Over the horizon - Aspects of automation
5 October 2017
Director of System Safety and Health, RSSB
Professor of Railway System Safety - University of Huddersfield
George Bearfield looks at risk - reality and perception - and how the industry needs to preserve its sound risk-based approach to safety as we consider a more automated future.
The risk-based approach to safety management in the UK is a beacon of successful safety regulation over the last half century, with steady improvements seen since the enactment of the Health and Safety at Work Act in 1974. To achieve the best safety outcome, safety risks must be evaluated and a rational approach taken to addressing them. Rail has benefitted as much as any industry sector from this enlightened approach and needs to continue this as we address future challenges.
Unfortunately, the current disputes on the safety of driver dispatch have become emotive and confused. In this article I set out a clear evaluation as to why driver dispatch of trains is unquestionably safe according to all criteria and what this tells us about how we need to address safety issues in the future, as more automated systems are brought into use.
How do we know if something is safe?
There are clear criteria for determining whether something is safe.
At the most basic level, regulation and case law sets out that the responsible party must estimate the risks, and determine the cost to reduce or eliminate them. Unusually for criminal law (with a growing threat of custodial sentences) there is a reverse burden of proof here. The responsible party must be able to show that it has taken all reasonably practicable steps to avoid or reduce the risk. The RSSB publication
'Taking Safe Decisions' provides the industry's agreed approach to undertaking this appraisal. Good practice and standards provide a 'short cut' to acceptable solutions where the issues and circumstances are well understood.
Although the setting of 'zero harm' as an aspiration and target is a good, forward thinking approach to driving improved safety culture, the legal framework sensibly realises that 'zero harm' is not currently achievable or realistic in all areas of risk. An honest, realistic approach to safety can therefore be applied, to ensure resources are used in the most effective way to benefit safety in the round.
How safe is Britain's rail?
Train operators undertake risk assessments and monitor all aspects of their operations including train dispatch. Systems are in place to manage risks, including monitoring, and investigating safety performance which are supervised and enforced by the rail regulator. This has been highly effective as seen by the relatively low safety risk in rail compared to other transport modes.
The infographic below shows the risk of being fatally injured per commute for a range of transport modes in Great Britain and for the average of the EU rail network. The fatality risk for each is proportionate to the area of each square.
Figure 1: Fatality risk in different transport modes (per traveller kilometre)
This infographic highlights two key things. Firstly, that the risk of being fatally injured in a commute is significantly less on the rail network than on any other land transport mode in Great Britain. Secondly, it is significantly less in Great Britain than when commuting on the EU rail network. GB rail companies are much less accepting of safety risk than any of their peers.
Is driver controlled operation a safe method of dispatch?
We can put the risk from train dispatch – by all methods - in context here by highlighting the proportion of passenger risk from it. To do this we need to use a compound measure of harm – Fatality and Weighted Injury – as dispatch related fatalities are thankfully very rare, so we need a richer source of data including major and minor injuries. This clearly shows that the risk from train dispatch is a very small part of the risk to which passengers are exposed on their rail journey (one fatality and weighted injury per 850 million journeys).
Dispatch-related risk as a proportion of total passenger risk
However, it is not the safety of train dispatch per se that has been the subject of such intense debate in recent disputes. The concern has been about
the difference between the risk from driver dispatch and guard dispatch. So what does the data tell us about that?
RSSB analysis (Figure 3) shows that the highest rate of more serious incidents occurred for guard dispatched trains from unstaffed platforms. However, given the high number of passenger journeys being taken, the real conclusion is that these rates are so low, that it is very difficult to provide any meaningful estimate of the difference. Put simply, any difference in risk that might exist is so small we it can't be estimated.
Figure 3: Number of fatal and major incidents involving passengers and public per billion passengers boarding or alighting
Are there other criteria for what risk is acceptable?
The level of safety required is ultimately determined by the regulatory framework. Broadly, the objective criteria the government use, for all sectors, is based on how the risks compare with other risks people are typically exposed to, known as 'tolerability'. According to the HSE document 'Reducing risks, protecting people':
"For members of the public who have a risk imposed on them 'in the wider interest of society' this limit is judged to be an order of… 1 in 10 000 [fatalities] per annum."
In other words, of 10,000 people engaged in that activity, one dies per year, which is like the background level of risk people are exposed to in their everyday lives.
The total fatality risk per year to a regular commuter due to all causes is estimated as 1 in 400,000 per annum. The train dispatch related risk (by any method) is approximately 1 in 6.7 million per year.
These risks are well below the level that the HSE define as 'negligible'. In other words, the risk associated with train dispatch (by any method) is a tiny fraction of the risk people face in their everyday lives, and far less than the chance of being struck by lightning for example.
So what has gone wrong in the understanding of this issue?
Industry decisions need to be logical, rational and consistent. But a decision made on robust safety arguments could be undermined by lack of public trust if the key facts are not understood by the public – a fact that is recognised within the 'Taking Safe Decisions' document. This has been the case with the debate around the safety of driver dispatched trains.
Also, the safety of driver dispatched trains has become confused with the very important, but distinct, issues of disability access and perceptions of public security which future articles as part of this series will consider.
The lessons for the future
Technology development is leading to increased automation. Electronic control systems and other initiatives such as train operated warning systems and obstacle detection at level crossings are all moving the rail industry in this direction.
Automation can help us overcome a range of challenges, for example unlocking capacity bottlenecks. When it comes to safety, automation will enable us to make the next step change by reducing exposure to harm, removing the variable element of human behaviour managing important controls, and making better use of our unique talents as human beings.
As an industry, we must continue to analyse data objectively and take timely, rational decisions. By applying this risk-based approach, we will continue to improve the health and safety on Britain's railway; and fear about safety should not be an obstacle for embracing innovation.
In the next article Michael Woods, Principal Operations Specialist R&D, and Claire Shooter, Senior Research Analyst, look at the role that automation can play in making the railway more accessible.