Maintenance Task Classification

Why the terminology needs clarifying

Maintenance is described using a long list of overlapping terms — preventative, corrective, predictive, proactive, scheduled, unscheduled, condition-based, reliability-centred. The international standard BS EN 13306:2017 defines much of this vocabulary and gives a taxonomy of maintenance types and activities, but it isn't complete: it doesn't fully address defect elimination or failure-finding tasks, and it doesn't treat "no scheduled maintenance" as the deliberate, justified decision it actually is. The taxonomy below fills those gaps, and ties each task type to the failure pattern and scheduling logic that make it applicable — which matters, because it's what lets you actually measure whether a task is working.

Preventative maintenance

Preventative tasks repeat at a periodic interval, and that interval can be set three ways: calendar-based (useful for round-the-clock operation or failure modes driven by environmental exposure), based on operational usage such as elapsed hours or cycles, or a hybrid of the two — for instance, a mix of exposure and operating age.

Scheduled restoration

A periodic intervention that restores a component's condition to a nominal value, applied while degradation is still gradual. Cleaning accumulated carbon dust from the brushes of rotating electrical machinery is a typical example — the dust reduces insulation resistance in the windings over time, and the cleaning task restores it.

Scheduled replacement

A periodic swap-out of a component for a new or refurbished one, used when the probability of failure rises sharply after a period of useful life and the component's condition can't be measured or restored in place. This only makes sense for genuinely wear-out failure patterns; applying it to components that actually fail randomly or prematurely is a common and costly mistake, since it adds cost without adding reliability. Filters that can't be restored in service and are swapped after a fixed period are a straightforward example.

Condition-based maintenance

Condition-based maintenance doesn't prevent failure — it detects the onset of failure, or tracks gradual deterioration, early enough that the worst consequences can be avoided. It splits into two forms.

Predictive maintenance relies on fixed sensors and continuous time-series data, so there's no fixed schedule — monitoring and analysis run constantly, with high automation and low ongoing cost once the initial investment is made. It's used both to diagnose an initiated failure (a bearing shock-overloaded and now showing up in vibration and temperature data) and to track long-term degradation with no distinct starting event, such as fatigue cracking that only becomes measurable once it reaches the surface, or accessible corrosion. Once diagnosed, the data also supports prognosis — estimating the remaining useful life, which can sharpen as new symptoms develop over time.

Scheduled condition-based tasks — inspections, sampling, NDE/NDT — are timed not by useful life but by the P-F interval: the gap between the point a potential failure can first be diagnosed and the point of functional failure. Convention sets the periodicity at half the P-F interval for non-safety failure modes, and a third for safety-critical ones, leaving enough remaining useful life to plan and act. This approach works across all failure patterns and is the only practical option for genuinely random failures. For premature failures, root-cause analysis should be the primary response, with condition-based maintenance used as a stopgap until the underlying cause is eliminated.

Failure-finding tasks

These matter most for protective devices — alarms, interlocks — that can fail silently between normal operation and maintenance activity. If a protective device has failed at the same time as the equipment it protects, the consequences of the combined failure can be severe, which is exactly the scenario a smoke detector failing during a house fire illustrates. Periodicity is set from the joint probability of both failures occurring together, reduced to whatever level an organisation's safety policy accepts, and the tasks themselves are usually simple functional checks — a test button, in the smoke detector case.

Failure-finding tasks are frequently under-recognised, particularly where physical safety checks sit under a different management system from the CMMS and so don't get counted as maintenance at all. That's a real risk: a periodicity that gets stretched because "it rarely finds anything" can quietly erode a safety margin nobody's tracking.

Servicing and zonal checks

Frequent, often daily or per-shift, these cover replenishing consumables, lubrication and cleaning — the last of which matters more than it sounds, since neglected cleaning raises contamination and fire risk in machinery spaces. Zonal checks, where experienced operators walk sections of plant using their own senses and judgement, catch a surprisingly high proportion of incipient failures. They're rarely captured in a CMMS or accounted for financially, which means cutting the workforce that performs them can produce an unreliability hit that shows up only after the fact.

No scheduled maintenance

Deciding not to schedule maintenance on a component is itself a legitimate, active decision — provided it's backed by analysis showing the failure risk is low and no practical, cost-effective maintenance task exists. Where the risk is unacceptable and no task is practical, the right response is usually a design modification rather than adding maintenance for its own sake.

Corrective maintenance

Deferred corrective maintenance applies where a failure has occurred but the consequences are small enough that operations can continue, at reduced performance if necessary, until recovery can be planned — a failed piece of redundant equipment is a typical case, and this work is often folded into existing planned outage windows. Immediate corrective maintenance applies where recovery is urgent because the consequences of continued operation are too severe to defer — a ship losing propulsion in a storm is the clearest example. Components that need urgent recovery are often better candidates for predictive maintenance instead, where the warning comes early enough to plan around.

Modification

Sometimes the right response to a recurring failure isn't a maintenance task at all, but a design change — to the component itself, to how it's operated, or to its environment. Suppressing dust and smoothing haul road surfaces on an open-cast mine site to reduce vibration stress on mobile equipment is a practical example of modifying the operating environment rather than the asset.

Defect elimination

Sometimes called proactive maintenance, this is the ongoing effort to find and remove preventable common causes of failure, rather than continuing to maintain around them. It runs two ways: root-cause analysis on any component showing a premature or random failure pattern where a wear-out pattern was expected, or a broader quality-management effort looking at how operations, maintenance and logistics handling could be improved — the kind of thinking behind a Total Quality Maintenance approach.

Classifying tasks this way isn't academic — it's what tells you whether a periodicity is set on sound logic or on habit. And once tasks are classified, the next question follows naturally: which failure pattern is each component actually in? Answering that reliably is where Weibull analysis inside IronMan® comes in.