Marine engine oil analysis: spotting wear before failure

A main engine does not fail all at once. A bearing wiping, a liner scuffing, an injector leaking: all of it leaves traces in the oil long before anything can be seen or heard. Oil analysis is the business of reading those traces.

The principle is mechanical, not magical. The lubricant circulates through the whole engine and touches every moving surface. It carries away the metal particles torn from the components, the products of combustion, water that has leaked from a cooling circuit, fuel that never burnt. A laboratory measures those elements. The chief engineer reads the report and decides.

That still requires the sample to be drawn properly, to reach the laboratory, for the result to come back, and for it to be tied to the right piece of equipment at the right running hours. This is where most shipboard oil analysis programmes come apart: rarely on the chemistry, almost always on the organisation. This article covers both.

Wear metals: reading the signature of what is wearing

Every moving part leaves a metallic signature in the oil that matches its alloy. Emission spectrometry measures those elements in parts per million. The report does not say “the bearing is about to go”: it says “lead has risen from 4 to 31 ppm”. Making the connection with the architecture of the engine is the engineer's job.

That connection depends on the machine. The same element does not have the same origin on a medium-speed four-stroke, on a reduction gearbox or on a deck hydraulic power pack. The table below sets out the correspondences most often seen on shipboard installations.

Trend beats absolute value

A single figure means almost nothing on its own. The ppm concentration depends on sump volume, on how much top-up oil has gone in since the last change, on elapsed time and on filtration. Two identical engines on two different vessels will not share the same baseline.

What matters is the gradient. Take an auxiliary engine whose iron sits between 18 and 22 ppm across four samples taken 250 hours apart. On the fifth it reads 55 ppm. Nothing about that figure is catastrophic in itself — plenty of engines run happily at 55 ppm of iron. What matters is that it has almost tripled over one interval, after a thousand hours of stability. That break in gradient is what justifies a resample and an inspection, not the crossing of a threshold.

For the gradient to be readable, three things must hold: the same sampling point, the same laboratory using the same method, and a regular interval expressed in running hours rather than weeks. A propulsion engine that has run 900 hours between samples and a harbour generator that has run 60 cannot be compared on a calendar. That is why the running hour counters have to be captured with each sample rather than reconstructed afterwards.

It is also worth thinking in wear rate rather than concentration: ppm gained per hundred running hours, corrected for the top-up oil added. A sump that is topped up regularly dilutes the metals and artificially flattens the trend. On large engines that consume oil this bias is permanent, and the only way to correct it is to record the top-ups. Tracking the volumes issued, held at tank and bunker level, therefore feeds directly into how the analyses are read.

Oil condition: water, fuel dilution, acidity and viscosity

The second half of the report is no longer about the engine but about the oil itself. An oil can be chemically exhausted with no metal having moved at all — and the reverse is equally true. The two readings are independent and should be made separately.

That last line deserves emphasis. Emission spectrometry only detects fine particles reliably. A gearbox tooth that is spalling produces debris far too large for the method to see: the report can look reassuring while the gear set is coming apart. On gearboxes and bearings, ask explicitly for a ferrous index or ferrography, and keep inspecting magnetic plugs.

Taking the sample: where most analyses are lost

A laboratory measures what it is sent. If the sample is not representative of the oil in service, the report is wrong — and a wrong report is worse than no analysis at all, because it reassures. Sampling quality has more influence on the result than the choice of laboratory.

Where to draw the sample

The right point is a dedicated sampling valve on a live pressurised line, downstream of the pump and upstream of the filter. Before the filter you see the particles the system is genuinely carrying; after it you see only what the filter let through. Fit that valve once and make it the single sampling point for that machine.

Where no dedicated valve exists, sample through the dipstick tube using a vacuum pump and fresh tubing, cut to a marked length so that it always reaches the same depth, roughly mid-level in the oil. Never from the bottom of the sump: the bottom concentrates water, sludge and heavy debris, and produces results that are consistently blackened. Never from the drain plug during an oil change either: what comes out then is the flushing of the sump, not the oil in service.

Under what conditions

Engine hot, at normal operating temperature, and running under steady load. Cold oil in an engine that has been stopped for hours has had time to settle out precisely what you are trying to measure. If the machine cannot be sampled running, sample immediately after shutdown, before settling begins, and note it on the label.

Before filling the bottle, flush the valve: run off enough oil to clear the stagnant contents of the valve and the dead leg, and put it in the waste container. Without that flush you are analysing the inside of a valve, not the contents of the system.

Use the bottle supplied by the laboratory, opened at the last moment, filled to about three quarters and closed again straight away. A bottle filled to the brim cannot be homogenised before testing; a reused bottle, however well rinsed, brings its own metals with it. Sample before any top-up and before any additive dosing, never after.

What the label must carry

A sample without a full identity is a wasted sample. The label, or the form that goes with it, must let someone who was never on board reconstruct the context: vessel, equipment tag exactly as it appears in the register, oil brand and grade, total machine hours, hours on the oil since the last change, top-up volume since that change, fuel in use at the time of sampling, date, and the name of the person who drew it.

