Lube oil analysis: five numbers that predict a bearing failure
Oil sampling is the cheapest condition-monitoring instrument on the plant. Five numbers, read together, will tell you a bearing is going before vibration does.
A routine oil sample is a few hundred dollars and a week of turnaround. A journal bearing failure on a gas turbine rotor is an unplanned outage measured in months. The gap between those two numbers is why the sample programme deserves more attention than it usually gets — provided someone reads the results as a set rather than as five separate pass/fail flags.
1. Wear metals, and specifically the ratio
Iron alone is uninteresting: it is everywhere in a steel system. Iron rising faster than copper points at a journal surface or a gear mesh; copper rising faster than iron points at a babbit-lined bearing. The ratio carries more information than either element on its own.
2. Particle counts against ISO 4406
A clean turbine oil sits around 16/14/11 or better. A jump of more than one code in a single interval usually means a filter bypass, a breather failure, or a new source of debris such as a degrading seal.
3. Water content
| Measurement | Normal | Investigate | Act |
|---|---|---|---|
| Water (Karl Fischer) | < 100 ppm | 100–250 ppm | > 250 ppm |
| Iron trend | stable | +25% per interval | doubling in one interval |
| Copper trend | stable | +25% per interval | doubling in one interval |
| ISO 4406 cleanliness | 16/14/11 | 17/15/12 | 19/17/14 |
| Viscosity @ 40 °C | ±5% of fill | ±10% | > ±10% |
4. Viscosity drift
Viscosity moving up means oxidation or coolant ingress; moving down means fuel dilution or shear of a multi-grade. Both change the film thickness in the bearing, and neither is visible from the gauge on the reservoir.
5. Acid number against the new-oil value
Track the rise, not the absolute figure. An oil at 0.35 mg KOH/g when the fill was 0.05 is oxidising faster than an oil at 0.40 that started at 0.30. When the rise accelerates alongside a water finding, the right answer is usually a reservoir flush rather than an additive top-up.
We caught a bad seal on a lube-oil cooler from a single potassium reading. That is a $2,000 repair against a $400,000 bearing.
Reliability engineer, combined-cycle plant
Keep every sample record with the unit's history and never discard an interval from the trend because the unit was shut down. Partial intervals through a start-up are exactly where contamination enters the system, and removing them flattens the very signature you are looking for.
- Lube oil
- Bearings
- Vibration
