Friday 9 May 2014

What is... Automatic Train Operation? (And Why Do We Still Need Train Drivers?)

I'm often asked... sometimes asked... well, somebody asked me once: what do train drivers do? How hard can it be? Why can't it all be automated?

First thing: driving a train isn't like driving a car. The key difference is in braking. When driving a car, you brake on sight - you see something in the way, be that a pedestrian in the road or a car ahead slowing down, and you brake. When driving a train, unless you're going no more than about 20mph you haven't a hope of braking on sight, because trains brake much more slowly.

More starkly: a car doing 70mph can stop in 75m (plus thinking distance); a train doing 125mph takes about 2,050m to stop - nearly a mile and a half. So if a train driver sees an obstacle in front of him, he's going to hit it.

More than that, though, if the driver doesn't know where the signals are, where the speed restrictions start, or exactly where the stations are, he's not going to know where to brake. As a simple example, a fast train heading towards Coventry station from Birmingham will typically start braking before it gets to Canley station (which is 1½ miles out from Coventry). If it's been raining, there won't be as much grip, so the driver has to brake sooner. There's no sign telling the driver where to brake; the driver has to know.

The term given to this requirement is route knowledge: before a driver can drive a train unsupervised on any railway line, they must have sufficient route knowledge for that specific route. A typical, reasonably simple, route might require about 60 journeys back and forth before the driver can "sign the route", meaning he has been assessed as having sufficient route knowledge to drive a train unsupervised on that particular route.

Some drivers will only sign one or two routes; some drivers will sign hundreds. As an example, drivers based at Oxford sign the Great Western Main Line from London Paddington to Oxford, as well as the branch lines to Banbury and Great Malvern; in total, about 150 route miles. That may sound like a lot, but that means an Oxford driver can't drive a train to Swindon, Bristol or Newbury, all common destinations for a train from Paddington.

As a result, timetabling trains isn't just a matter of planning where the trains themselves go; it also involves ensuring that it's driven by a driver who signs the route. This isn't too hard to plan in advance, but it makes life complicated when things go wrong: if the controllers at Paddington want to send a train to Bristol and the only people around are Oxford drivers, they're stuffed.

So surely it would make life a lot easier if all the trains drove themselves automatically? In theory, yes, and we are at the stage where the technology is there that would allow almost all passenger trains to be driven automatically; this is generally called Automatic Train Operation (ATO). However, ATO does have its drawbacks: it's neither easy nor cheap to install the technology, and it involves modifications to both the tracks and the trains.

The modifications to the trains are particularly tricky: while it's easy enough to build trains with ATO equipment, retrofitting older trains with ATO equipment - even if they're just a few years old - can be a nightmare. It's not nearly as simple as, say, adding an after-market CD player to your car. A reasonable analogy would be converting a manual gearbox car to an automatic gearbox: it's technically possible, but for the amount of trouble it would be to convert the old one, it's often easier to just buy a brand new one.

Moreover, since one of the advantages of automating the driving of trains is more efficient braking, ATO requires every train on a route to be converted (or replaced) before there's any tangible benefit over human drivers. The capacity of a line is determined by how closely you can run trains together, and that's determined by how well a train can brake. When there's more than one type of train on a line, the capacity is generally limited by the worst-performing train on a given line.

For example, the line between Coventry and Birmingham has 7 passenger trains an hour in each direction, plus two extra that just run between Birmingham International and Birmingham New Street. Trains can (and do) run as little as 4 minutes apart, so in theory the line could take several more trains. The catch is that they'd all have to plod along as slow as the train stopping at all nine intermediate stations: currently the fast trains take just 20 minutes, while the stopping trains can take 35 minutes.

Now, ATO might manage to improve that a little, by safely allowing trains to be closer together; but it would require all the trains that run on the line to be converted. However, the trains that run between Coventry and Birmingham aren't self-contained: trains can start their day in Birmingham and end up as far away as London, Liverpool, Exeter or Edinburgh. I estimate there are 250 passenger trains which could, in the course of a week, run between Coventry and Birmingham. (Never mind freight trains, which are even more complicated.)

However, on self-contained networks with a dedicated fleet of trains, ATO can work wonders. Three London Underground lines have ATO: the Victoria line has had ATO from when it opened in 1968, and was in fact the first railway line in the world worked entirely by ATO. The Central line was converted to ATO in 2001, shortly after the introduction of a new fleet of trains. The Jubilee line was converted to ATO in 2011; in this case the existing trains were modified to work with ATO.