The equipment tag matters most. “Port generator” is not enough on a vessel with three of them, one of which was replaced last year. What belongs on the bottle is the identifier from the equipment record — the same record that carries the job history.

Laboratory turnaround seen from a vessel

Ashore, you sample in the morning and have the report the next day. On board, you do not. The sample leaves at the next port call, travels by courier, reaches the laboratory, waits its turn, and the report comes back days or weeks after it was drawn. It often arrives once the vessel has sailed again, the engine team has changed over, and the machine has run several hundred more hours. That lag is not a failure of organisation: it is a fixed condition of the problem, and the follow-up has to be built around it.

The first consequence is a simple rule: the result belongs to the equipment and the hour counter at the moment of sampling, never to the date the report arrived. A report filed on its arrival date manufactures a false chronology. Three samples drawn at 5,200, 5,460 and 5,710 hours and received out of order three weeks apart only tell the right story if each is re-anchored to its original counter reading. That reconciliation is what makes the trend curve usable, and it is only possible if the counter was written on the bottle.

The second is that a register of outstanding samples has to be kept. A bottle that leaves without a record never comes back: nobody knows a report is missing, and on board the absence of a result reads as a normal result. The register carries, for each sample, the sampling date, the equipment, the counter reading, the despatch date and port, and then the date the report returned. Anything past the usual turnaround gets chased. That register sits naturally in the digital engine room logbook, alongside the readings that already document how the auxiliaries are running.

The third is deciding what to do while waiting. Three cases cover it. If the sample was routine, carry on and schedule the reading for when the report lands. If it was triggered by a doubt — a noise, a bearing temperature, oil consumption creeping up — do not wait for it: inspect with what is on board, and let the report confirm or rule out. If the doubt concerns water or glycol ingress, there are quick shipboard checks that give a yes-or-no answer in minutes; they do not replace the laboratory, but they stop you sailing three weeks on an assumption.

One last habit, and a cheap one: draw a duplicate and keep the second bottle on board, labelled, until the report comes back. When a result looks aberrant, you can then send the retained sample instead of restarting a measurement campaign that has, by then, lost all connection with the condition you observed.

Heavy fuel, two-stroke engines and how to read TBN

TBN is the parameter most often misread on board, because it has no absolute value: it is always read against the sulphur content of the fuel being burnt. The alkaline reserve of the oil exists to neutralise the sulphuric acid formed during combustion. The more sulphur in the fuel, the faster that reserve is consumed, and the higher the starting BN needs to be.

This is exactly what MARPOL Annex VI changed in lubrication practice. Since 1 January 2020 the maximum sulphur content of marine fuels has fallen from 3.50% to 0.50% by mass worldwide, with a 0.10% cap inside emission control areas. Vessels fitted with an exhaust gas cleaning system may continue to burn high sulphur fuel oil. A single fleet can therefore, today, operate side by side engines whose lubrication needs are nothing alike.

The risk runs in both directions. Too low a BN with a high sulphur fuel lets acid attack the liners: that is corrosive wear, and it accelerates. Too high a BN with a low sulphur fuel leaves the alkaline additive with no acid to neutralise; the excess deposits out, notably on the piston top land, and those hard deposits first polish and then scuff the liner. Two independent sources agree on this trade-off: the service letters issued by two-stroke engine builders and the technical literature published by the marine lubricant suppliers describe the same balance.

As an order of magnitude, two-stroke engine builders point towards low BN cylinder oils for fuels at 0.10% sulphur and below, an intermediate BN across the 0.10 to 0.50% band, and markedly higher BN grades for high sulphur fuel oil. The exact ranges vary by engine model and are set alongside the feed rate. The practical rule is therefore this: your engine builder's service letter overrides any general guidance, and it belongs filed against the equipment record, not in a separate binder.

On a crosshead two-stroke, incidentally, the analysis does not concern the same oil depending on what you are looking for. Cylinder oil is judged on the scrape-down oil collected under the pistons: residual BN and, above all, total iron, which reflects liner wear directly and lets the feed rate be tuned. System oil, separated by the stuffing box, is judged like any crankcase oil: water, viscosity, bearing metals. The two tests answer two different questions and cannot stand in for one another.

On medium-speed four-stroke engines, which power most service, fishing and passenger vessels, the reading is more direct: one oil, a TBN that falls, and an oil change called when the alkaline reserve drops below the maker's limit or when another parameter forces it. It is still worth watching the aluminium and silicon pair, which betrays catalytic fines carried in with the fuel: when both rise together and in similar proportions, it is the purifier that needs attention, not the engine.

Hydraulic oil, gearboxes and deck machinery: the blind spot

Analysis programmes usually stop at the engines. That is a pity, because the vessel's hydraulic systems respond better than anything to this kind of monitoring, and their failures are often the most expensive in operational terms: winches and windlasses, bow thruster, stabilisers, steering gear, deck cranes, fishing gear, controllable pitch propeller hub.

