What is preventive maintenance? Types, scheduling, and limits

What preventive maintenance means

Preventive maintenance (PM) is maintenance carried out to assess or mitigate degradation and to reduce the probability that an item will fail. This wording follows EN 13306, the European standard on maintenance terminology, which distinguishes it from corrective maintenance: maintenance carried out after a fault has been recognized, to restore the item to a state in which it can perform its required function^[1].

EN 13306 then divides preventive maintenance into two branches. Predetermined maintenance is done at established intervals of time or units of use, without first investigating the condition of the item. Condition-based maintenance starts from an assessment of the physical condition of the item, and the action depends on the result of that assessment. Predictive maintenance sits within the second branch: condition-based maintenance carried out following a forecast derived from repeated analysis of the item's degradation parameters^[1].

On the plant floor, "PM" is usually meant more narrowly. When a supervisor says "the compressor PM is next week", the meaning is almost always a fixed-interval task: change the oil, change the filter, check the belt. This article uses the broad definition from the standard, but most of the discussion concerns the fixed-interval kind, because that is what people usually mean when they talk about a PM schedule, and that is also where its limits are felt most.

Types of preventive maintenance by trigger

The most useful way to tell the types of PM apart is to look at what triggers a task. The trigger determines when a task falls due, what data must be available, and what kind of degradation the task can address. Three triggers are in common use.

Time-based (calendar)

Tasks are scheduled by the calendar: every week, every month, or every year. This trigger suits degradation that progresses with time whether or not the machine is used, such as lubricant aging in its reservoir or rubber hardening. It is also used for periodic inspections required by regulation. Its strength is that it is simple and easy to plan. Its weakness is that the calendar does not know whether a machine ran three shifts or sat idle all month.

Usage-based (operating hours or cycles)

Tasks are triggered by a counter: operating hours, number of cycles, number of shots on an injection molding machine, number of motor starts, or tonnage processed. This type is closer to how components actually wear, because the wear of a filter element, a bearing, or a mold depends more on workload than on the date. The condition is that the counter must be trustworthy, whether it comes from the machine's hour meter, the PLC, or disciplined operator records.

The following figures are illustrative. A compressor with a task every 2,000 operating hours that runs 20 hours a day will fall due roughly every 100 days. If production demand drops and the compressor runs only 10 hours a day, the same task falls due roughly every 200 days. A fixed calendar schedule does not capture this difference: it asks for the same work in both situations.

Condition-based

Tasks are triggered by the result of an inspection or measurement: filter differential pressure crossing a limit, vibration rising, or bearing temperature departing from its usual level. The check itself is still scheduled, but the corrective action waits for the condition. Moubray stresses that such checks must be more frequent than the P-F interval, the span between the point at which an impending failure can first be detected and the point at which the item loses its function^[2]. The article on predictive maintenance covers this in more depth.

Autonomous maintenance: the operator's role in PM

Not all preventive work has to be done by technicians. In Total Productive Maintenance (TPM), Nakajima makes autonomous maintenance (jishu hozen) one of its core activities: operators help maintain their own machines through daily work such as cleaning, inspecting, lubricating, and tightening bolts^[3].

The reasoning is practical. The operator is next to the machine for the whole shift. The operator is the first to hear a change in sound, see oil dripping on the floor, or feel an unusual vibration. In this framework, cleaning a machine also means inspecting it, because leaks, loose bolts, and cracks are easiest to see when the surface is clean. Nakajima describes autonomous maintenance as a step-by-step program, starting with initial cleaning and continuing until operators can carry out routine inspections on their own^[3].

This division of labor frees technicians for work that needs specialist skills and tools. The consequence is that every task in a PM plan must name its owner explicitly: operator, mechanical technician, electrical technician, or contractor. A task that is considered everyone's job tends to be done by no one.

How to build a PM plan

A PM plan is a list of tasks that answers a few simple questions for every asset: what is done, how, when, by whom, with what materials, and how the result is recorded. If any of these goes unanswered, the task will be delayed, done differently by different people, or completed without leaving any useful data. The elements of a PM plan are as follows.

