Executive summary
- PdM measures condition while the machine runs, to forecast failure.
- P is where deterioration becomes detectable; F is where the machine can no longer do its job.
- The gap between them is all the planning time you will ever get.
- Oil analysis and vibration typically give the longest P–F intervals.
- PdM pays when the asset is critical, downtime is expensive, and the plant can actually respond in time.
Predictive maintenance measures the condition of a machine while it is still running, and uses the trend in those measurements to forecast when it will fail. It differs from preventive maintenance, which replaces a component on a time cycle whether or not the component still has life in it.
P and F
Equipment failure is almost never instantaneous. The bearing that seized on Tuesday started deteriorating weeks earlier. P is the first point at which that deterioration can be detected by some instrument. F is the point at which the machine can no longer perform its function — which may arrive well before the component physically breaks.
The P–F interval is everything. If the technique gives you three months, you can order the part, plan the labour, and wait for a scheduled shutdown window. If it gives you two days, you simply know two days in advance what is about to happen — better than nothing, but it changes almost no plans.
The main techniques and what they buy
| Technique | What it detects | Typical P–F interval |
|---|---|---|
| Oil analysis | Wear particles, contamination, lubricant degradation | Months |
| Vibration analysis | Bearing deterioration, misalignment, imbalance, gear wear | Weeks to months |
| Thermography | Loose electrical connections, failing insulation, overheating bearings | Weeks |
| Ultrasound | Air leaks, electrical arcing, under-lubricated bearings | Weeks |
| Human senses on a structured route | Noise, smell, leaks, unusual vibration | Highly variable, but by far the cheapest |
Do not dismiss the last row
A trained operator walking a defined route every shift finds more developing faults than an expensive instrument used twice a year. This is exactly why the autonomous maintenance pillar comes first.
When PdM pays
PdM is not the answer for every machine. Its cost is the instrument, the skill to interpret the readings, and the discipline to collect data consistently. Four conditions should hold together before you invest.
- The asset is genuinely critical — if it stops, the line stops, and there is no standby.
- The failure develops gradually rather than suddenly, so there is a P–F curve to catch.
- The cost of an unplanned stop is clearly higher than the cost of monitoring.
- The plant can actually respond within the interval: parts available, people available, a shutdown window reachable.
The fourth is the one that fails most often. Plants buy the vibration meter, collect good data, correctly detect the developing fault — and then wait six weeks for the spare, by which time the machine has failed anyway. The problem is not the instrument. It is step 4 of the planned maintenance pillar, which was never done.
What PdM gives, and what it does not
- Gives: emergency work converted to planned work, which costs far less in time and people.
- Gives: full use of component life, instead of discarding parts that still had months in them.
- Gives: numerical evidence of whether a machine modification actually worked.
- Does not give: root cause. The meter says the bearing is failing; it does not say why it fails every six months.
- Does not give: a substitute for basic condition. A machine that is dirty, loose and under-lubricated generates so much abnormal signal that nothing can be read from it.
Accurate prediction on a machine whose basic condition has never been restored is accurate prediction of a failure we created ourselves.