• Home
  • Projects
  • A proactive maintenance AI classification solution for the rail industry

A proactive maintenance AI classification solution for the rail industry

Image of laptop with digital graphic
Speak to an expert about your challenge
Get in touch

Related markets and services

Amey’s AI classification model has brought significant value to the rail industry, specifically when it comes to rail maintenance. By helping to automate the most time intensive stage of the rail inspection process, we have helped rail organisations significantly improve the speed, accuracy, and efficiency of maintenance operations.

At a glance

  • Amey’s automated Data Quality Pipeline identifies data recording issues and presents them through a dashboard, enabling teams to plan reruns efficiently and avoid delays.
  • The team developed an end to end data processing pipeline that integrates a computer vision model to automatically predict rail type, ensuring accurate wear and tear calculations.
  • A machine learning model has replaced the manual data classification process, reducing rail type identification time from 160 hours to just 10 hours.
  • The solution has delivered a 16x reduction in the time required to generate reports, significantly improving efficiency and enabling around the clock backend data classification.
  • Automation of rail health status reporting has been accelerated, allowing earlier detection of potential defects and safety hazards.

Key metrics

  • 10

    hours required for rail type identification

  • 16x

    reduction in time spent generating reports

  • 15

    automated calculations used to predict rail type

The AI tool enables a more proactive approach to rail asset management, supporting earlier interventions, reducing the risk of disruption, and ultimately helping to deliver a more resilient and reliable rail network that improves outcomes for everyday passengers.

The challenge: manual process holding back rail maintenance

Rail maintenance teams often use laser scanners to create 3D models of rail tracks for structure analysis and defect detection. Once the data on the rail tracks is collected, engineers manually process and analyse the data to identify which type of rail track they are reviewing and what structural issues to look out for. This data is key for rail maintenance operations as it details the specific shape and structural attributes of the rail tracks and whether they are affected by any signs of wear and tear. This enables engineering teams to deliver intervention repairs before any further breakages or failures occur.

However, the initial phase of the process, which involves the manual matching of the rail data with its specific rail track profile, would take teams on average up to 160 hours to complete before the rail models could even be analysed for faults and damages. The extensive length of time this took would often increase the risk of significant delays on train services if structural rail track issues were not identified in good time.

As a solution to this challenge, Amey designed and developed an AI automation model that utilises a combination of machine learning (ML) and Natural Language Processing (NLP) technologies to automate the rail track profile identification process. This enables engineers to review and match the rail data with their correct profile templates in a way that is much faster and more efficient. This also includes activities where rail teams are manually entering and reviewing large groups of data sets which can otherwise be automated and streamlined, requiring minimal human intervention.

Amey’s solution: applying AI to transform rail data classification

As part of this project, we designed, developed, and trained a machine learning model that utilises natural language processing (NLP) to accurately classify rail track models against known track templates. Once the 3D models of the rail tracks were created, the data was fed into the machine learning model to analyse, compare, and predict which rail templates each of the data points belonged to. This resulted in the identification of their rail type. The tool carries out 15 different calculations to predict the rail type of data points.

This critical step, which previously relied on manual input, was now performed autonomously, accelerating the classification process and reducing analysis time from hours to just minutes.

The data on the rail tracks is stored and easily accessible for maintenance engineers via a web application and illustrated in various digital forms including dashboards. Once rail types are classified, maintenance teams can immediately begin structural assessments, identify defects, and schedule interventions based on data driven insights.

By streamlining the most time intensive part of the inspection process, the AI tool not only enhances operational efficiency, but it also enables a more proactive approach to rail maintenance across the network.

The outcome

Our AI powered rail classification tool has delivered tangible improvements for the rail maintenance process. It has enabled engineering teams to work at a much faster pace and with far greater efficiency, supporting a more proactive and data driven approach to maintaining rail infrastructure.

The solution has significantly reduced rail asset analysis time by automating the rail type classification process, cutting a task that previously took 160 hours down to just 10. This allows engineers to assess the health status of rail tracks much more quickly, enabling earlier interventions and more responsive maintenance. Operational efficiency has also improved, as removing the need for repetitive manual classification allows teams to focus on more strategic and critical activities such as defect analysis, planning, and intervention. This shift not only reduces overall maintenance time but also boosts productivity and optimises resource allocation.

Decision making has been strengthened through the use of a web based dashboard that hosts all classification data, giving engineering teams real time access to insights. With clearer visibility of asset condition and performance across rail types, teams can prioritise repairs based on risk and urgency, supporting smarter planning and investment. The accelerated identification and inspection process has also contributed to greater network resilience by reducing the risk of service disruptions caused by undetected faults. Earlier detection of wear and tear enables pre emptive maintenance, helping to protect the operational continuity of rail services.

AI is playing a transformative role in the rail sector, particularly in enhancing safety, efficiency, and reliability, all of which are increasingly important to passengers in today’s evolving travel landscape. Building on these improvements, our customer‑centric approach and digital‑led expertise have enabled us to automate and strengthen rail asset management processes through the development of our AI classification model. As a result, Amey’s AI solution is now a core component of asset maintenance operations, supporting engineering teams to proactively monitor their assets, intervene earlier, and reduce service disruptions across the network.

Speak to an expert about your challenge.