Most gearboxes die of something that was visible in the oil months before it was audible in the room. A wear metal count that doubles between two samples, a viscosity that climbs because the oil is oxidising, a water reading that creeps past a few hundred parts per million - each of those is a message, and none of them announces itself. Oil analysis is simply the habit of reading the message on a schedule instead of waiting for the failure to translate it.

This page is about building that habit as a programme rather than a one-off test: what to measure, where to take the sample, how often, and - the part most plants skip - how to convert a laboratory number into a decision somebody actually makes.

What an Oil Analysis Programme Actually Buys You

A single oil test tells you what the oil is like today. A programme tells you which direction it is moving, and direction is where the value sits. Two samples taken six months apart, read against each other, can show a bearing starting to lose its surface, a seal beginning to pass process water, or a sump running hotter than its grade was chosen for. None of those findings is visible from outside the case.

The economics are straightforward. A routine sample costs a small fraction of a service visit, and a fraction of a fraction of an unplanned stop. The reason programmes still fail is not cost - it is that nobody owns the calendar, nobody reads the report when it arrives, and the first interesting result is discovered in a drawer after the gearbox has already been opened.

So the first design decision is not technical. Decide who receives the report, what they are expected to do with it within a week, and where the trend file lives. A programme without an owner is a subscription, not a maintenance tool.

The Four Numbers That Matter in a Gearbox Sample

Laboratories offer dozens of tests. For industrial gear units, four families carry most of the information, and a fifth - the particle count - matters on critical drives where cleanliness is specified.

Wear metal ratios are more useful than absolute values. Iron rising with copper, on a unit with a bronze wheel, points at the worm mesh; iron rising alone with a stable particle count points at gear tooth contact or a bearing race; lead and tin together suggest a plain bearing surface rather than a rolling element. The laboratory's comment line is a starting point, not a verdict - the trend against your own baseline is the evidence.

Setting Sampling Points and Intervals

Sample from the same place every time, or the trend is fiction. On most industrial units the correct point is a live sample from a valve or sampling port in the return line, taken with the unit running and the oil warm, so the sample carries what is actually circulating rather than what settled in a corner. Where no port exists, a suction sample through the filler after a short run is the accepted compromise - and then it must stay the same compromise for every later sample.

Intervals should follow consequence, not convenience:

• Critical, continuous drives with no standby: sample quarterly, and at half that interval for the first 500 hours after a rebuild.
• Important drives with a spare or a bypass: every six months is a reasonable starting rhythm.
• Small units on intermittent duty with a cheap replacement cost: annual sampling, or condition-based change without analysis, is defensible.
• Any unit that has just been commissioned, repaired or had an oil change: take a baseline sample after a few weeks of running, once the oil has circulated and stabilised.

New oil is not a reference. Fresh oil carries its own additive package and its own particle count, and a plant that never records a baseline will spend years comparing samples to a catalogue figure instead of to its own machine.

Reading Trends, Not Single Reports

The single most common error in oil analysis is alarm at one number. A calcium or magnesium reading that jumps once usually means a top-up from a different product line, not a catastrophe. A particle count that spikes and returns means a filter change or a dusty week, not necessarily an ingress problem.

Trends behave differently. Three consecutive samples showing water climbing by equal steps is a seal path opening; iron rising in step with hours is normal wear, iron rising faster than hours is not. The practical rule is to act on the second confirming sample, not the first surprise, and to investigate immediately only when two different tests move in the same direction at once - water up and acid number up, or iron up and particle count up.

This is why the trend file matters more than the report. A spreadsheet with one row per sample, the four key values, hours run and any maintenance event is enough. It takes two minutes per sample and it turns a pile of PDFs into a history that a new maintenance engineer can inherit.

Turning a Report into a Maintenance Decision

A report becomes useful at the moment someone writes a sentence next to it: continue, shorten the interval, change the oil, inspect, or stop. Fixing those five responses in advance removes the ambiguity that otherwise parks reports in an inbox.

• Continue - all values inside the baseline band and no directional change. Keep the interval.
• Shorten the interval - one value drifting, cause not yet known. Re-sample at half the interval and check the obvious external causes: breather, seal area, oil level, ambient.
• Change the oil - viscosity or acid number out of band, or water above the limit for the application. Record the change in the trend file so the next sample is read against a known starting point.
• Inspect - wear metals rising with particle count, or a magnetic plug carrying paste rather than fuzz. Open the inspection cover before the noise starts.
• Stop - rapid change across two or more tests, or metal flakes on the plug. Continuing to run converts a rebuild into a replacement.

The decision is always cheaper than the delay. A gearbox opened on a trend finding costs bearings and a gasket set; the same gearbox opened after a seizure costs a shaft, a case and the production that stopped while it was being found.

Common Mistakes That Waste the Programme

Several habits quietly destroy the value of good laboratory work. Sampling from the drain plug after the oil has settled gives a reading of the sludge, not the circulating oil. Reusing a dirty sample bottle contaminates the result before it leaves the plant. Topping up with a different grade and saying nothing makes the next viscosity reading look like a fault in the machine rather than a decision by the store. Changing the oil and skipping the baseline sample leaves the next report with nothing to compare against.

Two more are structural. Running a programme on a unit that has never had its sump cleaned means the first samples are reading years of accumulated sediment. And ignoring the breather makes every water reading a mystery, because a saturated breather on a humid site transports moisture into the case every time the oil cools.

None of these mistakes requires equipment to fix. They require the sample to be taken the same way, by somebody who knows why, and the result to land on a desk where a decision is expected.

A One-Page Programme for a Small Plant

A plant with twenty gearboxes does not need a laboratory strategy; it needs a page on the wall and a calendar. The page carries the unit list, the sampling point for each, the interval, the four values that matter, and the person responsible for reading the report within a week. The calendar carries the dates.

The technical content is small. Choose one laboratory and keep it, so the methods stay comparable. Record a baseline after every oil change. Act on second confirmations. Keep the trend file with the machine history rather than in a folder labelled reports. Review the whole programme once a year against what it actually caught.

For a wider view of how these drives are specified in the first place - shaft fit, mounting, oil grade for the ambient band - the right-angle hollow shaft gearbox page walks through a worked frame selection, and the online selection tool will size a unit against your torque, ratio and duty class before the oil ever matters.

Frequently Asked Questions

How often should gearbox oil be sampled?

Quarterly for critical continuous drives, every six months for important units with a standby, annually for small intermittent units. After a rebuild or an oil change, take a baseline sample once the unit has run a few weeks, then resume the normal rhythm.

Is a single oil test worth doing?

Yes, as a snapshot of water, viscosity and wear metal content, but its value is limited without a comparison. One sample cannot show a trend; the second sample is where the information begins.

Can oil analysis replace opening the gearbox?

No. It replaces unnecessary opening. Analysis tells you when to look and what to look for; the inspection cover still has to come off to confirm a bearing or a seal that the report suspects.

What does water in gear oil actually do?

It reduces film strength, promotes corrosion on gear flanks and bearing races, and accelerates oxidation of the oil itself. Even modest water content shortens oil life and raises wear rates on the surfaces that carry the load.

Does synthetic oil change the sampling interval?

It usually allows a longer change interval, not a longer sampling interval. Synthetic grades resist oxidation better but they still collect water, dirt and wear metal, so the sampling rhythm stays and the interpretation shifts toward contamination rather than ageing.