Maintenance strategies are usually presented as a maturity ladder, with run to failure at the bottom and condition-based monitoring at the top, and that framing has done real damage. There are four strategies, they are all correct somewhere, and a site running one strategy across everything is wrong nearly everywhere: over-maintaining the cheap and replaceable while under-attending the few things that would genuinely hurt. The decision belongs to the asset, not to the site or the year. This page is the four and the questions that pick between them.
Run to failure, chosen deliberately
Right for assets that are cheap, held in spares, quick to swap and whose failure stops nothing and hurts nobody. What makes it professional rather than negligent is that somebody decided it, wrote it against the asset with the reason, and would revisit it if the consequence changed.
Time-based, the default and the blunt instrument
Fixed calendar intervals. Right where degradation follows time and weather rather than use, and right as a starting position where you have no better information. Its weakness is that it treats a duty pump and its standby identically, which is exactly backwards.
Usage-based, where degradation follows work
Running hours, cycles, throughput. Better than calendar for most rotating and cycling plant and it costs something real: somebody has to collect the readings. A usage interval with nobody reading meters is a calendar interval that is lying about itself.
Condition-based, expensive and narrow
Vibration, thermal, oil analysis: measure the thing and act on the measurement. Genuinely superior on high-criticality rotating plant and hard to justify anywhere else, because the sensors, the analysis and the skill to read it are all real costs. It is a strategy for the top of your criticality ranking, not for the site.
Questions people ask about maintenance strategies
Which strategy should most assets be on?
Time-based or usage-based, with a handful run to failure by decision and a handful on condition monitoring. A site where everything is on one strategy has not made the choices.
Is run to failure ever safe?
Where the failure has no safety, production or compliance consequence, yes. Where it has any of the three, no, and that is the test rather than the asset's price.
How do we move from time-based to usage-based?
Start collecting readings on the assets where it would matter most, and only convert intervals once you have a few months of data. Converting first produces intervals based on guessed usage.