A data-driven approach to sustainable operation, maintenance and development of the railway electrification system

A data-driven approach to sustainable operation, maintenance and development of the railway electrification system

Project partner: Network Rail and Virgin Trains

PI: Dr Alexeis Garcia-Perez

Summary

A consortium led by Coventry University developed a prototype solution for data-driven infrastructure monitoring in railways. An operational environment, namely the London-North West railway line, was used to study the utility of the technology by simulating the operational requirements and specifications required for the final system. The project demonstrated the feasibility of turning every train into an infrastructure monitoring train, informing the wider digital transformation strategy of Network Rail and the overall railway industry.

The Problem

As the railway industry moves towards a more data-driven decision making culture, an increasing proportion of its investment decisions is being based on facts and figures derived from a variety of data resources. As the infrastructure manager, regulatory requirements exist that demand regular monitoring of the infrastructure by Network Rail. The busiest routes require very frequent monitoring to ensure that the condition is maintained at a standard fit for transporting tens of thousands of passengers per day. Linear infrastructure, including track and Overhead Line Equipment, are monitored using sensors installed on the limited number of monitoring trains part of the Network Rail-managed fleet. The cost of running and maintaining this fleet cost tens of millions of pounds annually. In addition, there is significant cost associated with collating and processing this data, as parts of this process are performed manually.

It is therefore considered important for Network Rail to develop two key elements as part of their digital transformation strategy: systems that can collect and deal with the big data assets that could be captured from passenger-trains with sensor equipment installed; and developing algorithms that enable utilisation of the data to improve risk-based maintenance of the infrastructure.

The Coventry University – Network Rail Partnership

Back in 2014, Dr Alexeis Garcia-Perez had been awarded funding by a partnership between the Rail Safety and Standards Board (RSSB) and the Rail Research UK Association (RRUKA) for the study of the relationship between railway data and safety. As part of the KEEP SAFE project (KEEPSAFE/RSSB/13/RRUKA/1676), a series of workshops were conducted in Coventry with an aim to build a common understanding across the industry of the nature of railway data and safety in railway. Key industry players –including Network Rail, London Underground, Transport for London, the Office of Rail and Road and Thales participated in the project, informing the development of a safety prognosis tool and a knowledge-based approach to understanding safety in railway.

Success of the KEEPSAFE project led to a renewed partnership between Coventry University and Network Rail in 2017, this time for the purpose of contributing to a wider initiative aimed to improve the industry’s data-driven infrastructure management capabilities. The main area of focus of the new collaboration was the Overhead Line Equipment, an essential element of the railway infrastructure and one that has both a significant maintenance cost and a large impact of failure on the overall functioning of the railway network.

A Prototype Solution to Infrastructure Monitoring Using Passenger Trains

With research funding from Network Rail, the new project was led from the Faculty Research Centre for Business in Society by Alexeis Garcia-Perez, with the aim of demonstrating the feasibility of turning every train into an infrastructure monitoring train. A cross-industry consortium was formed which included in addition to Network Rail, partners such as Alstom Transport, Virgin Trains and Serco, as well as academics from different Faculties at Coventry University and from the Department of Engineering Science at the University of Oxford.

A number of sensors were fitted on the OLE infrastructure of a train operated by Virgin Trains. Data related to the health, usage and performance of the OLE system in the London-North West (LNW) railway line was collected as the train travelled between London and Carlisle over a period of six months.

Coventry University then led the consortium in their efforts to understand the issues pertinent to the collection, secure transmission and storage, and effective management of the railway infrastructure data to inform Network Rail’s infrastructure management strategy.

The secure data storage

A secure OLE data repository (Tactical Data Hub) was created within the Coventry University Information Technology infrastructure allowing for:

  • Secure transfer of data from the passenger train to the Coventry University repository
  • Secure hosting of the railway data and its availability to key industry players for the purpose of research
  • Secure data management and archiving processes.

The Tactical Data Hub contains OLE data in both its original (i.e. sensor) format and in CSV format for ease of use by other stakeholders from industry (e.g. Network Rail, Virgin Trains, Serco, Alstom, Thales, Vivacity) and academia (e.g. Coventry University, University of Oxford and University of Southampton).

Preliminary data analysis

Once stored at the Tactical Data Hub, a preliminary analysis of the OLE data was implemented in order to provide the consortium with an understanding of the quality of the data being collected and its potential to inform effective decision-making processes. This preliminary analysis was based on data cleansing processes currently conducted by Network Rail experts. Its outcomes led to further action by the consortium in an effort to improve the quality of the data collection process.

Data visualisation

Initial attempts to visualising the data for single- and multiple-runs of the passenger train over the same GPS locations showed that there was not enough valid data on every point of the LNW line to enable a meaningful evolutionary analysis of the health of the infrastructure.

Thus, the team designed a new approach to data visualisation that consisted of:

  • Primary analysis of all OLE data as collected (e.g. standardising the frequency of values) to ensure that it was possible to map every record with a GPS location.
  • Mapping of all OLE data, allowing users to select particular dates to be visualised at any given time.
  • Using colour coding to visualise the data in order to facilitate its visual inspection. Colours would be based on correlation between specific parameters and their expected thresholds.
  • Allowing the user to select a point (combination of a data point and its related geo-position) of potential interest to then dynamically generate diagrams representing the relevant parameters (e.g. Contact Forces and Acceleration against Velocity of the train) over time.

A data visualisation system was then developed using a non-intrusive approach (i.e. hosted in Coventry University’s IT infrastructure and accessible via a Web browser), to ensure that the system can be accessed by Network Rail engineers in the field, with no additional requirements.

Performance of the system was tested in an effort to provide Network Rail with an understanding of the scalability of the data transfer, conversion and visualisation processes.

Project Benefits

In providing a pilot for predictive maintenance in an operational environment (TRL 6), this project contributed to three main areas of development within Network Rail’s efforts to move away from traditional approaches towards intelligent maintenance of its infrastructure. These areas are (1) the secure storage and cross-industry sharing of OLE data; (2) the manipulation of big data sets related to the health of OLE infrastructure; and (3) the visualisation of key parameters within the data to inform decision-making. Implementation of this approach across the railway network could enable an optimisation of infrastructure maintenance, a reduction in resources costs and a reduction in disruption of services.

Finally, the project has set the basis for the systematic application of machine learning techniques across OLE assets data, sensor measurement data and traffic data. This also extends to consider the broader problem looking at both asset-related data – for example track measurements – and external environmental data. This better understanding of the assets and associated risk factors enables Network Rail to improve internal standards and associated information systems to flag up when risk to the infrastructure is high. This would enable a quantitative risk-based maintenance approach rather than approaches based on engineering experience alone.