Reliability-centered maintenance sets intervals only after identifying the failure mode, its consequences, and a technically effective task, with initial periods derived from reliability data and then adjusted in service [S1][S4].
RCM is suitable for critical assets, age-related failures, hidden functions, and condition-monitoring applications; run-to-failure is valid when failure consequences and corrective costs remain acceptable [S3][S4].
From Failure Mode to Candidate Task
RCM begins with the asset function and its required performance, then maps functional failures, failure modes, effects, and consequences before selecting maintenance [S4]. The interval is not selected independently of the task because the same failure mode may justify a calendar replacement, an on-condition intervention, a failure-finding test, or no preventive work [S1].
The three preventive task families are time-directed, on-condition, and failure-finding. Time-directed work replaces or restores an item at a fixed age without regard to as-found condition; on-condition work responds to a measurable deterioration; failure-finding work tests a protective function that operators may not otherwise see [S1]. RCM can also select run-to-failure, redesign, or a combination of tasks when fixed maintenance does not effectively reduce the identified risk [S1][S3].
A useful output contains more than a date: it identifies the failure mode, task, rationale, interval, owner, and the data that will trigger review [S4]. A comparison framework for CBM, PdM, and RCM helps separate monitoring methods from the RCM decision process that selects the maintenance strategy.
Time-Directed Intervals and MTBF
For a time-directed task, a simple initial estimate divides total operating time by the number of failures, producing mean time between failures, or MTBF, and then converts that reliability measure into a candidate replacement or restoration period [S1]. The calculation is only a starting point because the distribution of failures, operating exposure, and consequence of replacing an item before failure all affect the economic result [S1].
The aviation evidence behind RCM found that only about 11% of analyzed failure cases had a clear relationship with component age, while approximately 89% were independent of a simple time interval [S4]. A separate NASA summary covering airline, Swedish, and U.S. These findings weaken the assumption that every task should have a fixed calendar period based on elapsed age.
The 11% figure is evidence against blanket age-based scheduling, not a universal interval for every asset. Historical operating time and failure counts are necessary inputs, but the selected period must still be tested against the failure mechanism, remaining useful life, and cost of premature replacement [S1][S5].
On-Condition Intervals and the P-F Curve
On-condition task intervals are based on the expected P-F interval, which is the time between the point at which a failure becomes detectable and the point at which it progresses to functional failure [S6]. A vibration route that reliably identifies a developing imbalance before damage reaches the critical threshold therefore has a different interval from a route whose signal is intermittent or poorly correlated with failure.
The P-F interval sets the practical opportunity for action, not a guarantee that a fault will be found before failure. If the detection capability is weak, a short sampling period alone may not prevent functional failure; the monitoring method, trigger threshold, response time, and shutdown logic must operate as one chain [S3][S6].
On-condition work is generally attractive when a condition indicator can be measured reliably and an action threshold can be defined economically [S1]. It is poorly suited to failures that develop too quickly for the detection route, functions that are not observable under normal operation, or assets without a defensible condition trend. A failure-finding task is more appropriate for hidden protective functions, such as testing an emergency generator or fire-alarm system [S1].
Failure-Finding and Hidden-Function Tests
Failure-finding intervals depend on how long a hidden function can remain unavailable without creating unacceptable consequences and how reliably the test exposes the condition [S1]. The test should detect the loss of a required function, such as an emergency power or alarm function, before that failure permits secondary damage [S1].
Three data points control the decision: the required test frequency, the probability that the test detects the failure, and the maximum acceptable undetected period. A routine that only proves the test mechanism operated does not necessarily prove the protected function worked, so the test scope and pass criteria must match the failure mode being controlled [S1].
Do not replace an interval with more frequent testing when the test cannot detect the relevant failure or when the consequences of continued operation are not reduced. The corrective action is to repair or replace the failed item, redesign the test if it is ineffective, or escalate the consequence assessment if the protection is critical [S1][S4].
Cost, Consequence, and Task Selection
RCM compares the expected cost of failure with the cost of the candidate task and its technical effectiveness, so the interval is a risk decision rather than an arithmetic exercise [S4]. High-consequence failures justify greater analytical effort and often shorter intervals, while low-consequence failures may support run-to-failure when repair and restoration remain practical [S1][S5].
The decision must consider safety, environmental, operational, and production effects, as well as standby capacity, repair duration, spare-part availability, and labor [S4]. A shutdown consequence can justify redundant equipment or a design change even when no maintenance interval can reduce the underlying failure rate [S3]. Maintenance cannot raise inherent reliability above the design level, so an inadequate design must be addressed rather than hidden behind more frequent overhauls [S3].
For time-directed work, the period should be evaluated against premature replacement, failure before replacement, and task cost. For on-condition work, the period should be evaluated against P-F timing, detection capability, and response capacity. For failure-finding work, the period should be evaluated against the allowed undetected period and test effectiveness. No single metric selects all three intervals [S1][S6].
Adjusting Intervals With Field Evidence
An RCM interval is an initial engineering value, not a permanent calendar commitment. A field review should compare planned interval, actual work performed, as-found condition, defects found, failures before task, and operating exposure [S2][S4]. Data that shows a task consistently finds no condition justifying intervention may indicate excessive frequency, although the analyst must also test whether the task is capable of detecting the target failure [S1].
Intervals should also be reviewed when failure data, operating load, duty cycle, repair method, or equipment configuration changes. The RCM program is intended to feed every remaining failure back into the analysis, and a task that does not prevent its assigned failure mode should be replaced, redesigned, or removed rather than repeated at the same period [S3][S5].
Symptom-to-action discipline is straightforward: if failures occur earlier than the interval, investigate the failure pattern and detection assumptions; if the task finds no deterioration, verify its sensitivity before lengthening the interval; if a task introduces damage, reassess the maintenance concept; and if consequences are severe but the task is ineffective, escalate to redesign or additional protection [S1][S4].
Governance and Record Requirements
SAE JA1011 defines the minimum criteria for a process to be called RCM, while SAE JA1012 guides its application; the U.S. Department of Defense uses MIL-STD-3034 for its RCM process [S3]. The RCM record should identify the function, failure mode, consequences, selected task, interval, task owner, and monitoring or review data, making the decision auditable rather than reducing it to a work-order frequency [S4].
For each interval, retain the calculation inputs, including operating time, failure count, selected model, assumptions, and source data. For on-condition tasks, document the P-F basis, trigger, detection method, and response requirement. For failure-finding tests, document the function tested, test scope, acceptance criterion, and allowed undetected period [S1][S6].
The next review point should be triggered by either a date or an event. Track the next planned review, early failures before interval, repeat no-defect findings, changes in operating exposure, and any task that fails to prevent its assigned failure mode. Those records provide the next defensible adjustment to the interval [S2][S4].