The Victoria Line: Bathtubs and Breakdowns
Those travelling on the Victoria Line over recent weeks will have noticed that the line has experienced more than its regular share of disruption. Both the Evening Standard and the BBC have covered the issues but it is probably worth us giving a bit more background here as well (not least because the Standard’s article includes some numbers that are rather disingenuous).
Although London Underground have remained relatively tight-lipped on the more specific details, various sources seem to agree that the delays and breakdowns have at least in part resulted from issues affecting the new 2009 Stock units that have now entered service. Two specific areas have been highlighted so far as contributing to these failures – software issues and issues relating to sensitive door-edge activation.
In a sense, neither of these two factors has been entirely unforeseen. The 2009 Stock has been test-running for a while now, and both issues received a brief mention at TfL’s Rail and Underground Panel at the beginning of July:
Most of the incidents affecting the new fleet have been software issues and sensitive door edge activation, and have not caused major disruption to the line, which has operated over 97 per cent of its schedule
These recent failures have certainly highlighted that these issues still need to be fixed and indeed may well be bigger problems than had been anticipated. That the failures have manifested in rush hour (when Victoria Line trains become metaphorical sardine tins and excess pressure on doors during operation is more likely to occur) certainly seems to suggest that the issue with sensitive door edge activation is greater than previously thought.
Given the issues, it would be easy to suggest that the failures represent a wider failure with the £900m Victoria Line Upgrade, or with the new Rolling Stock itself. To do so, however, is probably unfair.
Talk to an engineer – be they of the digital or physical persuasion – about reliability and one phrase that will swiftly surface is that of the “Bathtub Curve.” The Curve, broadly speaking, is a convenient way of representing the overall reliability of a product or asset family throughout its life cycle – that is, not an individual product, but all of them.
The image below shows a basic visualisation of the curve – and indeed from where it got its name – plot out the product failures against the time that they happened within the family’s life cycle and it resembles the cross section of a bathtub.
The Curve can be (and often is) applied to a wide range of items – be it software, superconductors or (as is the case here) rolling stock. It can also be broken down broadly into three distinct time periods – early operation, normal operation and late operation. During each period, a number of factors affect reliability to a greater or lesser extent, ultimately contributing to the failure distribution that gives the Curve its distinctive shape.
It is the first of these periods that the 2009 Stock currently occupies. New in service, it is highly prone to “infant mortality” – failures relating to previously undiagnosed bugs, issues or problems with the design or manufacture. As time progresses, these issues are diagnosed and corrected, and the reliability of the family improves until it remains relative low and consistent during normal operation.
Indeed it is infant mortality that makes the Evening Standard’s coverage of the failures rather dubious – featuring, as it does, the headline:
New Victoria line trains ’23 times less reliable’ than the old ones
and within the article itself:
Transport for London has now admitted both problems were caused by the new trains, which have proved 23 times less reliable than the 43-year-old trains they are replacing
The origin of this particular statistical bombshell is easy to trace – the Rail and Underground Panel report linked to above also contains the following line:
Since the beginning of the year, the 2009 Stock has achieved 600km mean distance between failures, while the 1967 stock has achieved over 14,000km.
14,000 divided by 600 gives you 23 (give or take a decimal point).
Comparing those failure rates in such a way is, however, less than useful and (at worst) misleading. The 1967 stock has been in service for some time as can probably be guessed by the name, and whilst it is now entering the third stage of the bath tub curve (where wear-and-tear becomes an issue) it is far from not dead yet. Indeed London Underground’s well earned world-wide reputation for being able to sweat its assets (mostly born, unfortunately, out of necessity) means that it will not remain a fair comparison for some time to come. Indeed even comparisons with the 1967 stock’s own infant mortality rates are probably not entirely useful, due to the time distance that seperates the two.
Asking the Right Questions
Whilst it is important, therefore, to draw attention to these failures, it is important not to get too quickly drawn into conclusions over the reliability of the new rolling stock based solely upon them. In fact doing so diverts attention away from better questions that should be asked – not least because the Victoria Line is not the only line which will shortly see its rolling stock updated. The new S-Stock will soon begin to make its presence felt on the sub-surface lines – indeed there are rumours that it may even make a brief appearance this weekend – and both the Bakerloo and Piccadilly will (finances allowing) also see their rolling sStock updated. In all cases, these failures on the Victoria Line may lead to lessons that could be applied elsewhere.
The first of those questions is whether sufficient testing took place with the 2009 Stock before it entered passenger service, as extensive testing is the best antidote to infant mortality for obvious reasons – it catches bugs and problems before they manifest in real-life situations.
Again, however, this is a question that likely doesn’t have a simple answer. The 2009 Stock has certainly seen its fair share of testing already, and testing can never identify all failures because it is never a 100% accurate simulation of real-life behavior – just as Apple’s testing process failed to identify aerial issues with the new iPhone 4, it appears the 2009 Stock testing (unsurprisingly) could not truly emulate a full rush-hour passenger service pattern.
Secondly, given that infant mortality is indeed inevitable, it is worth asking whether the systems and processes in place to deal with it are adequate and how they will help reduce such failures as quickly as possible. Again, this is a question that may require a complex answer but it is also one that is well worth asking – because ultimately it is this that has the greatest impact on keeping the passenger experience on the Underground as good as possible through a difficult (and, given the current weather, sweaty) transition.
Living and Learning
Overall, therefore, it is important to take the recent failures on the Victoria Line in context – that is not to say that it is not important to ask questions as to why they occured, but that it is important to make sure that the right questions get asked. Problems during this period of change are certainly inevitable, but it is also important that they are minimised. The correct questions can help to ensure that the minds of passengers caught up in those problems are eased and, where necessary, that changes are made to help minimise issues even more over the coming months and years.
Thanks to Greg Tingey and all on uk.transport.london for various spots and discussion points. Those wanting to know more about Reliability Engineering and the Bathtub Curve may wish to read Dennis J Wilkins article “The Bathtub Curve and Product Failure Behavior” here and here.