Jeremy Revell and Richard Wheldon (Frazer-Nash Consultancy), and Luke Allen (eviFile) on their digital solution to disruption
As the UK’s railway system grows and provides enhanced and further services, the potential for disruption on the lines also increases. Frazer-Nash and eviFile have worked in partnership to produce a digital solution to overcome this.
One of the main challenges facing the railway industry today is the complex process of possession management – whereby trains are unable to run on sections of track whilst they undergo maintenance work.
The logistics of diverting, blocking, or closing sections of track can have implications across the network – with delays on main lines likely to cause disruptions, as well as the obvious financial implications of delayed engineering works.
Is this really a difficult or large problem?
There are thousands of engineering operations in any given year. Taking Network Rail as an example, in 2020/21, the company spent £1.6 billion on enhancements, £1.9 billion on maintenance, and £3.2 billion on renewals across the rail network (Office of Rail and Road, 2021).
Possessions can span long distances, with many tens of miles being the norm, and usually involve multiple worksites with many contractors working in parallel. It also goes without saying that these works require safe, traffic-free sites for maintenance activities to be carried out competently.
Whilst possessions are planned carefully in advance, ultimately, their successful delivery relies on plans working in practice. Disruption can occur due to machine faults, access issues, staff planning, or incorrect engineering arrangements – all demonstrating the complexity of planning possessions.
The timely organisation of deploying staff and equipment to worksites and minimising travelling distances are critical efficiency requirements. Barriers such as mutual road and rail access points, staff numbers, and equipment types can hinder these works. Consequently, issues in both timetabling and ensuring that engineering trains reach worksites at the correct time, and in the correct formation, can often arise.
How is this managed at present?
Modern project management practices can help with possession management to some degree, but traditionally, solutions have been underpinned by paper-based processes. Such processes, especially those involving numerous worksites over a large area, do not help those in charge of work readily understand the optimisation of such projects.
The Small Business Research Initiative (SBRI)
Recently, Frazer-Nash and eviFile bid for, and won, a Small Business Research Initiative (SBRI) competition funded by the Department for Transport (DfT) in an attempt to solve this problem. Part of a larger ‘first of a kind’ initiative, the competition aimed to accelerate innovation in the UK rail sector by enabling technologies to be readily and efficiently integrated into the railway system.
Using artificial intelligence, the team developed a tool to help optimise the work undertaken in possessions to increase efficiency and reliability, with the ultimate goal of empowering the industry to deliver robust infrastructure, on-budget works, less disruption to rail users, and value for money.
One of the fundamental challenges when undergoing possession management projects is the collection of key data from track. High quality, accurate information must be taken from both Network Rail PICOPS and engineers on the ground to deliver efficiencies that can be applied across any major possession, and can deliver a real difference for the industry.
After delivering a real time possession management solution for Network Rail, eviFile and Frazer-Nash worked in partnership to deliver on the Small Business Research Initiative (SBRI) competition.
Frazer Nash’s machine learning and data science principles, along with eviFile’s deep insight and field data engine expertise, provided a unique platform to collate, manage, and provide recommendations to possession planners.
By combing enterprise field data, the project team have been able to remove paper-based systems and build a long-term data set for effective analytics to aid improved planning and decision making around possessions. This removes the reliance on individuals’ knowledge of location-dependent rules and actions, as well as enabling a more data-driven approach to be applied to risk management and programming.
Utilising a data set of more than 50 possessions across multiple UK locations, and building a stakeholder group of Network Rail, tier 1 contractors, and internal stakeholders, the FOAK project aimed to ensure that data science and the needs of the users on the ground were combined to deliver a solution with a true impact.
It is our hope that by creating a standardised approach to data management capture, and reporting, we will have a material impact on how possessions are managed, delivered and optimised upon successful completion of this SBRI programme.
To deliver innovation to the rail possession management industry, Frazer-Nash applied their experience in advanced data analytics to enhancing eviFile’s possession management solution, already being used by TRU Alliance (Transpennine Route Upgrade), Alstom and VolkerRail.
Under the SBRI ‘first of a kind’ grant, Frazer-Nash is producing a stand-alone module that will work with eviFile’s platform to enable contractors, PICOPs (Person(s) in Charge of Possession), and possession stakeholders to make better informed management decisions.
Following stakeholder engagement, the team identified a shared need amongst stakeholders for greater confidence in the ability to complete essential possession works on time. The time available for possession works, (the ‘possession window’), is crucial, as small delays can often lead to the cancellation or descoping of major activities. Capitalising upon eviFile’s historic possession management data, Frazer-Nash is developing predictive AI algorithms to forecast the likelihood of a possession overrunning, given the available time window.
The developed module utilises Bayesian networks to build a digital twin and view of scheduled possession works, allowing real-time analytics and confidence metrics relating to the progress of possession works to be reported on. Bayesian networks are a graphical approach that allow the possession management problem to be treated probabilistically, in which the duration of
works can be uncertain. Furthermore, the underlying probability distributions can be incrementally updated as more possession data is collated over time.
The predictions work by effectively matching possession characteristics to historic works, for example: worksite complexity, activity types, expected durations, seasonal trends, and time of
day. The tool will present results through an intuitive UI using Power BI dashboards, allowing PICOPs and contractors to form a data-driven consensus. The developed algorithms will also allow contractors to test mitigation strategies (for example, through descoping tasks/intelligent resource reallocation) to reduce their risk of overrunning, with a view to supporting the rapid planning and replanning of complex possessions in real-time.
By transitioning to a data-driven approach, eviFile will be transformed from a descriptive and diagnostic platform to a predictive and prescriptive toolset. The outcomes of our project will allow contractors to operate to tight schedules and deadlines with greater confidence, identify bottlenecked activities at the earliest opportunity, inform, and test remedial plans of action when works are behind schedule.
If you have a problem that needs solving, get in touch with the experts at Frazer-Nash or eviFile to discuss your requirements, at firstname.lastname@example.org