What Is Predictive Maintenance?
Published on
Predictive maintenance is maintenance based on the measured condition of the equipment: you monitor it while it runs and step in when the readings say a part is on its way to failing. It differs from servicing on a calendar, which replaces parts whether they need it or not, and from repairing after a breakdown. This article explains how it works, where it pays off and where it does not.
Predictive maintenance: how it works
The idea is simple. A machine that is about to fail usually gives warning: it vibrates differently, runs hotter, draws more current or wears metal into its oil. Predictive maintenance measures those signs regularly, compares them with a baseline from when the equipment was healthy, and raises a flag when the readings drift.
The common techniques are the following.
- Vibration analysis: for rotating equipment such as motors, pumps, fans and gearboxes. It picks up imbalance, misalignment and bearing wear.
- Thermography: an infrared camera shows hot spots in electrical panels, bearings and steam lines.
- Oil analysis: a sample from a gearbox or hydraulic system shows wear particles and contamination.
- Ultrasound: detects leaks in compressed air and steam, and bearings that are starting to fail.
- Motor current analysis: shows problems in electric motors without stopping them.
The types of maintenance
The most common classification has three types.
- Corrective: you repair the equipment after it has failed.
- Preventive: you inspect and replace parts on a schedule, to head off the failure.
- Predictive: you measure the condition and act when the data calls for it.
Longer lists, of four, five or seven types, split these further or add improvement maintenance, but they all grow out of these three.
When predictive maintenance pays off
It is worth the effort where the equipment is critical, where a stop is expensive and where the failure gives warning in advance. Large rotating equipment, key pumps and compressors, and anything that halts the line when it breaks, are typical candidates.
It is not worth it everywhere. For cheap equipment that does not hold anything else up, running it until it fails and replacing it is often the sensible choice.
How to start
- Rank your equipment by criticality. What stops the plant, what is dangerous and what is expensive to fix.
- Pick the technique that fits each failure. Vibration for bearings, thermography for electrical faults.
- Take a baseline while the equipment is in good condition.
- Set the limits at which someone must act, and who that someone is.
- Act on the readings. A measurement that nobody acts on is only a cost.
Will maintenance be replaced by AI?
No, though AI is changing how the data is read. Software can analyse large numbers of readings and flag the ones that look abnormal, which saves time. But a bearing has to be changed by someone, a pipe has to be welded and a structure repaired, and that remains physical work.
Where we come in
Predictive maintenance tells you what to fix and when. Someone still has to do the repair, and often that means maintenance technicians, fitters and welders. We provide technicians for industrial maintenance in Spain and Portugal, on your site or as work to order. Tell us what you need.