ‘hunting’ for the exact stopping mark has been reduced and that the door open delay in automatic mode has similarly been reduced. Although I think there have been some software-related performance changes, I am not, unfortunately, really persuaded overall performance is much better. Taking the wheel-turning time between Goldhawk Road and Latimer Road only, I find the run time was between 2m 53s and 2m 55s when the trains were driven manually, and they are taking 8-10 seconds longer now. Even if we use the lower number, this is disappointing and represents an extra 3 seconds or so a stop, which mounts up over a long journey.
Of course, there may be some improvements in the pipeline and as mentioned in the original blog the performance will probably be livelier in the tunnel sections. There will also be service improvements resulting from the use of a train management system (I hope). Even so, the exuberant public relations story we have been given for the last few years while the new signalling was being planned did not hint at potentially slower journeys!
I have also experience on my trips out further heavy overcrowding between Baker Street and Edgware Road and much passenger puzzlement at Edgware Road as visitors to London try and get to grips with the confusing operation of the Circle Line here (without much help from staff, I may add). The Baker Street to Hammersmith and Gloucester Road part of the Underground really isn’t satisfactory for today’s flows and I wonder if something more imaginative could be considered. I know Crossrail will (one day) help, but even so Edgware Road is not a satisfactory in-town terminus in today’s conditions.