On the London Underground, the advantages of ATO can be seen very clearly. ATO's principal advantage is that it can stop the train at a station much more precisely: human drivers tend to be a bit cautious, since delaying people a few seconds is much better than crashing into something. Freed from human inhibition, the ATO system knows exactly where to brake, saving vital seconds.

Once the Jubilee line had been converted to ATO, the previous maximum frequency of 24tph (a train every 2½ minutes) was increased to 30tph (a train every 2 minutes). Now, 30 seconds may not sound like much, but that 30 seconds means a 25% increase in capacity (and, incidentally, a significant reduction in journey times as well).

The Central and Victoria lines operate at 33tph in the rush hours - about 110 seconds between trains - and are currently the most frequent lines in London; but such is the need for greater capacity that most other LU lines are being gradually converted to ATO: next will be the Northern line, which should be converted by the end of 2014.

That said, all the LU lines running under ATO still have drivers: although they can drive the train in an emergency, or when the ATO fails, their principal job is to open and close the doors. On the Docklands Light Railway (DLR), which has been completely driven by ATO since it opened in 1987, the doors are opened and closed by "train captains": they don't sit at the front of the train, but they are also available to drive the train in an emergency.

Truly unattended train operation, while technically feasible, often leaves passengers somewhat uneasy; the efficiency of ATO cannot be denied, but knowing that there's someone on board should everything go wrong affords a peace of mind - both for the operator and the passenger - that cannot easily be dismissed. Nonetheless, it does exist elsewhere in the world; Paris Metro Line 14 is completely driverless (and, incidentally, runs at 40tph - a train every 90 seconds!).

If you want to read more about the debate as to whether the London Underground should go "driverless", I highly recommend this post on London Reconnections.

What of Crossrail and Thameslink? In both cases the aim is for 24tph through the core. While many London Underground lines run at 24tph or more without ATO, it was decided to install ATO on Crossrail and Thameslink, with a brand-new fleet of trains brought in to operate the enhanced service from 2018. Due to the complexities of the National Rail network, ATO will only be used on the central core sections: on Crossrail, between Stratford, Abbey Wood and Paddington; on Thameslink, between London Bridge and St Pancras.

In a sense these are the only parts of the line which justify ATO, since further out the frequency of services is less; nonetheless, having part of the journey run by manual driving and part by ATO is an extra layer of complication, and it remains to be seen how it will work in practice. Crossrail, with only two branches to east and west, should work well. Thameslink, though, will have many more branches and has a much shorter core section; one late train could have significant ripple effects.

Thameslink will be served by a fleet of 115 trains (designated Class 700), which will be built by Siemens: of these, 55 are 12-car trains and 60 are 8-car trains. While the core will be capable of taking 12-car trains, there will be too many stations (particularly those on the Wimbledon loop) which cannot be lengthened beyond 8 cars. It will be interesting to see how the ATO system copes with different train lengths; all ATO operation in the UK so far is confined to a single type of train.

For Crossrail, 65 9-car trains (designated Class 345) have been ordered from Bombardier, with the option of a further 18 trains for future extensions. Although built by different firms, the two fleets will look quite similar internally: both will have fewer seats and lots of standing capacity, with full-width gangways between carriages to maximise capacity.

Of course, Thameslink is already served by a large fleet of about 110 4-car trains, most of which (the class 319s) date to the late 1980s and have plenty of serviceable use left in them. Once replaced by the new class 700 trains, these will be surplus to requirements for Thameslink and will be available for use elsewhere. Similarly, once Crossrail opens, it will displace a number of trains from existing services on the GWML and GEML.

The ideal solution is for these trains to move to another line to replace some of the oldest trains on the rest of the network, which can then be retired: this kind of "rolling stock cascade" means that more than one line benefits when new trains are introduced. There's just one slight problem: the class 319s are electric trains.

With the arrival of the new trains for Thameslink and Crossrail, the supply of electric trains will exceed the demand from electrified lines for them to run on. In the next post, I'll look at how electric railway lines work, and the benefits and disadvantages of electric operation; I'll also explain the programme of lines planned for "electrification" to rectify the imbalance.

