Aurizon's Keystone and IronMan®: data-driven reliability improvement

What Aurizon presented

Aurizon's Keystone program is a fleet-wide push to make maintenance tactics for its heavy-haul rollingstock decisions backed by evidence rather than habit — when to replace a component, when to leave it running, and where the next dollar of maintenance spend actually earns its keep. At IHHA 2025, Aurizon's own engineering team set out how that shift has played out in practice, and what changed once the analysis moved from a manual, months-long exercise to something running continuously across the fleet.

Where IronMan® fits

OXMT provided the analytical backbone behind that shift. IronMan® ingests the maintenance and reliability data Aurizon's teams were already generating — work orders, component histories, condition data — and applies machine learning and reliability-engineering methods (component-age analysis, Weibull and CROW-AMSAA modelling, and text mining across free-form maintenance notes) to structure it automatically. From there, the platform:

• Rapidly optimises maintenance tactics through a data-driven, rather than tenure-driven, approach
• Flags and reduces unnecessary preventative work and its associated cost
• Improves fleet reliability and availability by surfacing risk before it becomes a failure
• Gives planners and reliability engineers the clarity to act on a recommendation with confidence, instead of spending months building the case for it first

The published numbers

The figures below are the ones Aurizon's own paper put forward at IHHA 2025 — not a marketing estimate.

Why this matters beyond rail

Heavy-haul rail shares a structural problem with mining and other capital-intensive fleets: the data needed to run reliability-centred maintenance properly already exists somewhere in the CMMS, but it's too messy, too fragmented, or too time-consuming to turn into a decision. Aurizon's result is a concrete example of what happens once that barrier is removed — not a hypothetical ROI model, but a fleet-wide program with published, peer-presented figures behind it.

For a closer look at the reliability-engineering methods behind this — Weibull analysis, CROW-AMSAA, and RCM — see the Reliability module inside IronMan®.