Track Thinks in Networks. Trains Think in Hierarchies. One EAM Should Handle Both.

The rail asset management market is on course to double — from $3.1 billion in 2025 to $6.2 billion by 2035, growing at a 7.22% compound annual growth rate. Transit agencies across North America and Europe are investing in EAM platforms with more intensity than at any point in recent memory, driven by AI adoption, sensor deployment, and mounting pressure to replace aging infrastructure before it causes service disruptions.

The investment is warranted. What's less certain is whether the platforms receiving that investment are capable of handling both sides of the rail maintenance equation.

Two asset worlds in one operation

Run the assets in a large transit agency through any first-principles categorisation and two fundamentally different maintenance worlds emerge.

The first is linear: mainline track, overhead catenary, third rail, signal cables, communication infrastructure. These assets are spatially distributed, interconnected, and governed by network topology. A broken rail at one location doesn't just create a local work order — it propagates. It affects signal block availability, alters dispatch routing across the network, changes minimum headways across adjacent segments, and often triggers cascading schedule impacts across the entire line. Managing these assets well requires graph-aware reasoning: the ability to model which nodes and edges in a network are affected when any single element degrades or fails.

The second world is discrete: rolling stock, depots, traction motors, bogies, HVAC units, doors, pantographs. These assets follow a strict hierarchy — a vehicle consists of systems, systems contain assemblies, assemblies contain components — and maintenance follows parent-child logic. If a bogie frame develops a fault on one vehicle class, it raises the inspection priority for every vehicle in that class. ISO 14224 taxonomy, OEE measurement, and age-based replacement intervals all depend on that hierarchical structure being consistently maintained and applied. Performance at this level is measured in fleet availability, mean time between failures, and maintenance cost per vehicle kilometre.

Both worlds exist in the same organisation. Both require AI-driven decision support. Both need to be audit-ready for safety certification. The challenge is that these two paradigms require fundamentally different data models — and most EAM platforms were designed with one in mind, not both.

Where single-paradigm EAM falls short

Traditional CMMS-derived EAM platforms handle the discrete world reasonably well. They track work orders, manage parts inventories, and apply ISO 14224 hierarchies to rolling stock and depot equipment. What they cannot do is model what happens to a network when one infrastructure node fails.

GIS-native linear asset tools handle the inverse. They understand spatial relationships, segment topology, and network dependencies. But they were not built for the hierarchical failure logic that governs rolling stock maintenance or depot operations.

Most transit agencies resolve this by running both in parallel — a network-aware tool for infrastructure, a traditional EAM for rolling stock and facilities. The operational consequence is predictable: two data estates with incompatible schemas, maintenance decisions made in isolation, and an AI adoption ceiling that stops exactly at the boundary between systems.

The AI limitation is structural. A predictive model trained on rolling stock sensor data can forecast when a traction motor is likely to fail. What it cannot surface — if it has no visibility into the track network — is that the vehicle carrying that motor operates exclusively on the agency's only uninterrupted rapid transit corridor, and that taking it out of service during morning peak requires a network-level rerouting that affects twenty-two downstream services. That context does not exist in the rolling stock hierarchy. It lives in the network model. Producing a genuinely useful maintenance recommendation requires both.

The audit trail requirement adds pressure

ISO 55001 certification — which a growing number of transit authorities are pursuing as evidence of management system maturity — adds a documentation requirement that fragmented EAM deployments handle poorly. The standard demands that AI-informed maintenance decisions be traceable: linked to the asset condition data, risk assessment, and organisational context that drove the recommendation, not merely logged as a closed work order.

When decisions about track segments live in one platform and decisions about vehicles live in another, producing a unified, clause-level audit trail is a manual exercise. It is error-prone, time-consuming, and difficult to defend when a safety authority asks how a particular maintenance deferral was justified.

The UniEAM agentic platform was designed to close this gap. Every AI recommendation — whether it originates from a network-segment risk model or a vehicle-level condition assessment — carries a decision log capturing the reasoning, the input data, and the confidence level. The audit trail is unified by default, not assembled after the fact.

The case for a single reasoning layer

Rail transit represents the mixed-portfolio problem in its most legible form. No other sector has assets that split so cleanly between the linear world — track infrastructure that is fundamentally spatial and network-governed — and the discrete world — rolling stock and depots that are fundamentally hierarchical and component-governed.

The organisations that extract the most value from AI-driven maintenance are not those that buy the best rolling stock EAM and the best track management tool separately. They are those that consolidate on a platform where a single AI layer can reason across both worlds simultaneously — connecting a deteriorating track section to the vehicles that run on it, the depot that services them, and the network schedules that depend on all three.

For a broader look at the principles behind network-aware reasoning for infrastructure, see the guide to linear asset management.

Curious how UniEAM handles both? Request a demo.