Compared to a road vehicle,
- most trains are orders of magnitude more massive;
- the rolling resistance (the internal friction that stops a vehicle left to roll out) is much less;
- while the friction coefficient of the steel-wheel-on-steel-rail contact is much less than that of the rubber-wheel-on-asphalt contact.
Consequently, braking distances are necessarily long, and eventual collisions involve incredible energies. All this makes the safe stopping of moving vehicles the paramount safety issue on railways.
The most basic safe stopping system is signalling: drivers are instructed to stop with visual signals along the track. The signals are given by people or machines at points where the trains could meet upon another train or an obstacle, for example at the two ends of a station.
Stations can be far apart, so if you want to run trains more frequently, you need more signals. It's the simplest to place signals on the open line at regular intervals ("blocks"), with the distance chosen so that all trains using the line are themselves shorter and can stop within it: this is block signalling.
With the development of technology, this relationship turned upside down: train braking performance and speed limits were set thus that braking distances be shorter than existing block distances.
When signals are given and read by humans, there is room for carelessness and oversight. Engineers tried to create various automatic safety systems that bypass the signalman and the train driver, or at least ensure that they don't overlook something.
Most collisions and derailments happened at switches, where the paths of trains can meet. This risk was tackled at the signal side by creating a dependency between signal and switch adjusters: i.e. the signal for the track in the 'wrong' direction of a switch is locked in the stop position. The same way, successive block signals were connected. The first interlockings used actual mechanical locking, the state of the art is of course electronic, and allows for centralised (remote) switch control.
On the train side, the first control was the vigilance device: the driver has to regularly push a button or lever to confirm his consciousness, otherwise the train brakes automatically. As for other automatisms, many a chief engineer argued that they would lull drivers in a false sense of safety and reduce their vigilance. This argument is not without merit (nowadays many if not most accidents in Western Europe happen when safety systems are temporarily off for maintenance or due to malfunction), but a couple of preventable catastrophes convinced most major railways in Europe by the middle of the last century that automation is the lesser risk.
The first thing you want in terms of track-train interdependency is to control the passing of signals by trains. In both directions: the train should brake automatically when passing a stop signal, and a proceed signal should change to stop when passed by a train. Then you have automatic train protection (ATP). For this, you need to transmit information between track and train somehow.
The earliest, 19th century ATP systems used mechanical trips. Later, the idea came to use the closing of electrical circuits. To detect trains for the signalling, you can use isolated sections of the rails themselves. In the other direction, you need to create separate contact points, for example a short metal beam on the track touched by a brush on the passing locomotive (this is how France's Crocodile system works). It's even better though to have a contact-less means of information transfer. The simplest way is to use induction. Electromagnets are installed near signals along the track (called balises in modern systems) and on the leading locomotive/driving trailer of a train, information is transmitted as the latter pass over the former. (Early example: Germany's Indusi, modern examples: Germany's PZB90, France's KVB.)
On the left, on-board receivers for two different national ATP systems, both placed outside at the middle of bogies (for track magnets laid on the outside of the rails). Top: Switzerland's Integra-Signum, bottom: Indusi/PZB90 for Germany and Switzerland. On the right, an Indusi track magnet at the station of Payerbach-Reichenau, Austria (the two bogies actually belong to the same locomotive, which is shown in full at the bottom of the diary)
If the train already gets the signal information, a logical addition to automatic train control is cab signalling. The actual signals along the track are often obscured when the track is curvy, and are the easier to miss the faster the train goes, thus cab signalling makes a difference from around 120 km/h.
With all the previous automations in place, something else becomes possible: train dispatching from a control centre, which can adjust both signals and switches by remote control.
Block signalling with an ATP ensures that trains stay at a safe distance. However, it is inefficient. Let me demonstrate a worst-case scenario with the following diagram:
A locomotive on a single run almost reached signal A, which shows to proceed (green). The last signal the loco passed was signal B, which thus shows stop (red). Let's assume that the trackside ATP system detects train location only with the balises next to signals. Then, it cannot distinguish a block-length freight train with a loco that just reached signal B (meaning the train's end just passed signal C), and a single loco that almost reached signal A. Thus, the next signal, signal C has to be locked in stop position, too. Let's further assume that block signals also serve as warning signal for the next signal (D is yellow to warn that C is red) and that this is a strict system where the driver has to start braking at the warning signal or else automatic braking is initiated. As a result, the next train following the single loco has to brake at almost three block distances behind the rear end of the loco! Worse yet, if the next train is an electric multiple unit (EMU), with efficient friction-less braking that allows it a braking distance less than half the block distance, then it is kept at six times the braking distance from the loco ahead of it! What's more, consider that just after the EMU started braking, the loco will pass signal A and signal C will change to proceed – but the ATP of the EMU (unlike its driver) won't get this information until the next balise, so the driver will have to complete the braking!
To summarize, in conventional systems, the distance between successive trains can get way above what's necessary for safety for the following reasons:
- train location is known only with block length precision,
- train length can be much shorter than block length,
- due to the intermittent transfer of signal information, the train protection system cannot react to signal changes when the train is between two signals.