• Asset list and criticality. Start from a hierarchical asset register, then assess the impact of each asset stopping on production, safety, and quality. Critical assets get the most careful analysis.
• Task. Word it as an action that can be checked, for example "record the inlet air filter differential pressure", not "check the compressor".
• Method. Visual, measurement with an instrument, functional test, or replacement. The method determines the skills and tools required.
• Trigger and interval. Calendar, operating hours, cycles, or condition. Initial intervals usually come from the manufacturer's manual and are then adjusted to experience and the asset's history in your own plant.
• Owner. Operator, technician, or contractor, with an estimated duration so the workload can be planned.
• Checksheet. A uniform form with a written acceptance limit for every measurement, so that two different technicians judge the same condition in the same way.
• Spare parts and materials. Filters, oil, gaskets, or bearings that must be on hand before the task starts. A PM delayed because the parts have not arrived is a PM that was only half planned.

The table below shows the shape of a PM plan for a single rotary screw air compressor. Its content is illustrative, to show the structure, not a technical recommendation. The actual intervals and limits come from the manufacturer's manual for that compressor and are then adjusted to field experience.

Note that one asset can carry all three types of trigger at once. The operator's daily check is time-based, the oil change is usage-based, and the filter change can be condition-based if its differential pressure is monitored. A good PM plan does not pick one type for the whole asset; it picks the most suitable trigger for each failure mode.

The PM cycle: from plan to review

A PM plan on paper is only half the work. A PM program runs as a repeating five-stage cycle, and a weakness at one stage spills over into the next.

Scheduling places tasks in real time slots. Tasks that require the machine to stop must be coordinated with the production schedule, for example combined with a product changeover or a line's day off. This is where many PMs slip: the schedule is made by maintenance, but the machine is controlled by production.

Execution and recording should happen in the same place. A technician who fills in the checksheet beside the machine records the number they read; a technician who fills it in at the end of the shift records the number they remember. Recording must also keep the findings, not only check marks. Measured values recorded over time form a trend, while an "OK" mark does not.

Review closes the cycle. Recurring findings, tasks that never find anything, and failures that still occur between two PMs are the material for changing the plan. Without this stage, a PM program only repeats itself.

PM compliance as a KPI

The most common indicator of a PM program's discipline is PM compliance. The Society for Maintenance & Reliability Professionals (SMRP) defines it as the ratio of PM work orders completed within a defined time window to PM work orders scheduled^[4].

The following figures are illustrative. In one month, the utilities area has 48 scheduled PM work orders. Of these, 42 are completed within their window, 4 are completed late, and 2 are not done. PM compliance = 42 ÷ 48 × 100% = 87.5%. The four late work orders do not count as compliant, even though the work was eventually done.

PM compliance measures whether the schedule is carried out, not whether the schedule is right. A program full of unnecessary tasks can reach high compliance while still wasting labor hours. Compliance can also look good on paper if work orders that could not be done are rescheduled before they fall due. That is why this indicator is best read alongside failure data: are failures on the assets under PM actually decreasing?

The limits of preventive maintenance

Fixed-interval PM rests on one assumption: components become more likely to fail as they age, so replacing them before a certain age will prevent failure. The assumption is not wrong, but it holds more narrowly than is often imagined.

The study by Nowlan and Heap at United Airlines, which became the foundation of Reliability-Centered Maintenance (RCM), found that only about 11 percent of the components observed showed a clear relationship between age and failure^[5]. The rest failed at random or were most vulnerable early in their service life, and this early-life pattern was the one found most often^[5]. For components like these, there is no right replacement age, because their risk simply does not rise with time.

Moubray draws a less comfortable conclusion: intrusive work, such as dismantling, reassembling, or replacing components that are still working, can introduce early failures into a system that was previously stable^[2]. In the workshop this looks entirely ordinary: bolts tightened unevenly, a seal fitted askew, or dirt getting in while a gearbox cover is open. In other words, over-maintenance is not just a waste of labor hours and spare parts. It can add risk.

