Key takeaways
Short answer: Preventive and predictive maintenance both aim to stop failures before they happen, but they decide when to act very differently. Preventive maintenance runs on a fixed schedule — every X weeks or Y cycles — whether or not the asset needs it. Predictive maintenance watches real condition data, models the degradation, and forecasts the moment of failure so you can act just in time. PM trades wasted service for simplicity; PdM trades complexity for precision and earlier, smarter intervention. For the middle ground, see preventive vs condition-based maintenance.
Preventive maintenance is interval-based. You set a fixed schedule — by time (every 90 days) or by usage (every 10,000 cycles) — and the work order fires on that interval regardless of the asset's actual condition. Its strengths are simplicity and predictability: it is easy to plan and resource, requires no special technology, and reliably reduces the random failures of run-to-breakdown. Its weakness is that the interval is a fixed guess about a variable reality. Set it too long and failures still slip through between services; set it too short and you spend labour and parts servicing assets that were perfectly healthy — and every intervention is itself a fresh chance to introduce a fault. PM is robust and cheap, but inherently imprecise.
Predictive maintenance replaces the fixed interval with a forecast built from data. Sensors stream condition signals — vibration, temperature, oil chemistry, motor current, acoustics — and models analyse the trend to estimate not just that an asset is degrading but when it will fail, so you can intervene just before that point. Rather than service on the calendar, you service on the predicted remaining useful life. Done well, PdM both eliminates the wasted service of fixed schedules and catches developing failures earlier and more precisely than any interval could. The cost is real infrastructure: sensors, connectivity, historised data, and the analytics or models to turn signals into a trustworthy forecast — plus the discipline to act on it.
It helps to place predictive maintenance next to condition-based maintenance, because they are often confused. Condition-based maintenance acts when a measured indicator crosses a threshold — vibration exceeds this level, so service now. Predictive maintenance goes a step further: it uses the trend and a model to forecast when the indicator will reach failure, so it can plan the intervention in advance rather than reacting at the threshold. In other words, PdM is condition-based maintenance plus prediction. The distinction from preventive maintenance is sharper still: PM ignores condition entirely and acts on the calendar, while PdM acts on a data-driven forecast of the specific asset's future.
A critical motor is on a fixed quarterly PM — four services a year whether needed or not, and it still suffered two bearing failures last year that struck between services. Switch it to predictive: vibration and temperature sensors stream continuously, and a model trends the bearing signature. Six weeks out, the model forecasts the bearing will reach failure in roughly forty days. You schedule one planned replacement into a production gap two weeks before the predicted failure — and skip the blind quarterly services entirely. The result: no surprise breakdowns, far fewer interventions, and the repair done as a calm planned job rather than a 2 a.m. emergency. The forecast turned an unpredictable failure into a scheduled one.
The decision is economic, not ideological. Preventive maintenance is the right call for assets where failure is cheap, the schedule is reliable, and instrumentation would cost more than it saves — most low-criticality, predictable-wear items. Predictive maintenance earns its infrastructure on assets that are critical, expensive to fail, costly to over-service, or where failure is dangerous — there, the sensors and analytics pay back many times over. Most plants should run a portfolio: PM (or even run-to-failure) for the many low-criticality assets, PdM concentrated on the vital few. Starting PdM everywhere at once is a common, expensive mistake; start it where a single avoided failure justifies the whole investment.
Both strategies attack unplanned downtime, the largest availability loss in OEE for most plants. Preventive maintenance converts some random failures into planned work; predictive maintenance converts even more, earlier, with far less wasted service — turning would-be breakdowns into scheduled, minimal-impact interventions. The metric to watch is the ratio of planned to unplanned work: as PdM raises it, availability climbs and the six big losses shrink. The link runs through reliability — every forecast-and-prevent cycle raises effective MTBF and removes a future stop before it ever lands.
Fabrico connects maintenance strategy to the OEE it is meant to protect. It tracks PM schedules and completion, records the failure history that tells you whether intervals are right, and ties maintenance to live OEE so you can see whether a recurring downtime loss is a maintenance problem and whether moving an asset toward predictive actually raised availability. That closed loop — condition, action, result — is what tells you where predictive maintenance will pay off and proves it when it does. Book a demo to connect maintenance to availability.
Preventive maintenance services assets on a fixed time or usage schedule regardless of condition. Predictive maintenance uses condition data and models to forecast when a failure will occur and acts just before it. Preventive is interval-based; predictive is forecast-based.
They are closely related but not identical. Condition-based maintenance acts when an indicator crosses a threshold; predictive maintenance forecasts when failure will occur using the trend and a model. Predictive maintenance is essentially condition-based maintenance plus prediction.
No. Predictive maintenance is more precise but requires sensors, data, and analytics. It pays off on critical, expensive, or dangerous-to-fail assets, while simple preventive schedules are often the better value for low-criticality, predictable-wear items.
Condition sensors (vibration, temperature, oil, current, acoustics), connectivity and historised data, and the analytics or models to turn signals into a reliable forecast — plus the discipline to schedule work against the prediction.
Both reduce unplanned downtime, the biggest availability loss in OEE. They convert random failures into planned work — predictive doing so earlier and with less wasted service — which raises availability and shrinks the six big losses.
Zaloguj
Blog