One day, no doubt, automation will take over from human drivers in almost all aspects of driving trains. For most of the railway network that won't come soon; Crossrail and Thameslink are but the first tentative steps into automating the entire railway network. In the meantime, the cascade of rolling stock resulting from the new fleets for Thameslink and Crossrail will have more immediate impacts across the country: in the next post I'll begin to explain exactly where the ripples will be felt.

Previous post: What is... The Thameslink Programme?
Next post: What is... Electrification?

Tuesday 6 May 2014

What is... the Thameslink Programme?

The Thameslink Programme is a £5.5 billion project to triple the capacity of the Thameslink line, which runs north-south through central London via St Pancras, Farringdon, Blackfriars and London Bridge. If you have no earthly idea what Thameslink is, I suggest you start by reading my previous post, which explains the history and background of the line. In this post, I'll try and explain how the £5.5 billion will be spent.

Until 2009, the Thameslink core between Farringdon and Blackfriars could only cope with a maximum of 8 trains per hour (tph) in each direction. Furthermore, the platforms at King's Cross Thameslink, Farringdon and Blackfriars were only long enough to take 8-car trains, even though many of the stations further out could cope with 12-car trains. As such, the Thameslink trains are horrendously overcrowded.

Indeed, Thameslink was horrendously overcrowded within just a few short years of the line opening in 1988: passenger numbers quadrupled within the first year. So in November 1991, a major upgrade, then entitled "Thameslink 2000", was announced. Through numerous delays, first due to privatisation and then to complex planning inquiries, the programme was only approved in 2006 and funding was forthcoming the following year. It was quietly renamed "the Thameslink Programme" to hide the fact that it was running about 18 years late.

The plan had a simple aim: convert Thameslink into a proper RER-style cross-London line with 24 trains every hour, most of them 12 carriages long, running in each direction in the rush hour. This necessitated a variety of modifications:
  • more destinations, both north and south, were needed to send trains to;
  • longer platforms were required in the Thameslink core;
  • the branch to Moorgate had to close;
  • most importantly, London Bridge had to be rebuilt to give it the necessary capacity.
The need for additional destinations was simple: while the core could be upgraded to cope with 24tph, there wasn't any need for 24tph to Bedford or Brighton. To the north, Thameslink connected only to the Midland Main Line (MML) out of St Pancras, but that took it quite close to the East Coast Main Line (ECML) out of King's Cross. Peak-time services on the MML already amounted to 14tph, so a connection to the ECML would provide enough capacity to the north to run 24tph through the core.

As luck would have it, such a connection used to exist: the lines now used by Thameslink - the "City Widened Lines" - had not one but two historical connections to the ECML (as seen in this diagram), one of which had only closed in 1977. Thought was given to simply reinstating one of these connections, but ultimately it was decided against just putting back the old connection, for one simple reason: it would have to be a flat junction, and that would have constrained capacity too much to ensure that 24tph could be run reliably.

A flat junction on the railway is much like a right-turning lane on a dual carriageway: trains wanting to turn right must wait for a gap in trains coming the other direction to make the turn. Grade-separated junctions, usually called "flying junctions" on the railway, is akin to a proper motorway junction: the conflicting movements are replaced with a bridge or a tunnel.
The principal downside to having a flying junction is space: even in the diagram above it can be seen that a flying junction takes up more land, and land in central London is nothing if not scarce and expensive. And while the old connections to the ECML could have been used, there wasn't the space around them to make them into flying junctions.

Instead, a plan for something much grander came about: St Pancras would be comprehensively rebuilt for the Channel Tunnel Rail Link so that Eurostar trains could use the station. In the process, two new underground platforms would be provided for Thameslink services under St Pancras, replacing the inconveniently-sited and impossible-to-enlarge King's Cross Thameslink station. Just north of the new station, a flying junction with a connection to the ECML would be built - not an easy task given the incredible honeycomb of tunnels in the area, with six London Underground lines serving King's Cross-St Pancras.

In spite of the fact that the Thameslink programme still hadn't been finally approved, the construction at St Pancras was given the go-ahead as part of the Eurostar upgrade, with Thameslink services split in two for nine months while the new station was constructed. Thameslink duly moved into its new home at St Pancras in December 2007, with King's Cross Thameslink closing at the same time. The tunnels for the flying junction were bored, but left without track for the time being: they won't be needed until 2018.

