Take a 5.5 MW gas turbine generator where tin in the lube oil moves from 2 ppm to 11 ppm across two samples and the particle count drifts from 16/14/11 to 20/18/14, while the laboratory report sits unread because the machine is still running and still making load. That drift is enough warning to prevent a wiped journal bearing, if someone acts on it.
That is the whole argument for turbine lube oil analysis in one sentence. The oil carries the evidence of what is happening inside bearings, gears, seals and control valves weeks before vibration, temperature or a trip annunciator says anything. The cost of reading that evidence is a small, predictable annual sum. The cost of not reading it, on a machine of that size, is an order of magnitude larger once you count the repair and the weeks off line.
What turbine lube oil analysis actually measures, and what it cannot
Turbine oil does four jobs at once. It maintains the hydrodynamic film in journal and thrust bearings, it carries heat away from those bearings, it protects steel surfaces from rust and oxidation, and on most machines it is also the hydraulic working fluid for the governor and trip system. A single reservoir therefore feeds two very different duty requirements, and the tests fall into the same two families.
The first family looks at the oil itself: viscosity, acid number, oxidation stability, remaining antioxidant, water content, foaming and air release. These tell you whether the fluid can still do its job.
The second family looks at what the oil has picked up: wear metals, silicon, particle counts, ferrous debris and varnish potential. These tell you what the machine is doing to itself.
Be clear about the limits before you buy anything. Oil analysis will not measure babbitt thickness, bond integrity or bearing clearance, and it will not tell you which of four bearings is shedding metal unless you sample individual drain lines. It does not replace shaft position monitoring, vibration analysis or the physical inspections in your OEM schedule. What it buys you is lead time, and on a machine where an unplanned outage costs tens of millions of naira a day, lead time is the entire point.
If nobody at your plant currently holds a written oil specification, a sampling schedule and a set of action limits, that is the cheapest gap on the site to close. Request a technical proposal and we will scope a programme against your machines rather than a generic package.
The oil analysis tests that predict bearing failure
Bearings fail through a small number of mechanisms, and each one leaves a signature in the oil.
Viscosity is the first line. Turbine systems generally run ISO VG 32 or VG 46, and the film thickness that keeps babbitt away from a journal depends directly on viscosity at operating temperature. A drift of more than 5 percent from the new oil value deserves investigation and more than 10 percent is an action point. The usual causes in Nigeria are a top up with the wrong grade, dilution by fuel or water, or severe oxidation thickening the oil.
Water content by Karl Fischer titration is the one most plants underestimate. Free and emulsified water destroys film strength, accelerates oxidation and, on tin based babbitt, drives tin oxide corrosion that leaves a hard black scab on the running surface. Most turbine OEMs want water below 100 to 200 ppm. Gland steam leakage, cooler tube leaks and humid air drawn through a reservoir breather during load swings are the common routes in.
Elemental spectroscopy by ICP gives you the wear metal picture. Tin, antimony, lead and copper point at babbitt and bearing backing. Iron points at journals, gears, shafts or rust. Chromium and nickel suggest shaft coatings or alloy components. Silicon and aluminium usually mean ingested dust. Sodium and potassium suggest cooling water.
There is a trap here that costs plants real money. ICP spectroscopy only reliably detects particles below roughly 5 to 8 micrometres. A bearing that is shedding large fatigue flakes can return a comfortable elemental report while it is actively failing. That is why the wear metal result must always be read alongside a particle count, ferrous density or a patch test.
Particle counting to ISO 4406:2021 gives you the three number cleanliness code. Most turbine lube systems should be held at 16/14/11 or better, and governor and servo circuits usually want 14/12/9 or cleaner. A count climbing two codes over consecutive samples at unchanged filter condition means something upstream is generating debris.
Analytical ferrography is the follow up test, not the routine one. It puts the wear particles under a microscope and separates rubbing wear from cutting wear, fatigue spalling and severe sliding. When your trend has gone wrong and you need to know whether to plan an outage, this is the test that answers the question.
Worked example (hypothetical, continued): suppose the machine above had quarterly sampling in place and the samples were being taken correctly. What was missing was an owner for the report and a written action limit. Catching that tin trend at the second sample would mean a rotable shell exchange inside a short planned window. Missing it means a rotor lift, journal repair and bridging diesel, which costs several times more and takes far longer. The signals of an incipient wipe are covered in more detail in turbine bearing babbitt failure.
