For the past six years, the national rail public performance measure (PPM) has fallen steadily each year from 91.0 per cent in 2013 to 85.6 per cent in 2018. According to Transport Focus, the three main causes of passenger dissatisfaction are this fall in punctuality, increased fares and not being able to get a seat on a train. These are also the reasons why many consider that the current railway structure isn’t working and, indeed, some feel that nationalisation is the answer.
Against this background, the government commissioned the Williams review to consider the future structure of the industry. Although there is a political desire to change things for the better, improving punctuality requires much more than restructuring.
Each year, the UK railway network carries 1.7 billion passengers. Historically, the last time such numbers were carried was in 1920, when the railway network was twice its current size. To carry such numbers each day, UK rail operates 22,000 services. The complex interactions between these services on a crowded railway allow little time for service recovery. As a result, 70 per cent of all delays are now reactionary. Running a punctual railway has never been so difficult.
To enable potential researchers to familiarise themselves with the available data, RSSB created a ‘data sandbox’ which included datasets from various organisations as shown below. The intention is to make these available as a long-term industry resource soon.
- Attributed Delay Data;
- Performance Metrics;
- Network Rail Open Feeds including SCHEDULE (daily extracts and updates of train schedules), MOVEMENT (train positioning and movement event data), TD (train positioning data at signalling berth level), TSR (Temporary Speed Restrictions), VSTP (Very Short-Term Plan), RTPPM (Real-Time Public Performance Measure) and Train Planning Network Model;
- TD (train describer) data from Dec 2016 – May 2017;
- TRUST data from Dec 2016 – May 2017;
- GPS feeds;
- Upon request station, line speed, and tonnage data.
Rail Delivery Group (RDG)
- Darwin – real-time arrival and departure predictions and platform numbers;
- Knowledgebase – National Rail Enquiries database;
- Online Journey Planner;
- LENNON – ticketing and revenue database;
- National Rail Enquiries (NRE) data feeds.
Various train operators
- Genius – diagrams and allocations of trains data;
- Bugle – description and cause of delays;
- On-Train Data Recorder – station dwells and journey events;
- Traffic Management System data;
- Train describer data;
- Nexla and Orbita – train health, door opening/interlock times and energy consumption data;
- Web Gemini – train formation data;
- Passenger numbers – airbag and passenger count data;
- Reservations/ ticket sales.
- Unit movements data;
- Driver compliance System Retrieved Data;
- Warning Systems Data;
- Visual Cab 1 screenshot.
Transport Systems Catapult
- Mapping Grids (upon request);
- Mobile network data (upon request);
- National roadworks data (upon request);
- Haulage journeys data (upon request).
- Weather data
For these reasons, it could be said that poor punctuality that resulted in 15 billion delay minutes last year is an inherent feature of today’s railway. Yet something must be done. This is certainly the view of Network Rail’s new chief executive. Andrew Haines, who is committed to putting passengers first, has placed greater emphasis on train operations and is introducing a regional organisation to bring decision-making closer to customers.
To improve train performance, a National Task Force has been set up which brings together passenger and freight operators, Network Rail, the Office of Rail Regulation and the Department for Transport. The work of this task force has three overarching themes: better timetables, better assets and better operations.
An important aspect of improving operations is ensuring that rules for disrupted working are fit for purpose. This requires them to take account of modern communications and relatively new failure modes such as axle-counter failures. Rules also need to consider the overall system risk and so should not regard a stationary train as the safest situation, as crowded trains stopped for a long time introduce their own risks such as passengers evacuating themselves and crowded platforms. For this reason, slicker methods of degraded working are required.
This review of operational rules is one of the workstreams of the Enabling Better Network Performance Research Challenge (PERFORM) which is a cross-industry initiative led by RSSB. The other aspects are rail operations and variability (such as dwell time), understanding performance trends, managing disruption and getting value from the enormous amount of operational data that is generated each day. This was the subject of a £500,000 call for research in October for which a data sandbox was made available to interested participants.
The PERFORM programme was launched at RSSB’s recent “Enabling Better Network Performance” conference, which was attended by 150 delegates from industry, academia and the supply chain. In the opening keynote address, Tim Shoveller, then managing director of Stagecoach’s Rail Division (now managing director of Network Rail’s new North West and Central route), emphasised that the unprecedented performance challenges faced by the industry could only be solved by collaborative working. He was followed by Justin Willett, RSSB’s professional lead for operations and performance, who explained the PERFORM programme’s background, structure and governance.
The conference’s solution sessions included presentations from the five industry/academia teams that had been granted research funding from the October data sandbox research competition to develop novel data-driven solutions to improve network performance. There were also reports on other operations initiatives. After a discussion on how the industry should work together to improve performance, a further data sandbox research competition was launched.
Of the five research projects granted data sandbox funding, three concerned the impact of reactionary delays and two explored how machine learning could reduce station dwell times, which can be subject to wide variations. However, as these are generally less than the three-minute threshold, they are not usually monitored, even though they can have a significant impact on train performance.
The software being developed by the University of East Anglia, with support from Greater Anglia, will forecast how trains on the network are likely to be affected by current events and takes account of consequential impact on train crew availability. This will be used to help train controllers determine the knock-on effects of primary delays.
