Predictive maintenance: can data trends reduce disruption for freight trains?
New insights into existing data could predict which wagons need maintenance before major faults occur.
Sarah ShooterEditorial Content Manager, RSSB
New government targets call for a 75% increase in freight business by 2050. It’s an ambitious goal, but it’s also an opportunity for innovation in many areas, including safety, technology, efficiency, and data.
Many freight trains get pulled from service each year due to the condition of their wagons. This is essential for safety reasons, particularly for preventing derailment and damage to infrastructure. However, it can also be disruptive to schedules and budgets.
RSSB’s Data Insights team are working on a method to use existing industry data to predict where maintenance is needed before it causes issues on the network.
You may have read about Wheel Impact Load Detectors (WILD) in Research Highlights. These are sensor systems installed on Network Rail tracks around the country that measure wheel loads from passing traffic. There are over 70 WILD sites across the network, and they monitor rolling stock as it passes over them.
When the WILD system records a passing wagon that impacts a force of at least 350 kilonewtons, an alert is generated. The train is then pulled from the network. Some freight companies use lower thresholds to identify wagons which require maintenance. However, this is complicated because the wagon load will affect the measured force at WILD sites.
The WILD systems project has been running for several years already. In that time, it has collected a sizeable amount of data. The data from WILD systems is already contributing to the safety of the network by ensuring daily that rolling stock is fit for use. But being able to predict when a wagon needs maintenance before it reaches the tipping point would be even better.
RSSB’s Data Insights team are now exploiting that same data to bring even more benefit to the industry, with predictive analytics.
Predictive analytics is an advanced way of using data, statistics, modelling, and machine learning techniques. Put simply, it uses historical data to identify the likelihood of future outcomes and how they may present.
This type of data analysis is used in many industries, like healthcare, aviation, and energy. In aviation, predictive analysis has been used to support proactive maintenance of aircraft systems. You can read more about it in a cross-industry analysis from our Horizon Scanning team.
Read more
The key output for this project will be a decision support tool, through which companies can proactively manage the condition of their fleet. The tool will predict the future degradation of wheelsets. It will allow users to position resources and parts in the right place at the right time. Freight wagons can then be maintained before they become a risk to operations.
Most freight companies having quite different approaches to capturing maintenance data—both the systems used and the actual data stored. The WILD data, however, is universal. It gives a clear standard for comparison across all rolling stock. This means the data collected can be trusted to apply to all freight companies on the mainline network and benefit the whole industry by reducing schedule disruption and derailment risk.
Reducing the number of faulty wagons on the running line will also contribute to the longevity of the railheads themselves. They can be worn faster by larger imperfections in the wheelsets.
This project team is in the process of finalising the predictive models, and we are in discussions about a practical pilot of the tool.
We are planning to have the system in pilot with two freight companies by the end of 2024. At this stage, the goal will be to validate the solution, and share feedback and findings, before the technology is made available to the wider industry.
Read our Research Short about WILD data.
Read Horizon article
Do you want to suggest a topic for the Data Insights team to analyse?
Share your idea
