Predictive maintenance (PdM) plans maintenance from the measured condition of equipment rather than from the calendar. Technicians or sensors take regular readings, such as vibration, temperature or oil quality, and work is planned when the readings show the asset starting to degrade. The aim is to act before failure, but not before it is needed.
It is closely related to condition-based maintenance (CBM), which acts when a reading crosses a set limit; predictive maintenance also uses the trend to estimate time left before failure.
Predictive vs preventive maintenance
| Item | Preventive | Predictive |
|---|---|---|
| Trigger | Fixed time or usage interval | Measured condition and its trend |
| Example | Change gearbox oil every 6 months | Change gearbox oil when analysis shows wear metals or degraded additives |
| Upfront cost | Low | Higher: instruments, sensors, training or a specialist service |
| Best fit | Most important equipment | Critical, high-value equipment, especially rotating machinery |
Most organizations combine them: preventive maintenance for the bulk of their equipment, predictive maintenance for a short list of critical assets, and corrective maintenance where running to failure is cheapest.
Predictive maintenance techniques
| Technique | What it detects | Typical assets |
|---|---|---|
| Vibration analysis | Imbalance, misalignment, looseness, bearing and gear wear | Motors, pumps, fans, gearboxes, compressors |
| Infrared thermography | Hot spots from loose connections, overloads, friction or failing insulation | Electrical panels, switchgear, bearings, steam traps |
| Oil analysis | Wear metals, contamination, water, additive breakdown | Gearboxes, engines, hydraulic systems, transformers |
| Ultrasound | Air and steam leaks, early bearing faults, electrical arcing | Compressed-air systems, valves, bearings, switchgear |
Readings come from handheld instruments on a route, a specialist contractor's visits, or installed sensors feeding software.
The P-F curve
The P-F curve explains why predictive maintenance works. There is a point P (potential failure) where a defect first becomes detectable, and a later point F (functional failure) where the asset stops doing its job. The time between them is the P-F interval. A typical progression for a bearing:
- The defect starts, such as a tiny crack in a bearing race.
- Ultrasound and vibration pick it up (point P): the best time to plan the repair.
- Oil analysis and heat show it, later on the curve.
- Noise and visible heat appear, close to failure.
- Functional failure (point F): the asset stops.
To catch a defect, readings must be taken more often than the P-F interval; a common rule is to inspect at no more than half of it. If the P-F interval for a bearing is around eight weeks, monthly readings give at least one chance to see the problem and plan the fix.
When predictive maintenance pays off
- The asset is critical: failure stops production, affects safety or patients, or there is no standby.
- Failure is expensive: high downtime cost or long part lead times.
- The failure mode gives warning: it has a measurable P-F interval. Many electronic failures are random and do not.
- You have, or can buy, the expertise to interpret the data.
Examples by industry
- Manufacturing: vibration routes on a bottleneck line's motors and gearboxes.
- Hospitals and large buildings: thermography of distribution boards; oil analysis on standby generators.
- Fleets: engine oil analysis to set drain intervals.
Common mistakes
- Buying sensors before fixing the basics. If PM is not being done, more data will not help.
- Collecting data nobody reviews. Each point needs an owner and an action limit.
- Monitoring the wrong assets. Start from a criticality ranking.
- Not closing the loop. A finding should become a work order, and what the technician finds should feed back into the limits.
Spreadsheet vs software
Route readings are often logged in a spreadsheet with a trend chart per asset. That works for a short list. It breaks down with many measurement points and when readings are not linked to the work that follows. Dedicated condition-monitoring software handles the data and analytics; a CMMS handles the work that follows.
Predictive maintenance and Asetavo
Asetavo does not collect sensor data or perform condition monitoring or predictive analytics. It manages the assets and the work around them: the register, work orders with cost and downtime, PM and calibration schedules, and service history. Findings from a vibration route or an oil-analysis report can be raised as corrective work orders on the asset, with the report attached to the asset as a document. See Asetavo maintenance (Business plan).