The development of a decision support tool using neural network technology to model reactionary delays is the research project led by Liverpool John Moores University, in collaboration with Merseyrail. The third reactionary delay project is a method to visualise the cause and consequence of knock-on delays under different scenarios to understand the delay dependency between locations. This is being developed by City, University of London and Risk Solutions, with support from Great Western Railway.
Using machine learning to analyse train performance data to the second is the aim of a project led by Middlesex University, with support from Southeastern. This is integrating the vast amount of available raw data to model train operation that will provide useful information to engineers and operators to enable them to act to reduce delays.
Artificial intelligence is also being used by a team led by University of Southampton, in collaboration with South Western Railway. It is using a range of data sources to develop a real-time visualised alert system which could identify unexpected sites that could be targeted for mitigation measures.
An interesting point raised in the discussion about these initiatives is the impact on passengers from actions taken to recover the service, which can include skip stopping and terminating services before their end destination. As such actions can cause significant disruption to some passengers, minimising train disruption does not necessarily minimise overall passenger disruption. However, the conference was advised that the research to determine the best operational strategy to recover from service disruption is not considering the impact on passengers adversely affected by actions to recover the service.
ADCI, RAATS and T1135/54
Other performance improving projects described at the conference were automated driver competence indicators (ADCI), considering red signal approaches, improving operational decision making and planning for disruption.
The ADCI project is being trialled by LNER and c2c and is based on data analysis by the University of Huddersfield using software developed by Cogitare that is now ready for industry roll out. It aims to use on-train data recorders to assess driving technique in respect of safety, energy consumption and punctuality. It will provide an app to enable drivers to assess their own performance and enables targeted support to be provided to individual drivers. The project will also identify common performance issue along the driver’s route.
The industry has done much to reduce signals passed at danger (SPADs) which, until recently, were normalised by train miles. A more meaningful approach considers how many red signal approaches result in SPADs. To facilitate this, the University of Huddersfield has developed the web based RAATS tool (Red Aspect Approaches To Signals) which uses the train describer data available under Network Rail’s open data initiative. As well as improving SPAD analysis, RAATS provides valuable performance data by, for example, showing where red signals routinely delay trains.
Supporting front-line operators, who often have to make real time decisions based on incomplete information, is the purpose of RSSB research project T1135 which has developed the G-FORCE decision making tool which is named after the steps it involves: G- go or no go?; F- facts; O- options; R- risks; C – choose and E – evaluate.
Another RSSB research project, T1154, considered ways of planning for disruption. This has developed a best-practice toolkit which is being tested by Greater Anglia, GTR and ScotRail. It considers four levels of disruption, five defined phases of disruption, decision making processes, the overall management of contingency plans and the processes and training needed to support them.
The next part of the conference considered various system and processes underpinning the ongoing performance initiatives.
One such enabler is ITED (Industry Train Event Data). In his presentation, Dominic Medway, Network Rail’s operational performance and analysis manager, explained how the ITED will enable to-the-second analysis of all aspects of operational performance. He advised that ITED is expected to go live in late summer 2019.
Crew and Stock systems were the subject of the presentation by Andrew Graham, who is the digital railway operations support for the Rail Delivery Group (RDG). This highlighted the range of systems currently in use which include verbal communications and pen and paper as well as digital systems. As many systems are not interconnected, changes are advised in an ad-hoc manner and, with continually changing demands, it is difficult for operators to keep track of crew and stock alterations and to share information with each other.
To address these issues, RDG has, following cross-industry consultation, recently published a concept of operations for a common crew and stock system which needs to be further developed to operate with Network Rail’s traffic management systems (TMS).
The important of this requirement was reinforced by Jonathan Scott, project director for Network Rail’s Digital Railway programme, who made it clear that TMS requires strong operational input. Jonathan’s presentation concerned the lessons from the first TMS deployments. These are the Thales Aramis system which, was deployed in Wales in March 2019 and is about to be introduced in Anglia, and the Resonate Luminate system that went live on the Paddington to Bristol route in June 2018.
He considered that there were positive indicators of operational benefits from this early use of TMS, especially in the identification of timetable anomalies, and that the biggest benefit has been the lessons learned for the deployment of other TMS, especially management of operational and business change inputs.
Data Sandbox Plus
A further funding opportunity for data-driven operational research was explained by RSSB’s senior partnerships and research manager, Giulia Lorenzini. This data sandbox plus call for research aims to build on the experience of the 2017 data sandbox research and seeks solutions to the following key challenges:
- Predicting and minimising operational delays;
- Understanding train movements;
- Reducing dwell time variations;
- Management of disruptions;
- Better measurement and understanding of performance and delays;
- Any other challenges identified by relevant organisations.
RSSB is encouraging the feasibility and demonstrator projects, for which funding from RSSB of respectively 80 and 60 per cent is available. There are two rounds of applications for which the closing dates are 5 July and 6 December with the winner to be announced in August 2019 and January 2020.
The “enabling network performance conference” certainly made it clear that infrastructure, trains and their passengers generate a vast amount of data. Examples are: each time a signal changes or a point moves, each time a train starts, stops, passes a signal or its doors open, each time someone buys a ticket or goes through a ticket gate.
The challenge is how best to use all this data. The five winners of the original data sandbox competition provided some of the answers. It will be interesting to see what solutions will come from the further research projects to be funded by data sandbox plus.