The tests that predict governor and control valve failure
This is the half of turbine lube oil analysis that most Nigerian plants skip, and it is where the expensive trips come from.
Governor servo valves, trip blocks and control valve actuators work with clearances measured in single micrometres. They do not fail because the oil got dirty in the ordinary sense. They fail because of varnish, a soft, sticky, insoluble degradation product that plates onto metal surfaces, closes those clearances and makes a spool stick.
Membrane Patch Colorimetry to ASTM D7843 is the test that quantifies varnish potential. A sample is filtered through a membrane and the colour of the deposit is measured as a delta E value. As a working guide, below 15 is normal, 15 to 30 warrants monitoring, 30 to 40 is abnormal and above 40 is critical. The number is temperature and handling sensitive, so the laboratory procedure must be consistent sample to sample or the trend is meaningless.
RPVOT to ASTM D2272 measures remaining oxidation stability against the new oil value. Below 50 percent of new is a caution point and below 25 percent is usually an action point. It is a useful long horizon test but it is slow and it can be blind to certain additive systems.
RULER, or linear sweep voltammetry to ASTM D6971, measures the remaining amine and phenolic antioxidant directly. It is faster than RPVOT and it shows depletion earlier, which matters because varnish forms once the antioxidant package is exhausted, not before.
Air release to ASTM D3427 and foaming to ASTM D892 deserve more attention than they get. Entrained air compresses under load, which makes a governor sluggish and unpredictable, and it also drives micro dieseling, where bubbles collapse adiabatically and locally cook the oil into varnish precursors. A rising air release time is an early warning on both counts.
Worked example (hypothetical): a 12 MW condensing steam turbine begins tripping on control valve response during load changes, several times over a few weeks. Oil analysis returns an MPC value of 42 and RPVOT at 22 percent of the new oil figure, and the oil has not been tested for either parameter since commissioning.
In that case, an ion exchange and fine filtration campaign with an antioxidant reformulation costs a fraction of a full charge replacement, and both routes together typically cost less than the trips have already cost in lost production. Work out your own figure using the method in plant downtime cost per hour before you decide what a test package is worth.
Test schedule and limits
The table below is a planning envelope for a mid sized industrial steam or gas turbine at a Nigerian site. Limits must be confirmed against your OEM manual and your new oil baseline, because a limit without a baseline is a guess. Costs move with laboratory, sample volume, turnaround time and the exchange rate.
| Test | What it catches | Guide caution / action point | Frequency |
|---|---|---|---|
| Viscosity at 40°C (D445) | Wrong grade, dilution, oxidation thickening | 5% / 10% change from new | Quarterly |
| Water by Karl Fischer (D6304) | Gland leaks, cooler leaks, humid ingress | 150 ppm / 300 ppm | Quarterly |
| Acid number (D664) | Oxidation, additive depletion | +0.1 / +0.3 mg KOH/g over new | Quarterly |
| Particle count (ISO 4406) | Debris generation, filter bypass | 17/15/12 / 19/17/14 | Quarterly |
| Elemental spectroscopy (D5185) | Bearing and gear wear, dust, coolant | Trend based, no fixed limit | Quarterly |
| RULER antioxidant (D6971) | Time left before varnish starts | 50% / 25% of new | Half yearly |
| MPC varnish potential (D7843) | Governor and servo valve sticking | delta E 30 / 40 | Half yearly |
| Air release and foam (D3427, D892) | Sluggish governor, micro dieseling | 10 min / 15 min air release | Annually |
| Analytical ferrography | Wear mode when a trend goes bad | Diagnostic, by exception | On trigger |
Cost per sample moves with laboratory, turnaround time and the exchange rate; what stays constant is the frequency needed to catch a trend before it becomes a failure. A full quarterly routine package plus the half yearly extended tests is a modest, predictable annual line for most single machine plants. Online moisture and particle sensors on the main reservoir cost more to install but supplement laboratory work rather than replacing it. See our cost guide for how programme scope drives the total.
Lube oil sampling: where, how often, and why the method decides the result
A bad sample produces a confident wrong answer, which is worse than no sample. Three rules cover most of it.
Sample from a live, turbulent line while the machine is at normal operating temperature and load. The main return header before the filters is the standard system point. Individual bearing drain lines are what let you isolate which bearing is wearing, and fitting permanent sampling valves on those lines during a planned outage is a small job with a large diagnostic payoff.
