Pump Reliability & Predictive Maintenance (PlusSmart) · 9 min read · Updated November 2026
The P-F Curve in Practice: Which Technology Sees the Fault First
The P-F curve is the one reliability diagram every maintenance manager knows and few use to choose instruments. Each monitoring technology detects a developing fault at a different point between P and F — oil debris and electrical signatures early, overall vibration later, heat and noise last. Putting numbers on the intervals changes what you buy and how often you look.
The P-F curve, from reliability-centred maintenance, plots a machine's condition against time. Point P is where a defect becomes detectable; point F is functional failure. The P-F interval between them is the time you have to detect, plan and act. The curve's lesson is simple: the earlier your technology sees P, the longer the interval and the cheaper the repair. Its second lesson is the one most programmes miss: your inspection frequency must be shorter than the P-F interval of the technology you use, or you will find the fault only in hindsight.
Typical P-F intervals for rotating equipment
| Technology | What it sees | Typical P-F interval | Notes |
|---|---|---|---|
| Oil wear-debris monitoring (online) | First metal particles from pitting gear teeth, bearing raceways | Months (3–9 for gearboxes) | Earliest for oil-lubricated equipment; sees surface fatigue before it changes vibration |
| Oil quality (online) | Water ingress, oxidation, additive depletion, wrong oil | Weeks to months | Precursor of wear rather than the wear itself |
| Electrical signature analysis (ESA/MCSA) | Rotor-bar cracks, stator faults, eccentricity, bearing and load faults via current modulation, supply unbalance | Weeks to months | Only method for electrical faults; measured in the MCC |
| High-frequency vibration / acceleration envelope | Bearing defects (race, ball, cage), gear mesh | Weeks to months (1–6) | Needs sensors close to the bearing with good mounting |
| Overall velocity vibration (ISO 20816) | Unbalance, misalignment, looseness, advanced bearing damage | Weeks (2–8) | What route-based programmes mostly measure |
| Drive data (torque, power, speed) | Operating-point problems: dead-head, dry running, cavitation, run-out | Seconds to minutes for events; weeks for trends | Prevents the damage rather than detecting it |
| Temperature | Bearing overheating, winding heating | Days (1–14) | Late; useful as a second vote |
| Noise, visual, smell | Severe damage, leaks | Hours to days | F is close |
| Oil laboratory sample (quarterly route) | Same as online oil, sampled | Whatever is left after a 90-day sampling interval | The interval problem: P can occur the day after a sample |
The intervals are ranges, not promises — a bearing with a lubrication failure can go from P to F in days whatever you measure. But the ordering holds across industries: for gearboxes and oil-lubricated bearings, debris first; for motors, current first; for pumps and fans with grease-lubricated bearings, acceleration first, velocity second, temperature last.
The inspection-interval rule
If a technology's P-F interval is eight weeks and you measure monthly, you may see P once before F — or not at all if P fell just after a round. The rule of thumb is to inspect at no more than half the P-F interval, ideally a third. That is the quiet argument for continuous or near-continuous monitoring: a wireless sensor reading every ten minutes has an inspection interval of ten minutes, so the full P-F interval is available for planning. Route-based programmes are not wrong; they are just limited to faults with long intervals, and they miss anything that develops between rounds — which is exactly what happens on pumps after a process upset.
Combining technologies: the "which first" question
PlusSmart's selection table by failure mode comes from this logic:
- Unbalance, misalignment, looseness — vibration (excellent), ESA (good via load modulation), drive data (limited).
- Bearing and gear wear — oil debris (excellent, earliest, for oil-lubricated), vibration (good), ESA (good).
- Oil degradation, water ingress — oil quality only.
- Rotor bars, stator turns — ESA (excellent), vibration (limited).
- Cavitation, dry running, dead-head — drive data (excellent: prevented by control), ESA (good), vibration (sees the consequences).
- Power quality, unbalance — ESA (excellent), drive (partial).
A critical gearbox-driven mill or compressor therefore wants oil debris plus vibration plus ESA; a pump fleet wants drive-based protection plus vibration, with ESA for inaccessible motors; a fan wants vibration plus ESA. No single sensor covers every failure mode, and buying one technology for everything leaves blind spots on the P-F curve.
From detection to prediction
Detecting P is the first half. The second is estimating where on the curve the asset is and how fast it is moving, which turns an alarm into a date: "replace this bearing in the next four to six weeks" is a maintenance plan, "high vibration" is not. That estimate comes from trend rate, the fault type (bearing defects accelerate; unbalance from build-up is linear), and experience with the machine class. PlusSmart Analytics ranks assets from good to bad with that remaining-time estimate and the cost of letting each one fail, which is what the planner needs to put the work into the next shutdown instead of the next emergency.
Applying it
List your critical machines, write the dominant failure mode next to each, choose the technology that sees that mode earliest, and set the measurement interval at a third of the P-F interval. Most plants end up with online oil on the few gearboxes, wireless vibration on the pumps and fans, ESA on the motors they cannot reach, and drive-based protection wherever a VFD exists. That is the PlusSmart toolkit, and it was assembled from this curve.
Need engineering help with this?
Our engineers apply this to real projects every week — selection, specification, commissioning. Send your question or a part number.
Quick answers
Is the P-F interval the same for every bearing?
No. It depends on the failure mechanism, load, speed and lubrication. Fatigue spalling on a well-lubricated bearing can give months of warning; a lubrication failure can give days. Use the ranges for planning and the trend rate for the actual decision.
Can temperature alone be used for condition monitoring?
Only as a last line. Temperature rises late on the P-F curve — typically days before failure — so it confirms a problem rather than warning of one. It is useful as a second vote alongside vibration or current.
Why does online oil monitoring beat quarterly samples?
Because P-F intervals for gear and bearing damage are often shorter than 90 days, a quarterly sample can miss the whole window. An online sensor reads daily, so the full interval is available for planning the repair.