industrial//maintenance//preventive maintenance

Preventive maintenance is the maintenance strategy that replaces or overhauls a part at fixed intervals of time, operating hours or cycles, whatever its condition, and it is the default for parts whose wear is predictable: filters, belts, brake pads, the oil of an engine, the bearings of a fleet with a well-known life. It needs no sensors and no models, only a calendar and a statistic of how long such parts last, which is why it is the first strategy most plants adopt beyond running to failure.

Preventive maintenance is the maintenance strategy that replaces or overhauls a part at fixed intervals of time, operating hours or cycles, whatever its condition, and it is the default for parts whose wear is predictable: filters, belts, brake pads, the oil of an engine, the bearings of a fleet with a well-known life. It needs no sensors and no models, only a calendar and a statistic of how long such parts last, which is why it is the first strategy most plants adopt beyond running to failure.

Whether it helps depends entirely on how the failure rate changes with age, the hazard function. If old parts fail more often than new ones (wear, fatigue, corrosion), replacing a part before it reaches the steep part of its life prevents failures. If the hazard is constant, failures are random in time: a part with 10,000 hours is exactly as likely to fail tomorrow as one installed yesterday (exponential distribution), and replacing it buys a new part with the same risk. With a falling hazard, early defects dominate and replacement makes things worse, swapping a part that survived its weak period for one just entering it.

The failure rate decides whether preventive replacement makes sense.

With a constant rate, changing healthy parts only spends spares; only a Weibull shape parameter above 1, with lives that are not too scattered, justifies a calendar.

The scatter matters as much as the trend. If wear-out lives range from 2,000 to 12,000 hours, a fixed interval short enough to prevent most failures throws away most of each part's life; that waste is what condition monitoring recovers, by replacing each part when it, rather than the average part, shows wear.

The interval comes from an explicit trade: the cost of each planned replacement against the cost of each failure in service times its probability before the interval ends, read from the fitted life distribution. Fitting that distribution correctly requires the units that have not failed yet (censored data).

Airline data showed that age-related wear-out is the minority pattern for complex components (bathtub curve), which is why reliability-centred maintenance moved much aviation maintenance away from hard time limits; where it does apply, a well-computed calendar is a fine choice before anything predictive (maintenance).