Glossary · Safe work, maintenance and competence
Predictive maintenance
Also known as: PdM
German: Vorausschauende Instandhaltung
In maintenance, predictive maintenance is maintenance based on condition data and analysis used to predict degradation or failure, so work is done when the data indicates it is needed.
- Machine safety
- Maintenance
In one sentence
Predictive maintenance uses condition data and analysis to forecast failures. For safety parts, the prediction itself needs validation.
Example
Vibration sensors on the spindle bearings of a CNC machine feed a model that predicts bearing wear; the bearing is replaced at the next planned stop instead of at a fixed interval.
How it applies
- Machine safety: A prediction can be wrong. Where a component carries a safety function, fixed limits from the design, such as switching cycles, still apply unless the prediction method is shown to be adequate.
- Integrated systems: Predictions depend on sensors and data pipelines. Sensor drift, gaps or stale values can delay maintenance silently; define what happens when data is missing.
- AI and retrieval: Many predictive models use machine learning. Their inputs, versions and performance should be monitored like any other system in use, see post-deployment monitoring.
- Technical documentation: The maintenance manual explains which parts are maintained on prediction and which on fixed intervals, and what the user does with alerts.
Predictive vs. preventive maintenance
Preventive maintenance is planned at fixed times, usage or conditions. Predictive maintenance schedules work from an analysis of condition data. Many classifications treat it as a form of condition-based preventive maintenance.