A growing number of organisations, from manufacturing companies to defence institutions, want to detect failures early and schedule maintenance predictively. The technology to do so exists: sensors continuously record machine behaviour and AI models identify patterns indicating wear or impending faults. In practice, however, a widespread problem persists: organisations have access to more data than ever before, but that data is difficult to combine or interpret. Two measurements of the same type of component, taken at different locations or points in time, are not automatically comparable. Differences in sensors, operating conditions and registration methods make it difficult to identify reliable patterns.

The core of the problem lies with metadata: information about when a measurement was taken, under what circumstances, with which type of sensor, and on which component. In most organisations, this information is not recorded in a standardised way. As a result, measurement series cannot be compared, maintenance logs cannot be interpreted automatically, and AI models remain dependent on simulated data. Nathan Houwaart, researcher at the Industrial Digital Twins research group at HvA, summarises the challenge: “Failure of critical systems costs money and time, and sometimes compromises safety. Better data are the first step towards maintenance that prevents failures rather than repairs them.”

The Industrial Digital Twins research group and the NLDA addressed this problem together. Defence has a direct operational motivation for improved data quality: the reliability of technical systems, ranging from vehicles to aircraft and ship components, is of critical importance. Unexpected failures carry consequences that are both financial and safety-related. The researchers identified the minimum information needed to make maintenance metadata usable: what contextual information must be recorded, how it should be documented, and in what format. The result is a structured framework describing how maintenance-related measurement data can be recorded and shared unambiguously.

The newly developed standard has relevance beyond the HvA and NLDA partnership. The HvA is contributing the findings to the Dutch Prognostics Lab, a national initiative funded by the Ministry of Defence, in collaboration with the University of Twente, The Hague University of Applied Sciences, NLR and the Rijnland Regional Water Authority. This lab focuses on standardising measurement and data practices for predictive maintenance across the Netherlands, with the aim of bridging the gap between theory and practice. The collaboration fits within the broader mission of the research group: developing practical tools that enable industrial organisations to use Digital Twin technology for more efficient and cost-effective maintenance, resulting in fewer unexpected failures, lower costs and smarter deployment of scarce technical specialists.