What is the P-F curve in reliability-centered maintenance?
The P-F curve illustrates deterioration between the point when an incipient or potential failure becomes detectable (P) and the point when the required function is lost (F). The P-F interval is the available warning time for a specific failure mode and detection method under defined operating conditions. It is usually uncertain, not a fixed equipment-wide value, and must be compared with inspection frequency and the complete response lead time.
Define functional failure before measuring the warning interval.
P is the earliest detection state that a particular monitoring method can identify with usable reliability. F is the failure to meet the defined performance standard, not necessarily catastrophic destruction. An asset may still be running when it has already functionally failed its required accuracy, capacity or quality limit.
Different methods detect different stages: oil debris analysis, vibration, ultrasound, thermal measurement and human observation need not share the same P. Consequently, the P-F interval belongs to a failure mode, operating context and detection method together.
An interval shorter than P-F is necessary but not sufficient.
A frequently taught screening heuristic is to sample at a fraction of the shortest credible P-F interval, sometimes one half. This is not a universal standard. Detection probability, sensor sensitivity, variability, operating duty and the time needed to execute work determine an appropriate cadence.
Assess the complete path: degradation begins, a sample or alarm occurs, the finding is recognized, an acceptable decision is reached, materials and access are available, and corrective work restores or protects the function. The monitoring program fails if the organization detects P but cannot act before F.
| Time component | Planning question | Typical source |
|---|---|---|
| Detection delay | How long until the next usable sample or alarm? | Route interval or alarm system records |
| Diagnosis and confirmation | How quickly can an abnormal signal be validated? | Analyst backlog, repeat measurement history |
| Planning and approval | How long to authorize a safe work package? | CMMS readiness and risk approval workflow |
| Parts and access | Will materials and production access exist in time? | Procurement lead time and outage calendar |
| Intervention and restoration | How long to correct, test and return to service? | Job plans, actual duration and acceptance record |
Worked example: bearing deterioration with uncertain warning time.
Fictional example: historical cases suggest vibration abnormalities become detectable between 18 and 40 operating days before a bearing can no longer meet duty. A monthly route could miss the shorter events entirely. A seven-day route reduces the maximum sampling delay, but it is not by itself a sufficient maintenance strategy.
Suppose the site needs two days for assessment, ten for parts and scheduling and one to intervene. A 13-day response path plus a possible seven-day detection delay consumes up to 20 days, already greater than the lower bound of the illustrative 18-day P-F interval. The team should investigate earlier detection, stocking, access or an approved different policy. The values are hypothetical and do not prescribe inspection frequencies.
The observation that a bearing survived four weeks after an alarm does not prove every bearing has a four-week P-F interval. Some faults deteriorate non-monotonically or too rapidly for periodic inspection.
Avoid treating a single successful save as validation of an interval. Record missed detections, false positives, unplanned failures, degradation rate variation and executed corrective lead times.
Record what the chosen technique can and cannot see.
Vibration can detect some rotating-element defects but not all lubrication, electrical or sudden failure modes. Oil debris analysis can reveal some wear processes while missing others. A technically credible program maps each critical mode to an observable signal, method capability and threshold approved for the asset.
Where data are limited, use scenario ranges and evidence labels rather than reporting a precise 21-day interval. Compare severity and consequence with a 'no reliable warning' outcome before accepting monitoring as the control.
Turn findings into planned interventions and refresh the strategy.
Capture asset, failure mode, point of detection, method, trend, threshold, alert date, work authorization date, restoration date and observed condition. This creates real response-time evidence. Use PM and CMMS feedback to test whether the actual detection-to-functional-failure warning is still adequate.
Escalate to redesign or another approved policy when deterioration is not detectable early enough, the response capacity cannot be made fast enough, or the operating consequence is unacceptable. The visual curve does not override an OEM requirement or process-safety constraint.
What does a defensible P-F assessment decide?
Whether a specified detection method gives enough credible warning, at an economically and operationally feasible inspection cadence, for the facility to act before functional failure. Unknown P-F behavior remains an uncertainty, not permission for a nominal interval.
- Define F as loss of required function.
- Link P to a specific sensor or inspection method.
- Assess the entire response delay, not just the inspection route.
- Revalidate with both intercepted and missed failures.
Technical references
- NASA Reliability-Centered Maintenance Guide — Failure-based maintenance strategies, inspection principles and program review.
- SAE JA1011_202411: Evaluation Criteria for RCM — Evaluation criteria for genuine RCM processes.
Choose detection and response together.
Use site-specific findings and execution lead times to build realistic, evidence-backed warning coverage.
RCM learning path See facility-specific XRVE services