Never sample from the bottom of a reservoir, from a drain plug or from a filter housing. Bottom samples collect settled water and sludge, and the result reflects the sump rather than the circulating oil.
Use pre cleaned bottles rated for particle count work, flush the sampling port before drawing, and record hours run, load, oil temperature, ambient conditions and any top up volume since the last sample. A particle count taken into an ordinary bottle is not a particle count.
On frequency, ASTM D4378 sets out the accepted practice for in service monitoring of mineral turbine oils and is the document to write into your maintenance procedure and your tender specifications. As a starting position, quarterly routine testing with an annual extended package suits most industrial turbines. Move to monthly when a trend is running, after any oil system intervention, following a top up of more than 10 percent of the charge, or on a machine whose failure would stop the plant.
Nigerian sites have two local factors worth planning around. Harmattan season pushes silica into any reservoir with a tired or missing desiccant breather, which is part of the wider picture in harmattan dust turbine derating. And oil bought outside a controlled supply chain is not always the grade on the drum, so a baseline test on new oil before it goes into the machine is money well spent.
Worked example (hypothetical): a sample shows silicon rising from 3 ppm to 28 ppm with a particle count at 21/19/16 and filter differential pressure climbing. The cause is a reservoir breather that has been running without desiccant for a couple of seasons. A new breather assembly and a kidney loop filtration pass is a modest job that returns the system to 16/14/11 inside a week. Left alone, that silica would go through the bearings and the servo valves.
Turning turbine oil analysis results into a maintenance decision
Data with no owner changes nothing. Assign one named engineer to receive every report, compare it against the previous two samples and record a decision, even when that decision is no action.
Judge trends, not single numbers. Two consecutive samples moving in the same direction at unchanged operating conditions is a stronger signal than one result crossing a limit, and one result crossing a limit is often a sampling error until the resample confirms it.
Read the oil alongside the machine. Rising tin plus a bearing metal temperature drifting upward plus a shaft centreline shift is a bearing case with three independent confirmations, and it belongs in the next outage scope. Rising MPC plus slow valve response plus a filter differential pressure climb is a varnish case. The diagnostic sequence for the vibration side sits in steam turbine high vibration.
Some responses are site work and some are not. A breather change, a filter change, a resample or a kidney loop pass is routine maintenance. Bearing inspection, oil flushing to a measured cleanliness target, servo valve overhaul and rotor work need an outage window, calibrated tooling and a qualified turbine engineer on site. Oil results tell you which of those you are facing, and they usually tell you early enough to plan it rather than react to it.
Oil condition sits inside the same scope as clearances, alignment and inspection intervals in a steam turbine overhaul, and the same discipline applies across pumps, compressors and gearboxes under rotating equipment services. If you want a sampling plan, action limits and a laboratory scope written against your specific machines, book a plant assessment.
Frequently Asked Questions
How often should turbine lube oil be sampled?
Quarterly routine testing with an annual extended package is a reasonable baseline for most industrial steam and gas turbines. Move to monthly sampling when a parameter is trending, after any work on the oil system, following a large top up, or on a machine whose failure would halt production. Frequency should be set against downtime cost and machine criticality, not against oil volume.
Can turbine lube oil analysis tell me which bearing is wearing?
Only if you sample the individual bearing drain lines. A sample from the main return header tells you that the system is generating wear metal but not where it originates, because everything mixes in the reservoir. Fitting permanent sampling valves on each drain line during a planned outage is inexpensive and turns a general warning into a specific one. Without that, you pair the oil result with bearing metal temperatures and shaft position data to narrow it down.
What is varnish, and will a standard oil analysis package detect it?
Varnish is a soft insoluble oxidation product that plates onto metal surfaces and closes the tight clearances in governor servo valves, trip blocks and actuators. A standard package of viscosity, water, acid number and spectroscopy will usually not detect it, because varnish forms from dissolved degradation products rather than measurable particles. You need Membrane Patch Colorimetry for varnish potential and RULER or RPVOT for remaining antioxidant. Both belong on any turbine with an electro hydraulic control system.
Does good oil analysis mean I can extend my turbine inspection intervals?
Not on its own. Oil analysis reports the condition of the lubricant and the debris it carries, but it cannot measure babbitt bond, bearing clearance, blade condition or casing distortion, all of which require physical inspection. It can support a condition based case for adjusting an interval, but that case has to be built with the OEM position, vibration history and inspection findings together. The realistic benefit is fewer surprises inside your existing schedule, which is discussed further in turbine inspection intervals.