What it is
Condition-Based Maintenance (predictive maintenance): instead of intervening by calendar, a signal that precedes functional failure is monitored (vibration, temperature, oil particles, performance trend, acoustic emission) and intervention happens when the condition crosses the threshold — inside the P-F curve window (the interval between the detectable failure and the functional failure).
When it is the right decision (Framework B)
Three simultaneous conditions (Moubray, ch. 7; SAE JA1011):
- A detectable P-F exists — the failure mode emits a measurable signal before compromising the function;
- The P-F interval is practicable — there is time to detect, plan and act (inspecting every P-F/2 guarantees at least one reading inside the window);
- Detection is reliable — the chosen technique actually sees point P on that machine and operating condition.
It applies both to Wear-out with P-F and to Random with P-F failures (IT-MNT-001's "Random with detectable P-F" category): what CBM demands is the advance warning, not a predictable age.
Periodicity and pitfalls
- Periodicity = P-F/2 — of the P-F interval of the technique in use: ultrasound sees point P before spectral analysis, which sees it before temperature; changing technique changes the P-F and therefore the route.
- Baseline first: without the healthy-machine reference (vibration signature, acceptance performance curve), the alarm threshold is a guess.
- CBM does not fix causes: detecting early buys planning time — recurrence demands root-cause analysis (the canonical example is cavitation: CBM detects it, but the correction is operational/design).
On the platform
The strategy selector walks these questions for any failure mode; the P-F calculator sizes the route from the estimated interval.