At Farringdon, the platforms were only long enough to take 8 carriages: extending them to the north was impossible, due to the severe gradient. Extending them to the south, however, would entail removing the flat junction to Moorgate. From the point of view of improving Thameslink services this was actually a benefit: without the trains to Moorgate, there would be more room for trains to Blackfriars and beyond. Passengers for Moorgate would still be able to change at Farringdon and get the Metropolitan line to their destination.

At Blackfriars, there were five platforms, all only long enough for 8 carriages; but three were bay platforms, only accessible to trains terminating from the south. Unfortunately the bay platforms were to the east of the through platforms, meaning any terminating trains had to cross the path of Thameslink services (another flat junction). Ideally the bay platforms would be to the west of the through platforms; to achieve that, the plan involved extending Blackfriars considerably to the south, to permit the through lines to be slued across.

In fact, extending Blackfriars to the south came up across a large obstacle: the River Thames. The platforms already encroached a little onto the bridge carrying the railway over the Thames; the Thameslink Programme called for them to be extended all the way across the river. Indeed, the plans included a new entrance to Blackfriars station on the South Bank, with the station having four platforms - two bays on the west side, and two through platforms on the east side.

Phase 1 of the Thameslink Programme, begun in 2009 and finished in time for the Olympics in 2012, involved shutting the Moorgate branch and the bay platforms at Blackfriars and diverting those services through the core (increasing peak services to 15tph) to facilitate the construction works at Farringdon and Blackfriars. In addition, the line was completely resignalled to permit trains to run much closer together. The works were not without disruption, requiring nearly three years of weekend closures through the Thameslink core.

But the disruption there pales in comparison to that caused by Phase 2: begun in 2013 and due for completion in 2018, Phase 2 is focussed almost entirely on London Bridge. Currently, Thameslink trains currently have to share tracks with trains to and from Charing Cross, and then cross over to the line towards East Croydon in a double flat junction. Prior to the works, the layout looked like this:
London Bridge track layout in 2009
London Bridge track layout in 2009

By 2018, Thameslink will have its own dedicated platforms and tracks through London Bridge, and the layout will look like this:
London Bridge track layout in 2018 (provisional)
London Bridge track layout in 2018 (provisional)

(To see clearly exactly what's changing, try opening both and flicking back and forth between the two.)

As can be seen, Thameslink had pretty much no track of its own through London Bridge. In a hugely complex plan, two extra tracks will be constructed between Metropolitan Junction and London Bridge for Charing Cross trains, with Thameslink taking over the existing two tracks. The lines run over the top of Borough Market, and finding space for the extra viaducts has been an absolute nightmare: the formation for the new tracks was put in place in 2011, involving one of the most complicated bridge slides ever done (see these pictures).

A couple of miles out at South Bermondsey, a flyover will be constructed to permit Thameslink trains to jump over the Charing Cross trains without interrupting them. This involves completely reconfiguring which track is which, and will not be an easy task.

Crucially, though, there aren't enough through platforms at London Bridge for those lines on their own to actually be any use: there were 6 through platforms and 9 bays for services terminating from the south. That will be turned into 9 through platforms and just 6 bays, through a massive gradual reconstruction; only three platforms will be closed at any one time, and a full peak service will continue to run throughout (though some trains will not call at London Bridge during the works).

Over the course of 2013 and 2014, the nine old bay platforms will be shut and the six newly-repositioned bay platforms opened in their place. From January 2015 to August 2016, no trains to and from Charing Cross will call at London Bridge, to permit the new through platforms 6-9 to be constructed. From then until early 2018, no trains to and from Cannon Street will call at London Bridge, permitting the rest of the through platforms to be reconstructed. Notably, only five of the six existing through platforms will remain: platform 1 will be removed, permitting all the platforms to be straightened somewhat (they are currently quite curved).

Perhaps the most annoying thing during the work will be the diversion of Thameslink services. As discussed previously, there isn't enough capacity for Thameslink services to serve London Bridge in the peak, and so instead they run via Elephant and Castle. But for the three years from 2015-2018 when the through platforms are being reconstructed, no Thameslink services will run via London Bridge at all, even in the off-peak.