There are some patches for intermittent ATP systems to get around the problems (multiple balises per block, infill loops etc.), but they aren't ideal and can get expensive. Also, all of the above capacity-limiting problems get more serious with increasing speeds: block distances have to accommodate ever longer braking distances. This triggered the development of continuous train control in the second half of the 20th century. The three basic innovations needed for these systems were:
- some means of sending signal information continuously;
- some means of determining train location with more precision (this can be the wheel sensor of the locomotive);
- last but not least: signals and balises don't give the binary proceed/stop information, but a target speed (between zero to maximum line speed).
Both the first and last can solve the signal change problem. The second gives some leeway in brake initiation. The last one provides for the possibility of block distances shorter than braking distances: trains don't have to go from top speed to zero within a block, instead successive block limits can instruct a step-by-step speed reduction.
Continuous train control systems on European mainline railways use one of two technologies:
- Some intermittent train control systems already use the (isolated) rails within a block as electrical circuit for the detection of trains. These can be used for information transfer in the opposite direction, too: by sending a (modulated) AC current into the rails, they will function as an antenna. (Examples: the Netherlands' ATP-EG, Hungary's EVM-120, and the much more advanced TVM430 system used on French high-speed lines.)
- You can also lay a pair of cables in the middle of the track to serve as antenna. The benefit is that by switching the cables at regular distances (100 m or even less), you have a trackside means of finer-than-block train localisation (the train counts the polarisation changes). There is one such system, the LZB system, which originated in Germany but is standardised by the International Union of Railways (UIC) (currently installed on high-speed and upgraded mainlines in Germany, Austria and Spain).
Crappy shot, but you can see one of the LZB cables along the middle of the track (as well as an Indusi track magnet). Shot at 200 km/h near St. Peter-Seidenstetten, along the quadruple-tracked section of Austria's Westbahn
Continuous train control is still not the ultimate in efficiency: if you can locate the end of all trains, and calculate potential braking distances continuously (in dependence of the train properties, the momentary speed, and the gradient and the weather conditions on the track ahead), then you can allow trains to always approach each other at the minimum safe distance. This is called a moving block system.
The two preconditions are relatively easy to implement on urban rail systems, where moving block systems already exist (example: CBTC). On mainline railways, there is no such system yet, only one semi-moving-block system: CIR-ELKE, an 'update' of the LZB system, which uses the cable switch points to create virtual blocks that are a fraction of the actual blocks (and shorter than the trains).
Although developed for higher speeds, these advanced systems also come in handy for capacity enhancement on busy conventional lines. CIR-ELKE is used on the Karlsruhe-Basel line along the Rhine Valley, which sees heavy freight traffic (see Corridors for freight), as well as on the central artery of Munich's suburban rapid transit network, where trains are scheduled at 2-minute headways (see Local Rail Extras I and this Salon comment); and the focus of the diary, the also continuous ETCS L2, is used on the Betuweroute, a new freight-only line all across the Netherlands to Rotterdam's port (again see Corridors for freight).
Going wireless: ERTMS, ETCS
A bit over two decades ago, a pan-European team initiated the development of digital wireless technologies for railways. This was later adopted by the EU, in its most important initiative to standardise European railways for seamless cross-border traffic. For train control, the main benefit of wireless would be saving on expensive and vulnerable trackside equipment (impacts from stuff falling or dangling from trains, rain, snow, floods and cable thieves are a constant headache).
Then modern third-generation wireless technology was first to be deployed for communications, as a replacement for the multitude of train radio systems (which are used for voice communication between locomotive drivers and dispatchers at stations or control centres). This is GSM-R (with "R" for "Railway"), the first element of the European Rail Traffic Management System (ERTMS). The other is a new ATP system, the prosaically named European Train Control System (ETCS). Ultimately, ETCS is intended to enable all the frills of train control, from vigilance device to moving block implementation, by using GSM-R for communication between track and train. However, we are far from that, and the technology is developed so that it can be implemented at different levels, which build on each other:
- Level 0 covers on-board equipment with (limited) compatibility with existing national systems. The idea is to transform the track information gathered by the receiver of the national system into ETCS data format, and send it to the ETCS on-board computer. This already can mean savings: in the locomotives of Siemens's new Vectron platform, it is possible to install the receivers of ten different systems simultaneously, but all can be connected to a single ETCS computer. Bombardier's four-system TRAXX locos also use the ETCS computer when running under LZB.
- Level 1 is an intermittent ATP: the basic trackside equipment is called Eurobalise, it is an electromagnet placed in the middle of the tracks, and can transmit a large variety of information. ETCS L1 is in operation on pilot lines in several countries, mostly upgraded mainlines.
- Level 2 is a continuous train control system: it still uses the Eurobalises, but chiefly as location markers only, and most of the train-track information exchange is via GSM-R. ETCS L2 is in operation on a dozen lines, all of them newly built ones, mostly high-speed.
- Level 3 is a moving block system, in which trains are located with high precision with the help of on-board equipment, balises aren't needed any more. At this stage it is a distant dream.
