Maintenance
Types of maintenance
Periodic, counter-based, condition-based: how to choose between them and why combining them on the same plant pays off.
Maintenance is planned in three ways, and the difference lies in what triggers the job: a date, accumulated use, or a measurement that signals wear. None of the three is better than the others in absolute terms: it depends on how the component wears out.
| Logic | What triggers the job | Typical use |
|---|---|---|
| Periodic | A calendar due date | Lubrication, inspections, overhauls |
| Counter-based | A threshold of hours or cycles | Filters, oils, tools subject to wear |
| Condition-based | A measurement beyond a threshold | Vibration, temperatures, fluid quality |
Periodic
The job comes back at fixed intervals, regardless of how much the machine has worked. It is scheduled from the date of the last recorded job, and the due date approaches on its own.
When it works best: when the frequency is known, stable and prescribed, and when what degrades the component is time rather than use. Safety inspections stay periodic even when better data exists, because their whole purpose is not to depend on any other logic.
The limit: it ignores actual usage. A machine idle for two weeks accumulates due time as if it had been producing, and a machine pushed across three shifts accumulates it too slowly.
Counter-based
The job is triggered when a cumulative quantity reaches a threshold: run hours, cycles completed, Units produced. The quantity comes from device data, so the count follows the actual work instead of an estimate.
Recurring examples: a filter to replace every so many run hours, a tool to change every so many Units processed, a belt to check every so many cycles.
When it works best: when wear is proportional to use. On plants with variable load or frequent stops it is more accurate than periodic, and it generally saves unnecessary jobs without raising risk.
The prerequisite: you need a reliable counter. If the count resets at every restart or is lost during stops, the threshold is never reached, or is reached at random.
Condition-based
The job is planned when a measurement shows the component is genuinely deteriorating: vibration rises, a bearing temperature climbs, a fluid loses its properties. It is the most advanced of the three, and the only one that reacts to the real state of the machine instead of a forecast.
It takes more preparation than the others: you need a suitable sensor, and you need to work out where to set the threshold. The page on condition-based maintenance goes into detail.
Combining them is the norm, not the exception
All three coexist on the same plant, and often the same component needs two of them.
The most useful case is condition alongside a periodic due date: the job is brought forward if the measurement calls for it, but it still comes at the due date even if the measurement never signalled anything. That way data serves to bring jobs forward, not to postpone them, and no component is left ignored because its sensor was silent.
Where to start
A maintenance plan does not start out complete. A path that works:
- put what is prescribed on the calendar, so no obligations are missed;
- move usage-related jobs to counters as soon as the data is available;
- add condition on the few components that cost the most when they fail, and for which a measurement exists that precedes the failure.
The third step is the one worth most, and that is why it is best done last: on a plant where the first and second are not in order, condition alarms arrive in a process that does not yet know how to react.