Servicing a machine too early wastes money on parts and needless downtime. Servicing it too late means a breakdown in full production. Between the two lies the right moment, and that is the whole point of condition-based maintenance.
Mimorian models equipment and captures the know-how of teams to help intervene at the right moment, on the right component. Condition-based maintenance means intervening according to the actual state of the machine, when a measured parameter crosses a threshold, rather than at a fixed date. The open questions: which parameter to monitor, where to set the threshold, and on which equipment the effort is worth it. Here is how to decide.
Where condition-based maintenance sits among maintenance strategies
Condition-based maintenance is one form of preventive maintenance, alongside systematic maintenance (at a fixed date) and upstream of predictive maintenance (which adds a forecast of the degradation). For the full picture, from corrective to predictive, see our comparison of maintenance strategies.
What sets condition-based maintenance apart: you intervene when the actual state of the machine calls for it, neither at a fixed date nor after the breakdown. That still requires measuring that state, and knowing how to interpret it.
Which signals should you monitor to intervene at the right moment?
Condition-based maintenance relies on a measurable parameter that betrays the degradation before the breakdown. The most common:
- Vibration: on rotating machinery (motor, pump, fan), an imbalance, a bearing defect or a misalignment raises the vibration signature well before the failure. It is the most widely used signal in condition monitoring [Source: ISO 17359, 2018].
- Oil quality: oil analysis (metal particles, viscosity, contamination) reveals the internal wear of a gearbox or a hydraulic circuit.
- Temperature: a hot spot on a bearing, a motor or an electrical cabinet signals emerging friction or overload.
- Counting: number of cycles, number of operating hours, number of start-ups. A usage threshold replaces or complements the physical measurement.
The principle stays the same whatever the signal: you set a trigger threshold, and you intervene when it is crossed, neither before nor after.
Condition-based or predictive: the threshold boundary
The difference comes down to one word, the forecast. In condition-based maintenance, you react to a threshold crossed now. In predictive maintenance, a model extrapolates the curve to announce when the threshold will be crossed. The EN 13306 standard in fact classes predictive maintenance as a form of condition-based maintenance carried out following a forecast [Source: EN 13306, 2018].
In both cases, the decision rests on a threshold. The real question is its origin: does that threshold come from an isolated sensor, or from the actual history of your breakdowns?
Choose according to equipment criticality
A piece of equipment whose stoppage is costly and which degrades gradually is the ideal candidate for condition-based maintenance. Non-critical, redundant or cheap-to-repair equipment does not justify the effort: run-to-failure is enough. Between the two, systematic preventive maintenance remains a reasonable compromise.
This grid avoids two symmetrical kinds of waste: instrumenting and monitoring machines that are not worth it, and letting a bottleneck asset break for lack of follow-up. A hydraulic unit serviced every six months can thus break between two visits, and the bill for a single episode climbs fast, in the order of several tens of thousands of euros.
The role of AI: the right threshold, and the means to justify it
To run condition-based maintenance, you need to know the actual state and be able to interpret it. A vibration or temperature threshold only makes sense against the machine's history: what is normal on one motor is not on another. By cross-referencing structured knowledge about a piece of equipment (its known failure modes, its intervention history) with the monitoring of its state, AI spots the moment when a part deserves an intervention, and on which component. Organisations that are mature in condition monitoring gain on average 9 % in machine availability [Source: PwC/Mainnovation, 2018].
The most reliable threshold draws on the facts as much as on the sensor. When a bearing has failed three times in six months at the same operating speed, the history sets the right interval better than a vibration curve alone. AI finally brings a less visible benefit: it gives the team what it needs to justify a replacement in advance, with figures to back it up, to a management that hesitates to commit the spend. Acting at the right moment becomes a reasoned decision.
Frequently asked questions
When should you use condition-based maintenance rather than systematic maintenance?
On critical equipment that degrades gradually and whose stoppage is costly. Systematic maintenance remains simpler for safety components or parts with a known service life.
Which parameters are monitored in condition-based maintenance?
Most often vibration, oil quality, temperature, or a count of cycles and operating hours. The chosen parameter must betray the degradation before the breakdown.
Do you need sensors for condition-based maintenance?
Not always. A count of operating hours or the breakdown history already gives a useful threshold. Sensors refine the measurement on the most critical equipment.
Conclusion
Condition-based maintenance is the right answer for critical equipment that degrades gradually. The rest falls under systematic preventive maintenance or accepted run-to-failure. Chosen well by criticality, and backed by a threshold built from the facts, it protects production better at an equal maintenance budget.
To place condition-based maintenance among all the strategies, see our comparison of maintenance strategies. For the framework of trustworthy AI in maintenance, our guide to trustworthy AI for maintenance. On what an unplanned stoppage really costs, see our real cost of unplanned downtime in industry.