But once the rebuilding is complete, Thameslink trains will be able to run through London Bridge all day, with up to 18tph serving London Bridge (with the rest going via Elephant and Castle). Beyond London Bridge most trains will continue to head south towards East Croydon and on to various destinations off the Brighton Main Line, but the capability will be there for trains to head towards Kent as well (with a flying junction at Bermondsey). Once London Bridge is completed, the connection to the ECML will finally be opened for use and Thameslink trains will serve a wide variety of destinations.

You might notice I'm being slightly coy about exactly where it's going: that's because it hasn't really been decided yet. While the infrastructure upgrades for the core are clearly defined, the end network is not (unlike Crossrail). Describing the probable destinations, and the politics involved therein, would be another article in itself; I'll leave it to the good people at London Reconnections, who have a fascinating article on the possible destinations of Thameslink post-2018, which I highly recommend.

In the meantime, though, we are right in the middle of a hugely exciting but also horrendously disruptive upgrade of London Bridge, that will unlock the potential of Thameslink in a way that this country has never seen before. The London railway network post-2019, once both Thameslink and Crossrail are operational, will be quite a different place, and I look forward to it all coming to fruition. To keep track of what's going on, you can check the Thameslink programme website.

Both Crossrail and Thameslink will achieve their 24tph through brand-new fleets of trains running with Automatic Train Operation (ATO). What is ATO, I hear you cry? Head on over to the next post...

Previous post: What is... Thameslink?
Next post: What is... Automatic Train Operation?

Thursday 1 May 2014

What is... Thameslink?

Thameslink is a railway line connecting north and south London via St Pancras, Farringdon, Blackfriars and London Bridge. Think of it as a north-south version of Crossrail, but it's a bit old and naff by comparison. As such it's undergoing a huge £5.5 billion upgrade called the Thameslink Programme, lasting nine years, that will effectively triple Thameslink's capacity.

From when it opened in 1988 until 2009, the core section of Thameslink only got 8 trains per hour (tph) in each direction in the rush hours. By contrast, Crossrail will get 24tph in each direction (in rush hours) from the day it opens. But then Thameslink was shoehorned into an existing network not remotely designed for it: two completely separate networks north and south of the river were combined into one, meaning 8tph was the limit.

But let's start from the beginning. First, here's a map of the central section of Thameslink:
(Map based on OpenStreetMap; © OpenStreetMap contributors)
The main Thameslink line is in blue; the now-closed lines to Moorgate and Holborn Viaduct are shown in green, and the connection to the ECML which will open in 2018 is shown in yellow.

Prior to 1988, suburban services into London on the Midland Main Line (MML) from Bedford and Luton ran to St Pancras and Moorgate. A few of them went to the high level station at St Pancras, rebuilt a just few years ago for the arrival of Eurostar trains. But most used the the "City Widened Lines", a remnant of the old Metropolitan Railway, to go underneath St Pancras to Farringdon and Moorgate, which was much more convenient for the multitude of City commuters (hence the name). While trains to Moorgate couldn't call at St Pancras, they called instead at King's Cross Thameslink, a cramped pair of platforms just east of King's Cross proper on the City Widened Lines.

South of the river, most of the suburban services from Kent and Sussex used a variety of stations: about half went to Victoria; the other half went to London Bridge and on to either Charing Cross or Cannon Street. Just a small handful of trains used the two stations at Blackfriars and Holborn Viaduct, both lying on the west edge of the City and mostly eschewed in favour of Cannon Street by City commuters.

But connecting Holborn Viaduct and Farringdon, there lay the disused Snow Hill tunnel, built in 1866 to connect the Metropolitan Railway to the lins south of the river, but closed in 1916 to all but a handful of goods trains. The Greater London Council under Ken Livingstone had been campaigning for some time for its reopening, but initially that meant trying to get two completely separate regions of British Rail to work together.

In the 1980s, however, the railway was restructured: rather than five regions of BR, new "sectors" were created. Most importantly, Network SouthEast took over responsibility for all commuter services across London and the south-east, giving the railway network one voice for commuters. And Chris Green, as head of NSE, gave his enthusiastic backing to Thameslink: it symbolised perfectly the uniting of London's commuter routes into one coherent network.

