Preventive or predictive maintenance: intervening on a calendar or on evidence

Comparison criteria

Criteria are stated explicitly and neither option is declared a winner: which one fits depends on the constraint that binds hardest in your operation.

Condition monitoring earns nothing until somebody acts on it

The common failure in condition-based programmes is not technical. Sensors are fitted, data accumulates, a dashboard exists, and an alert appears in an inbox belonging to nobody in particular. Weeks later the bearing fails and the trend is found afterwards in the history. What makes the approach work is the unglamorous part: a named owner for each alert, a defined response to each severity, a route from alert to a planned work order, and a review of what the data said before every failure that got through. Without that loop, monitoring measures the deterioration accurately and changes nothing about the outcome.

Not every failure announces itself in advance

Condition-based methods detect deterioration that develops over a period long enough to catch. A great many industrial failures do not behave that way: a component fractures, a control board fails, a contaminant enters, an operator error damages something. For these, monitoring provides no warning, and the useful responses are different — design changes, protective devices, spares held for rapid replacement, or accepting the failure and planning the recovery. Deciding which of your critical assets actually deteriorate detectably, and which simply stop, is the analysis that determines where a monitoring investment can return anything at all.

Most plants end up running several regimes at once

A sensible maintenance strategy applies different approaches to different equipment, and the differentiator is criticality combined with failure behaviour. Assets whose failure stops the plant and which deteriorate detectably justify monitoring. Assets that wear predictably and are cheap to service suit a fixed interval. Assets that are inexpensive, redundant and quick to replace can reasonably be run until they fail, which is a decision rather than a default. The work is going through the equipment list and deciding, with the reasoning written down, so that a successor can see why a machine is treated the way it is.

Frequently asked questions

What has to exist before condition monitoring can pay for itself?

A measurement that reflects the deterioration you care about, a baseline captured while the asset is healthy, enough history to distinguish a trend from ordinary variation, and someone with the skill and time to interpret it. Beyond that, you need a path from an alert to scheduled work, with authority to take the machine out of service. Programmes that install sensing without securing the interpretation and the response usually generate data that documents failures rather than preventing them.

How is a maintenance interval decided when there is no failure history?

Manufacturer guidance is the usual starting point, adjusted for how hard you actually run the equipment and the conditions it sits in, since published intervals assume conditions that may not match yours. From there the interval should be reviewed against what technicians find: components replaced with obvious life remaining suggest lengthening it, deterioration found at each visit suggests shortening it. Recording the condition at each intervention is what turns an assumed interval into an evidence-based one over time.

Does condition-based work remove the need for scheduled intervention?

No, and treating it as a replacement is a common overreach. Lubrication, cleaning, calibration, statutory inspections and tasks that restore a known condition still need to happen on a cycle regardless of what the data says. Monitoring changes when component replacement happens and gives warning of developing faults; it does not remove routine servicing, and it does not cover failure modes that give no advance signal. The regimes are complementary, applied to different tasks on the same asset.