The P-F curve plots an asset's condition falling as a failure develops over time. Point P is the potential failure, the first moment the developing failure can be detected. Point F is functional failure, the moment the asset can no longer do its job. The gap between them, the P-F interval is the warning window you have to catch the problem and act.
That window is the whole reason condition monitoring works, and it is the number that should set every inspection frequency in a plant. Inspect faster than the window and you catch developing failures with time to plan a repair. Inspect slower than the window and you will keep being surprised, no matter how much monitoring gear you buy. This guide explains the curve, where it came from, how detection technologies sit on it, and how to turn a P-F interval into an inspection schedule.
What is the P-F curve?
The P-F curve is a graph of condition against time for a single failure mode. Early in the mode's development the asset is healthy; as the failure progresses, a crack grows, a bearing spalls, a winding degrades, condition declines, slowly at first and then steeply. The curve marks two points on that decline. Point P is where the degradation first becomes detectable by some technique. Point F, further down, is functional failure: the asset has stopped meeting its required standard of performance.
The key insight is that P comes well before F. A bearing does not go from perfect to seized in an instant; it announces itself, through ultrasonic emissions, then vibration, then heat and noise, for some period before it actually fails. The P-F curve is the picture of that announcement, and the P-F interval is how much notice it gives.
What is the P-F interval, exactly?
The P-F interval is the elapsed time between the potential failure (P) and functional failure (F) for a given failure mode. It is measured in whatever unit fits the mode, hours, days, cycles, or starts. A slowly propagating fatigue crack might have a P-F interval of months; a lubrication failure on a high-speed bearing might have one of days. The interval is a property of the failure mode and the detection method together: a more sensitive technique detects P earlier, which lengthens the usable window.
One refinement matters in practice: the net P-F interval. Detecting P is not the same as fixing the problem. You still need time to diagnose, plan, get parts, and schedule the repair. The net P-F interval subtracts that response time from the raw interval, leaving the time you can actually spend monitoring before you must act. If the P-F interval is 30 days but it takes 10 days to get the part and a window, your net interval is 20, and your inspection frequency has to fit inside that, not the full 30.
Where does the P-F curve come from?
The P-F curve comes from Reliability-Centered Maintenance the 1978 report by F. Stanley Nowlan and Howard F. Heap of United Airlines, sponsored by the U.S. Department of Defense and released publicly that December (report AD-A066579). Their study of airline maintenance data reshaped the field: it showed that most failures are not age-related, so replacing parts on fixed schedules often does nothing, and that the productive alternative for many failure modes is to detect the onset of failure and act inside the warning window. The P-F curve is how they described that window.
That report is also the source of the modern task taxonomy still used today: an RCM program has essentially four task types, inspect to detect a potential failure (on-condition tasks, the P-F idea), restore or rework before a maximum age, discard before a maximum age, and check for hidden failures that have already occurred. On-condition monitoring, riding the P-F interval, is the one that made condition-based maintenance and predictive maintenance possible.
How do different detection methods sit on the curve?
Different technologies detect the same failure at different points, so they give different amounts of warning. On a developing bearing failure, ultrasonic and vibration methods pick up the earliest microscopic signs; wear debris shows up in the oil somewhat later; then heat, then audible noise a human can hear, then the bearing is hot to the touch, and then it fails. Each method sits at its own point on the curve, with its own lead time.
| Detection method | Detects | Relative warning |
|---|---|---|
| Ultrasonic / acoustic emission | Earliest friction and micro-defect energy | Longest |
| Vibration analysis | Bearing and rotating-element defects | Long |
| Oil / wear-particle analysis | Metal debris shed into the lubricant | Long-to-moderate |
| Thermography | Abnormal heat from friction or electrical fault | Moderate |
| Audible noise / feel | Late-stage noise, vibration, heat | Short |
| Smoke / seizure | Failure in progress | None |
This is also why oil analysis earns its place: wear debris shows up on the curve well before you can hear or feel the failure, and reading the wear metals in the oil from a properly pulled sample, puts you near point P rather than near point F.
How do you set an inspection interval from the P-F interval? Five steps
- Define the failure mode. The P-F interval belongs to a specific mode, not to a machine in general. "Bearing failure from lubrication starvation" has a different interval than "bearing failure from misalignment." Start from the modes that actually hurt you.
- Estimate the P-F interval. Use OEM data, failure history, published guidance for the mode, and engineering judgment. It is an estimate, and a rough one is far better than none.
- Pick the detection method that gives enough window. If the P-F interval on your chosen method is too short to act inside, move earlier on the curve, a more sensitive technique detects P sooner and lengthens the window.
- Set the inspection interval to no more than half the P-F interval. Inspecting at half the interval guarantees at least one inspection lands inside the window; many programs use one-third or less for critical assets to leave margin for a missed reading. Inspect at or beyond the full P-F interval and you will step right over failures.
- Subtract response time and validate. Use the net P-F interval, the window minus the time to plan and execute the repair, as the real budget, then check performance against outcomes and adjust. Catches inside the window prove the interval; failures between inspections say tighten it.
What does the P-F curve not tell you?
The P-F curve does not predict when a failure will start, only how much warning you get once it does. That is a crucial limit. Nowlan and Heap's own finding was that the large majority of failure modes are not age-related; they strike at random ages, a pattern the bathtub curve only partly captures, which is exactly why fixed-interval replacement so often fails to help. The P-F approach sidesteps that by watching condition instead of counting time, but it only works for failure modes that actually give a detectable warning. Some failures are effectively instantaneous and have no usable P-F interval, for those, the answer is redundancy, design changes, or accepting run-to-failure, not more inspection.
It also assumes the decline is orderly enough to catch. Real curves are noisy, P-F intervals vary sample to sample, and a method can miss P. That is why the interval carries margin (half or less) and why condition monitoring lives alongside, not instead of, a sound preventive maintenance schedule and honest MTBF tracking.
Why the P-F interval is the number that sets your schedule
Nearly every condition-monitoring frequency in a plant traces back to a P-F interval, whether or not anyone wrote it down. The concept is not academic: it is the analytical basis for choosing what to monitor and how often, and it comes straight from the founding Reliability-Centered Maintenance report by Nowlan and Heap, sponsored by the U.S. Department of Defense (Nowlan & Heap, 1978, AD-A066579). Getting the frequency right is where the money is: the U.S. Department of Energy's FEMP O&M guidance, maintained by Pacific Northwest National Laboratory, reports that condition-driven maintenance offers savings that can exceed 30-40% over reactive maintenance, with predictive programs adding 8-12% over preventive-only (PNNL, O&M Best Practices), but only when inspections actually fall inside the window.
Where the P-F curve fits your reliability program
The P-F curve turns condition monitoring from a collection of gadgets into a decision. It tells you which detection method to use (the one that finds P early enough to act), how often to inspect (inside the net P-F interval), and when a reading matters (when it shows the curve heading down). Feed it good inputs, clean oil samples disciplined routes, and it drives the whole equipment reliability program.
The practical failure is not understanding the curve; it is executing it: inspections due on the right interval, readings trended against the curve, and a finding at point P turned into a planned repair before point F. When routes live on paper and readings live in three systems, the window closes unnoticed. Connecting condition readings to the work-order that acts on them, so a reading near P becomes a scheduled job, and inspection compliance shows up on your maintenance KPIs and PM compliance is the pattern on our platform overview; the CLS case study shows it running in a plant.