ProRail is building a long term roadmap in which data and AI are structurally embedded in maintenance and planning. At its core is the shift from reactive and time based maintenance to predictive and prescriptive maintenance. Not repairing once something fails, but anticipating when intervention is needed and acting with precision. KPMG supports ProRail in turning this strategy into tangible results through a series of successful initiatives.
The maintenance approach is strongly data driven and focused on maximising asset availability at predictable cost. A concrete example is the video inspection train, which captures the entire rail network twice a year using high resolution cameras. These images, together with environmental and process data, are processed on an Azure based big data platform designed and developed in collaboration with KPMG.
Within this platform, AI models automatically recognise and assess assets. Using computer vision, sleepers are identified and analysed for condition and degradation. Where physical sensors are unavailable or not cost effective, ProRail applies soft sensing. These virtual sensors use machine learning models to predict asset condition based on existing data, without the need for additional hardware.
To ensure reliable use in production, an MLOps approach has been established. The full model lifecycle is managed, from training and validation through to deployment and monitoring. This ensures models remain scalable, reproducible and suitable for sustained use in operational decision making.
Digital Twins and advanced planning algorithms enable ProRail to translate these insights into concrete maintenance and capacity decisions. Data is therefore not only valuable for analysis. It is an essential management instrument.
At the same time, employees are actively involved in adopting new ways of working. Digitalisation is not an IT project. It is an organisation wide development in which learning and continuous improvement are central.
