Use of AI to Achieve a Reliability-Centered Maintenance Regime

Case Study, 09.02.2022

In August 2021, Network Rail granted full product acceptance to KONUX after two successful trials. The trials began in August 2020 with the goal to demonstrate the capability of the The KONUX Predictive Maintenance System for Rail Switches in asset management and maintenance validation.

The Challenge

One of the main challenges many infrastructure managers share is that the current inspection and maintenance regime is based mainly on snapshots of measurement taken at fixed time intervals. This results in highly biased insights into switch (S&C) conditions due to variability of different train types, speeds, and temperatures, preventing optimization of maintenance decisions.

The lack of insights into the actual condition of the assets could lead to, for example, unnecessary tamping on healthy assets. This does not only accelerate ballast degradation but also prevents infrastructure managers from deploying the resources where they are most needed.

Also, the interval-based regime makes little information available for evaluating maintenance effectiveness. When unsuccessful activities are left unnoticed for weeks or months until the next cycle, the asset can degrade faster in the meantime and, in the worst case, the damage could leave no other option than replacing the asset.

To have an unbiased picture of switch conditions and make maintenance decisions on a sound basis, it is crucial to have constant insights into the switch conditions as well as the actual usage of the switches.

To explore solutions against such challenges, Network Rail embarked on trials with KONUX.

The Solution

The trials were run in the East Midlands and East Coast routes, with the KONUX IIoT devices installed on both wooden and concrete sleepers with different installation methods. The demonstrated use cases were trackbed health monitoring, true load identification (usage-based inspection planning), and trackbed maintenance validation.


KONUX IIoT device installed on the concrete sleeper (left) & User interface (right)

For trackbed health monitoring, the KONUX system demonstrated the capabilities to monitor:

  • the ballast stability based on the vertical displacement of the sleepers
  • degradation cycle to help identify a root cause if the cycle is too short
  • acceleration forces to detect early signs of wear on frogs (‘crossings’ in the UK) & blades

For true load identification, the KONUX system demonstrated the capabilities to identify:

  • the actual usage of the switches, revealing the insights into actual tonnage, train speed, and the number of trains
  • the most pressured switches that require more attention than other assets in the same track category
  • the impact of different train types

For maintenance validation, the KONUX system demonstrated the capabilities to:

  • Evaluate the performance and sustainability of the tamping activities
  • Benchmark procedures, tools, or subcontractors based on the performance

The Results

The trials demonstrated clear cost-saving benefits: on the East Coast route, the KONUX system helped avoid unnecessary tamping activity based on the data indicating low vertical displacement and no sign of voiding.

This also prevented track closure and speed restrictions that might have otherwise affected track availability and capacity. Such results have demonstrated the KONUX system’s potential to improve the current inspection and maintenance regime to become more efficient. The testing revealed a potential 20% reduction in delay minutes.

In terms of maintenance validation, the trial successfully validated the ineffectiveness of a tamping action based on the data of the rising vertical displacement close to the pre-maintenance level in just two days. This insight helped identify ineffective practices, prevent potential failures by acting on time and reduce costs by eliminating repeated actions.


Would you like to learn more about the case? Get in touch with our team here.

This site uses cookies. By continuing to browse the site, you are agreeing to our use of cookies.

Okay