In 1988, the new Thameslink service began, with trains linking Bedford and Brighton, Luton and Purley, and Cricklewood and Sevenoaks. The service was somewhat tentative at first: the off-peak service consisted of just 6tph, and they ended up clumped so that there were some 20-minute gaps. The morning rush-hour actually had fewer trains through the core, because Thameslink was merely added on top of existing rush-hour service patterns, rather than replacing them.

For example, while many of the Midland City services from Bedford and Luton now ran through the Thameslink core to Blackfriars, London Bridge and even all the way to Brighton, there was too much commuter demand to Moorgate to just close it outright and force everyone to use Farringdon. As such, Bedford-Moorgate services remained a feature of the timetable for decades to come, at first just in the peaks, but within a few years a half-hourly all day service had returned to Moorgate.

On the other side of the river, Blackfriars and Holborn Viaduct similarly continued to have rush-hour trains which terminated there and didn't run through the core to Farringdon and beyond. There was, however, a key difference between the two: Blackfriars had platforms for the through Thameslink trains to call, but they skirted Holborn Viaduct without calling, easily visible from the end of the platforms.

Having two under-used terminals was now rather unnecessary, and the decision was taken to close Holborn Viaduct and divert the remaining peak trains to Blackfriars or through the Thameslink core. But just closing Holborn Viaduct outright would lead to the Ludgate Hill area lacking any kind of railway station; it lay in the no-man's land between the Circle and Central lines, unserved by the Underground at all.

So, in an audacious plan, Network SouthEast decided to bury part of the Thameslink line just south of Holborn Viaduct - in the Ludgate Hill area, at the east end of Fleet Street - and provide a replacement station on Thameslink, underground between Farringdon and Blackfriars, so that Holborn Viaduct could finally be put out of its misery, releasing the land for valuable City development. This required ripping up the newly-reopened Thameslink line, digging out a tunnel "box" for the station, and then covering it over - not an easy task.

Holborn Viaduct station was closed on January 26th 1990, consigned to the history books. Over the next four months, the railway around Ludgate Hill was completely rebuilt. Incredibly, services continued to run through the Thameslink core during construction works, with just a two-week closure in May 1990 to finish everything off. The new station, originally St Paul's Thameslink but renamed City Thameslink soon after, opened on May 29th 1990.

By far the biggest problem for Thameslink, however, was the lack of capacity through London Bridge. With well-established commuter networks ferrying passengers by their thousands into Cannon Street and Charing Cross every weekday morning, all of which had to pass through the seven through lines on the north side of London Bridge station, rail bosses were unwilling to reduce existing services to make room for the new Thameslink services.

It's not hard to see why: the regimented service of trains arriving at two-minute intervals into Charing Cross has barely changed since the mid-1970s, principally because it works so well at getting commuters into London. Reducing the service would only serve to make overcrowding worse; and since any improvement in Thameslink services had to be matched by a corresponding decrease in services to Charing Cross, the decision was taken to give the incumbent services priority over new ones.

So from the outset, and to this day, Bedford-Brighton services ran via London Bridge in the off-peak only, when there is room to thread them through the Charing Cross services. In rush hours, the same services ran instead via Elephant and Castle, avoiding London Bridge entirely. Operationally this works reasonably well, but it deprived Thameslink commuters access to one of the biggest London stations.

In spite of all that, however, Thameslink services were well-used from the outset: those places lucky enough to get a Thameslink service saw a quadrupling in usage within just a year of opening the line. The destinations changed a number of times, but by the turn of the millennium the off-peak service had settled down to 4tph Bedford-Brighton and 4tph Luton-Sutton, the latter via the Wimbledon loop.

Indeed, Thameslink ended up becoming something of a victim of its own success: with no capacity available to increase peak services, and the platforms at Farringdon and Blackfriars only long enough for 8-car trains, something big had to happen if Thameslink was ever to solve its overcrowding problem.

Enter the Thameslink Programme: a comprehensive upgrade programme, designed to remove the capacity bottlenecks along the route and give Thameslink capacity of its own, obviating the need to share capacity with other services. Such is the extent of the upgrade that it deserves its own blogpost: click here to continue reading about the Thameslink Programme.

In the meantime, if you want to read more into the creation of Thameslink, the Network SouthEast chronology is a year-by-year history of NSE with many fascinating tidbits; indeed, the whole site is an invaluable reference.

Previous post: What is... Crossrail?
Next post: What is... the Thameslink Programme?