Opinion Travel & Transport

After the MTR test-run crash, how can we trust computers to keep us safe on trains?

Brighton, the south coast town to which I used to resort for fun and diversion when I lived 20 miles up the road, has few claims to historical fame. The Prince Regent slept there, and rarely slept alone.

But in 1862 it was the scene of a historically important accident. Trains leaving Brighton in the general direction of London have to face a long climb up the South Downs, before they reach the Clayton Tunnel, which gave, and still gives, access to the more easy-going scenery of the Weald.

In the 1860s trains operated on the “time interval” system. A train leaving the station would be given a five-minute start before another one would be sent after it. And that was it. The trains were quite slow.

Clayton Tunnel

Clayton Tunnel. Photo: Barry Marsh.

On the day in question three trains left the Brighton station and headed for the tunnel. Train number one passed the signal box before the tunnel and entered the portal. The signal at that point should have automatically moved to “stop”. It did not. A gadget was provided to warn the signalman that this important event had not occurred but by the time he realised what was going on train number two had passed the signal. So he waved a red flag from his window as the train went by.

Minutes passed. Train number three appeared, and waited at the signal which he had manually moved to “stop”. At this point the signalman made a serious mistake. He sent a message to his counterpart at the other end of the tunnel asking if the tunnel was clear. This gentleman, having just seen train number one storm past his box, replied that it was.

So our signalman duly sent train number three into the tunnel. Unfortunately, the driver of train number two had seen the red flag. He had stopped his train, a procedure which took some time in those days. And then he started reversing back towards the tunnel entrance.

The resulting collision inside the tunnel caused 300 casualties. As generally happened with coal-burning engines and wooden carriages the wreckage caught fire. Rather surprisingly only 23 people died but the railways were in their infancy – in its day this was the worst accident ever.

The accident achieved instant fame because it was the inspiration for a popular ghost story by Charles Dickens, “The Signalman”. The tunnel is still rumoured to be haunted.

More seriously, in consequence of the crash the time-interval system was rapidly abandoned, and replaced by what was known as the “absolute block” system. Under this the track is divided into sections called “blocks” and the signalling system is set up so that only one train is ever allowed into a block at a time.

Of course, this did not prevent all accidents, or even all collisions between trains running on the same track. Human beings can be very ingenious, or very careless.

Times have changed, signal boxes have disappeared, equipment has become more sophisticated and knowledgeable. But still the block system has been a fundamental feature of safe railway operation for some 150 years.

MTR crashed trains

MTR crashed trains. Photo: Hong Kong Fire Services Department.

So I was a little disturbed to read, after the latest MTR incident, that the corporation was testing some software which would abandon this fundamental principle, and allow the trains to chase each other round the network, saved from disaster only by an omniscient computer.

I understand that this is a way of increasing the capacity of the system, but the latest accident was on what is known as a scissors crossing. Most of the time an MTR train is following another MTR train and has another following behind it. The worst likely problem is that one driver will have to brake sharply to avoid his predecessor.

A scissors crossing has more exciting possibilities, because it is a junction where trains can go in one direction or the other but not both. You can, as it were, have a train going Route A-to-Route B or a train going Route B-to-Route A. If two trains attempt this simultaneously they will collide.

Unfortunately, a network like the MTR needs a lot of these. If a train arrives at the end of the line in, say, Kennedy Town driving on the left it cannot simply go back the way it came. It needs to switch to driving on its new left. This can be done in a number of different ways but the easiest and cheapest is to have a scissors crossing just outside the station so that trains can go from the in-bound track to the out-bound one either before or after visiting the station.


Photo showing a train crashed into another train on the MTR system panel. Photo: Facebook/Sam Yip.

Clearly, this is a problem from a signalling point of view, because this gives us eight possible routes through the resulting junction (four in each direction) of which two pairs do not conflict with each other and four involve a possible collision.

It would, I suppose, be surprising if we could not find a computer somewhere capable of managing this tricky situation. But because of the consequences of error, it will have to be extremely reliable, which rules out any of the computers of my acquaintance.

One wonders why they did not avoid this complexity by separating the cross-overs. The scissors looks good on a model but it is really just two cross-overs super-imposed. The left-to-right one doesn’t have to be in the same place as the right-to-left one. Perhaps that would be more expensive.

Well I suppose they know what they are doing. It is nice to hear that the MTR is trying to improve things. You have to wonder, though, if this is a good time to introduce a spiffy new system when they seem to be having so much trouble operating the present one.

After the MTR test-run crash, how can we trust computers to keep us safe on trains?