This does not mean fixed-interval PM should be abandoned. Moubray explains that scheduled replacement or restoration is technically feasible when a failure mode has an identifiable age at which the probability of failure rises sharply, and most items survive to that age^[2]. Mechanical wear that scales with usage often meets that condition. For hidden failures, such as a safety valve or protective switch whose failure only becomes apparent when it is needed, periodic functional testing is still required^[2].

Reviewing and optimizing a PM program

PM programs often grow by addition. Every major breakdown produces a new task, and tasks are rarely removed. The result is a long schedule, partly overlapping, and partly of unknown origin because the people who added the tasks have since moved to other roles.

RCM offers a structured way to reorganize a program like that. Moubray frames it as a series of questions about each asset: what its functions are, how it can fail to fulfill them, what causes each failure, what happens when each failure occurs, how much each failure matters, what can be done to prevent or detect it, and what should be done if no suitable task can be found^[2]. The answers determine the type of task for each failure mode, including a deliberate decision to let some components run to failure.

Smith and Hinchcliffe describe how to apply RCM to plants already in operation, with the existing PM program as the point of comparison: the tasks produced by the analysis are compared with the current tasks, and a decision is made on which to keep, change, add, or remove^[6]. This approach is more realistic for a plant that cannot build its program from scratch.

Not every plant needs to run full RCM on every asset. For non-critical assets, a simple review is often enough. The following questions can be asked of every task on the schedule.

• What failure mode does this task prevent or detect?
• Has this task ever found anything in its history?
• Do failures still occur between two executions of this task?
• Could condition monitoring replace replacement by calendar or operating hours?

Steps to build a PM program

The sequence below can be used to build a new PM program, or to reorganize one already in place, starting from a single area with critical assets.

1. 1Build the asset register and set criticality. Assess the impact of each asset stopping on production, safety, and quality, then start with the assets that have the greatest impact.
2. 2Gather material for each critical asset. Manufacturer's manuals, failure history from the CMMS or the workshop logbook, and the knowledge of senior technicians.
3. 3Define tasks per failure mode. For each way the asset usually fails, choose the appropriate trigger: time, usage, condition, or functional test. If no task makes sense, decide deliberately to let it run to failure.
4. 4Assign owners, durations, and materials. Split tasks between operators and technicians, then make sure spare parts are on hand before each task falls due.
5. 5Create checksheets with acceptance limits. Every measurement needs a written limit and a place to record its actual value.
6. 6Schedule together with production. Agree on machine downtime windows for tasks that require a stop, before the schedule is issued.
7. 7Measure and review. Track PM compliance and failures between PMs, then review the plan periodically: remove tasks that serve no purpose, change intervals that do not fit, and add tasks only for failure modes that have been analyzed.

Common mistakes in PM programs

• Copying the manual without adjustment. Manufacturer's manuals are written for general operating conditions. A plant with dusty air, a three-shift load, or machines that rarely run needs to adjust intervals based on its own experience.
• Tasks that cannot be checked. "Check machine condition" does not tell the technician what to measure or when a result counts as abnormal.
• Checksheets that hold only check marks. An "OK" mark does not store a trend. Measured values recorded over time let a decline in condition show before it becomes a failure.
• Adding a PM after every breakdown. Without root cause analysis, a new task may not target the actual failure mode, and the schedule keeps swelling.
• Chasing PM compliance as a goal. High compliance built on meaningless tasks only moves the waste into a tidier report.

Conclusion

Preventive maintenance is the effort to act before a failure occurs, and its trigger can be time, usage, or machine condition. Its value is not set by the length of the schedule but by the precision of each task: whether it targets a real failure mode, with a suitable trigger, carried out by a clear owner, and leaving data that can be reviewed.

The findings of Nowlan and Heap are a reminder that most failures cannot be prevented simply by replacing components by the calendar^[5], and that unnecessary teardowns can add risk. A healthy PM program therefore always moves in a cycle: planned, carried out, recorded, then reviewed and pruned. For a plant that is just starting, one area with critical assets and checksheets that record measured values is a sufficient starting point.