While a freight train leaves the station of Győr, in the foreground, you can see a pair of Eurobalises in the middle of the track next to a signal pole. The Vienna-Budapest line was one of the first equipped with ETCS L1
While train control standardisation looks like a great idea, there are three big practical problems:
- We are trying to introduce a newly developed system (in level 2, even an entirely new technology) in a field where reliability is a must. That means the system must reach a higher standard of stability than expected of most new technology, even issues discovered after years of operation will count as teething problems. That also means that the development of standards must come to an end.
- We are trying to implement the traffic rules of all European countries, with all the local specialities, be them ones developed for Arctic weather in the north of Sweden or ones developed for unstable DC supply in Italy. This universality makes ETCS necessarily more complicated than any existing national system, and the necessity of various additions may not be recognised for years.
- Last but not least, train control systems are expensive long-term investment, with amortisation times measured in decades. Replacing all existing national systems with ETCS before they get aged is not cost-efficient and prone to fall victim to budget cuts.
The focus of the diary is an example for the first problem. The Madrid-Lleida section of the Madrid-Barcelona high-speed line was supposed to go into operation with maximum speeds of 350 km/h under ETCS L2 supervision in 2002. However, not even the construction of the system was finished by that date, and a conventional system was installed so that the line could be opened a year late. ETCS was later commissioned at L1, permitting 250 km/h from May 2006, raised in steps to 300 km/h in the next 12 months.
As an example of the problems of complexity, in Belgium and the Netherlands, the opening of two new high-speed lines was delayed for two years (see Delays come to an end (EU HSR 2009)) and the aforementioned Betuweroute started up below capacity, because there were communication problems between track- and train-based equipment of different manufacturers, and the standards were updated, necessitating a costly retrofit. Additionally, when the operator of "Fyra" local services on the Dutch-Belgian high-speed lines was forced to deploy locomotive-pulled trains as interim service (ordered EMUs were – and are – late), the certification of the locomotives for both ETCS L1 and L2 was drawn out, and even with certification, the trains often ran late.
As an example for ETCS falling victim to budget cuts, there is everyone's favourite government in Germany, intending to delay ETCS deployment on transit corridors in spite of an international agreement for the Rotterdam-Genoa corridor and industry outcry. In the meantime, I read some more on the background, which underlines the cost-effectiveness problem. The original intention was to install level 1, but that would have reduced capacity on the section now under CIR-ELKE, so it was decided a year ago to go for level 2. Level 2, however, would have been very expensive, without bringing benefits over legacy systems, partly because incompatible signals would have to be replaced and partly because even the GSM-R network would have to be rebuilt. The latter is a result of being a pioneer of that system: presently beacons don't provide a very strong signal, and a frequency of the later built UMTS network can interfere. So Germany's transport ministry would prefer Level 0.
All in all, we have the strange situation that ETCS, by being state-of-the-art with multiple suppliers that can be made to compete, is a frequent choice on new or upgraded lines in Developing World countries, but it is nowhere near being the envisaged standardiser of European railways.
ETCS still getting going?
In spite of all the bad news, there are some signs that ETCS deployment won't grind to a halt, though it won't be as fast as envisaged.
One is the commissioning of ETCS L2 between Madrid and Lleida (with the rest to Barcelona to follow in December). On one hand, at the raised speed, this represents an application more demanding than for legacy systems (only TVM430 has been used regularly at speeds higher than the present 310 km/h, but this will be raised). On the other hand, when ETCS L2 trials resumed in 2008, the national railway (RENFE) and the track authority (ADIF) stated clearly that there is no target date and the system will enter commercial operation only if and when reliability is evidenced.
Meanwhile, two smaller countries with aged intermittent ATPs are moving ahead with a complete switch to ETCS. Switzerland – which has the best experience, with problem-free ETCS L2 operation at up to 200 km/h on two lines for four years now – will deploy a simplified version of ETCS L1 across the network by 2017, the contract was awarded last month. Meanwhile, two years ago Denmark decided to equip the entire network with ETCS L2 by 2021, with the GSM-R contract awarded last year.
For its part, Austria plans to have ETCS L2 on all its newly built lines and at least ETCS L1 on all pan-European corridor lines that don't have LZB. The first phase is to go on-line in 2013, with all modern locomotives and trainsets equipped for L2. Trials are on-going, and at a conference I attended two weeks ago, representatives of all manufacturers swore that everything is according to schedule. Due to the Austrian developments, Hungary's MÁV is also equipping all its modern locos for ETCS L2.
186 106, a quadruple-voltage loco from Bombardier's TRAXX platform belonging to Munich-based leasing company Railpool and normally running in the Rotterdam port traffic, on exhibition at Sopron in Hungary for the conference I attended two weeks ago. The loco was in Hungary for compatibility trials between its on-board ETCS and the trackside ETCS L1 equipment on the Budapest-Vienna line.
At this point, I think a comprehensive ETCS deployment at least on mainlines will take decades, but will come.
:: :: :: :: ::
Check the Train Blogging index page for a (hopefully) complete list of ET diaries and stories related to railways and trains.