Glossary
Predictive maintenance
Using sensor data and models to predict when equipment is likely to fail so it can be serviced before it breaks down.
Also called: PdM
Predictive maintenance uses data from equipment, vibration, temperature, pressure, sound, or electrical signatures, to estimate when a machine is likely to fail, so repairs can be scheduled just before that point rather than on a fixed calendar or after a breakdown. It sits between reactive maintenance, fixing things after they fail, and preventive maintenance, servicing on a fixed schedule regardless of actual condition.
Models typically flag anomaly detection-style deviations from a machine's normal operating signature, or estimate remaining useful life using historical failure data and current sensor readings, often benchmarked against a machine's mean time between failures. Some organizations pair this with a digital twin, a live virtual model of the equipment, to simulate wear and stress under different operating conditions before committing to a maintenance schedule.
Predictive maintenance matters because unplanned downtime is expensive and unscheduled failures can also be safety incidents, while over-maintaining healthy equipment wastes labor and parts; done well, it improves overall equipment effectiveness by reducing both unplanned stoppages and unnecessary servicing. The main pitfalls are false positives that erode trust in the system when flagged equipment turns out to be fine, and models trained on too little failure history to reliably distinguish a genuine precursor to failure from normal operating noise.
Last reviewed September 22, 2026