If vibration analysis is the stethoscope of predictive maintenance, oil analysis is the blood test. The lubricant circulates through every moving surface in the machine, picks up the wreckage of wear, the contamination that gets in, and the chemistry of its own degradation — then carries that evidence to a sample bottle. A single oil sample, read correctly, tells you which component is wearing, what is getting into the oil, and whether the oil itself is still fit for service.
Done well it catches gearbox and bearing failures weeks before vibration does, for around twenty to forty dollars a sample. Done badly it produces a monthly PDF of numbers nobody acts on. The difference is almost entirely in sampling discipline and knowing what the tests mean.
What Oil Analysis Detects
Three families of information live in every sample.
1. Wear metals — which component is dying
As surfaces wear, they shed microscopic metal particles into the oil. The element tells you the material; the material tells you the component.
- Iron — the dominant wear metal in most machines. Cylinders, gears, shafts, bearing races, liners.
- Copper — bronze bushings, thrust washers, bearing cages, valve guides.
- Lead — babbit-lined journal bearings. Lead in a journal-bearing machine is a flashing red light.
- Tin — often paired with lead in babbit, or in bronze (copper-tin alloy) components.
- Aluminium — piston skirts, aluminium bearings, some gear cases.
- Chromium — hardened surfaces — chrome plating, piston rings, some roller bearings.
- Nickel / Molybdenum / Manganese — alloying elements of steel; rising levels point to specific steel-alloy components.
A rising iron trend on a gearbox points at gears or bearings. Add copper and you suspect a bearing cage or bushing. Add lead and the journal bearing is the prime suspect. The combination is a fingerprint, not a single number.
2. Contamination — what is getting in
- Silicon — usually dirt / dust ingress (silica). Also present from silicone sealant or antifoam additive, so check the source before panicking.
- Sodium — often coolant ingress (glycol mixes carry sodium).
- Potassium — another coolant marker; sodium-plus-potassium together is a strong coolant-leak signature.
- Water — kills lubrication film, rusts surfaces, attacks additives. Measured by Karl Fischer for precision, or a quick crackle test for a yes/no.
- Fuel dilution — unburnt fuel in engine oil thins the viscosity and drops film strength.
- Soot — in diesels, high soot indicates combustion blow-by or an overdue oil change.
3. Lubricant degradation — is the oil itself still good
- Viscosity — the single most important property. Too thin (fuel dilution, wrong grade, shear) and the film collapses. Too thick (oxidation, soot, contamination) and flow starves.
- TAN (Total Acid Number) — rising TAN means oxidation and acidic by-products attacking surfaces.
- TBN (Total Base Number) — in engine oils, the reserve alkalinity that neutralises acid. Falling TBN means the additive pack is depleting; when it runs out, corrosion accelerates.
- Oxidation / Nitration — the oil chemically degrading under heat and oxygen. Rises with high operating temperature and extended drain intervals.
- Additive depletion — magnesium, zinc, phosphorus, calcium, barium are additive elements (detergents, anti-wear, dispersants). A dropping trend means the additive pack is being consumed.
The Tests — What to Order
A competent lab offers tiered test packages. The core set for most machines:
- Elemental (spectrometric) analysis — ICP or atomic absorption. Gives you the wear-metal and additive-element table. This is the backbone.
- Particle quantifier / ISO cleanliness code — counts particles by size range. Critical for hydraulic and turbine oils where cleanliness is everything.
- Water content — Karl Fischer for precision; crackle test for cheap screening.
- Viscosity at 40 °C (and 100 °C for engines) — the must-have.
- TAN and/or TBN — depending on oil type.
For deep root-cause work on a flagged machine, add analytical ferrography — prepares a slide of the particles and looks at them under a microscope. Shape and colour distinguish cutting wear, sliding wear, rolling-contact fatigue, and corrosive wear. It is the difference between "there is a lot of iron" and "there is severe sliding wear from a heavily loaded gear." Order it when elemental analysis flags a problem, not on every routine sample.
Sampling — Where Most Programs Die
The single biggest source of useless oil-analysis data is bad sampling. Garbage in, garbage out, and oil sampling has a dozen ways to be garbage.
- Sample live oil, not dead oil. Draw from a circulating line or a properly flushed sample valve, not the drain plug at the bottom of the sump where everything settles. Drain-plug samples over-report sediment and wear.
- Sample the same way, every time. Same valve, same point in the run cycle, same machine state (warm, recently running). A sample taken cold-after-weekend and one taken hot-under-load are not comparable.
- Flush the sample valve first. The first oil out carries the debris trapped in the valve, not the oil in the machine. Discard it.
- Clean bottles, clean hands. A dirty bottle or a funnel wiped with a dirty rag contaminates the silicon and iron readings before the oil even reaches the lab.
- Do not sample right after an oil change or a top-up. Fresh oil dilutes the evidence to near-zero. Wait until the oil has been in service long enough to carry information.
- Consistent interval. Monthly for critical assets, quarterly for the rest. Irregular sampling destroys the trend, and the trend is the whole point.
The One Rule: Trend, Don’t Threshold
Every lab report comes with alarm limits — "iron normal to 50 ppm." Those generic limits are a starting point, not a verdict. A gearbox that has run at 45 ppm iron for two years and suddenly jumps to 90 is a problem; the same machine sitting at 60 ppm rock-steady since new might be perfectly healthy for its whole life.
Trend over threshold. The rate of change, the shift from that machine's own baseline, is what predicts failure. A single high reading is a question, not a verdict — re-sample and confirm before you open the machine. A steadily rising trend is an answer: something is wearing, and you have weeks-to-months to plan the intervention.
This is also why sampling the same way every time matters so much: inconsistent sampling produces a noisy trend, and noisy trends hide real shifts.
Worked Example
A reduction gearbox on a conveyor shows iron at 30-35 ppm for eighteen months — its happy baseline. Copper sits at 4 ppm. At month nineteen, iron ticks to 48, copper to 9. Month twenty: iron 72, copper 16, and the lab flags a few large ferrous particles on the particle counter. Vibration on the gearbox is still within normal limits.
Read together: rising iron plus rising copper on a gearbox points at a bearing (copper = cage/bushing) generating ferrous debris (iron = race/rollers) that vibration has not caught yet because the defect is still early. The trend says: plan a bearing change at the next available outage, in the next few weeks, do not wait. Cost of catching it: two twenty-dollar samples and ten minutes of reading. Cost of missing it: a seized bearing, a shredded gearset, and an unplanned line shutdown.
Where Oil Analysis Sits in the Stack
Oil analysis is strongest where vibration is weakest and vice versa. Vibration catches mechanical looseness, imbalance, bearing fault frequencies, and misalignment — surface-motion problems. Oil analysis catches wear debris, contamination, and lubricant breakdown — material-and-fluid problems, often earlier in the failure progression. Wear-particle generation frequently precedes the vibration signature by weeks. The mature program runs both, plus temperature, and reads them together. (For how those signals combine into a work-order trigger, see our piece on remaining useful life.)
How OpexMX Handles It
OpexMX ingests oil-analysis lab results per asset, trends every element against that machine's own baseline (not generic limits), and raises an alert when the trend shifts — distinguishing "high but steady for years" from "jumping off baseline." A confirmed trend auto-opens a predictive work order against the right asset, pre-linked to the likely component, with the lab report attached. The result is the oil program most teams wish they had: numbers that turn into work orders instead of PDFs that nobody opens.
See how OpexMX turns oil-analysis results into trend-triggered work orders →