Preventive, predictive, condition based: which to use where

Preventive maintenance works to a schedule. Condition based maintenance works to a measurement. Predictive maintenance watches how a measurement is moving and acts before it crosses the line. Reactive maintenance waits for the failure. All four are correct somewhere, and picking per asset is the actual skill.

The test, in three questions

Ask these about one piece of equipment and the answer usually falls out.

• •What happens when it fails: is anybody hurt, is production stopped, or is it an inconvenience?
• •Does it fail gradually with a warning you could measure, or does it just stop?
• •Can you afford a sensor and somebody to read what it says?

Run to failure is a real strategy

For a light fitting, a tap washer or a cheap fan in a store room, waiting for the failure is the cheapest correct answer. Putting those on a schedule spends money to prevent an outcome nobody minds. The mistake is doing it by accident rather than on purpose, which means nobody has stock of the part.

Preventive is right for most things

Where failure matters and the wear is roughly predictable with use or time, a schedule wins. Filters, belts, lubrication, seals, batteries. It is also the only option where the law sets an interval, whatever the condition says.

Condition based needs somebody to read it

A pressure differential across a filter tells you when to change the filter better than any calendar. That only helps if the reading gets taken and something happens when it is out of range. A reading that goes into a system nobody looks at has cost you the sensor and bought nothing.

Predictive is worth it for a short list

Vibration, thermal imaging and oil analysis pay back on equipment where the failure is expensive and the warning period is long enough to act on. That is usually large rotating plant: chillers, big pumps, compressors, motors over a certain size.

Do the arithmetic before buying. What does one failure of this machine cost, how often does it happen, and what proportion would you actually catch? If that product does not clearly beat the cost of the programme, the money belongs somewhere else.

The order to do this in

Get the record right, then find out which equipment is hurting you, then instrument that equipment. Teams that buy sensors first end up with a data stream pointed at the same gap the paperwork was falling into.