The difference lies in the dominant failure mode. On an engine you are looking for wear. On a hydraulic system the root cause is nearly always particulate contamination: the internal clearances of a piston pump or a servo valve are measured in micrometres, and it is particles of that size that destroy them. The primary parameter is therefore not spectrometry but particle counting, expressed under ISO 4406 as a three-number code corresponding to the populations of particles above 4, 6 and 14 micrometres. The maker of the most sensitive component in the circuit sets the code you must not exceed; that target comes from the component, not from a generic figure.

Water comes next. A deck hydraulic system breathes through its reservoir, in salt air and across wide temperature swings. The water that condenses there corrodes, encourages varnish on spool valves and destroys the additive package. On exposed deck machinery, water content deserves the same attention as cleanliness.

Gearboxes follow a third logic again. Gear teeth do not wear diffusely: they spall. The debris produced is too large for spectrometry, which makes a ferrous index or ferrography indispensable alongside magnetic plug and filter inspection. A gear set can be degrading badly while the spectrometric report stays perfectly normal.

One further circuit deserves a place in the plan: stern tube oil. There the priority is water content and sodium, which flag a loss of sealing at the aft seal — information far better obtained from a routine analysis than from an unexplained rise in tank level. Each of these circuits should appear as a distinct item in the equipment register, with its own analysis frequency and its own trend curve.

Oil analysis, vibration and thermography: three different views

Oil analysis is only one of the three condition monitoring techniques in common shipboard use. They overlap partially, and it is that overlap which lets a diagnosis be confirmed. None of them replaces the other two.

In practice the useful combination is straightforward. Oil analysis says something is happening inside; vibration says where, on which rotating element; thermography confirms whether it is running hot. A line shaft bearing whose lead is climbing in the oil, whose vibration spectrum is filling in, and which shows warmer than its neighbour on a thermal image: that is no longer a hypothesis. The method and the measurement points are set out in the article on vibration analysis and thermography on board.

The three techniques also need to feed the same equipment file. Three separate spreadsheets kept by three different people never cross-check at the right moment. That is one of the arguments for a single condition monitoring record, whether the input is a laboratory result, a manual reading or data pushed automatically by the vessel's sensors — a subject covered in detail in the article on using NMEA 2000 data for condition-based maintenance.

From laboratory result to work order

An oil analysis programme that never generates a work order is not a maintenance tool: it is a cost line. The question is not whether reports arrive, but what happens to them in the two days that follow.

Three outcomes cover almost every case.

• Normal: nothing to do, but the result is filed against the equipment record with its counter reading. A normal point has value: it builds the baseline the next one will be measured against.
• Monitor: a parameter is drifting without crossing the critical limit. Shorten the sampling interval and attach a light inspection to the next routine visit — magnetic plug, filter cut open and examined, temperature check.
• Critical: limit crossed or a clear break in gradient. Raise a dated work order, assigned, with the report attached.

That condition-based work order has to carry more than an instruction. It carries the equipment tag, the counter reading at sampling, the measured values and the limit crossed, the last three results, and the resample to be taken after the job. An engineer who opens it three weeks later in another port must be able to understand why it was raised without tracking down whoever raised it. That is precisely what a work order held in the CMMS does and an email does not.

Limits themselves are built in two stages. You start with the maker's figures, which are the only ones available at the outset and remain the contractual reference. After a first year of monitoring you have each machine's own history, and you almost always find that some units live normally above or below the generic values. At that point you tighten the limits onto what the machine actually does. Moving from the theoretical limit to the observed limit is what separates a programme that is endured from one that is useful — the same shift described in the article on moving from corrective to condition-based maintenance.

Then there is the fit with the existing maintenance plan. Oil analysis does not abolish preventive work: it moves part of it. Routine oil changes can be extended where the analysis supports it and the maker permits it, but inspections, safety rounds and statutory checks stay on the calendar. Good practice is to enter the sampling itself as a recurring preventive task, triggered by the hour counter, exactly like a survey. Without that, sampling is the first thing dropped when the port call is short. Building that plan is covered step by step in the four-step preventive maintenance plan guide.

Setting up the programme on board

A useful oil analysis programme comes down to a handful of decisions, taken once and held to.

1. Choose the equipment to monitor: main and auxiliary engines, gearboxes, hydraulic power packs on critical deck machinery, stern tube. Start with whatever stops the vessel when it fails.
2. Fit a dedicated sampling valve on each, in the right place, and never move it.
3. Set the interval in running hours and enter it in the preventive maintenance plan as a task in its own right.
4. Define what the label must carry, and refuse any incomplete sample — including your own.
5. Keep the register of samples despatched and chase whatever fails to come back.
6. Tie every result to the equipment and to the counter reading on the day of sampling, not to the date of receipt.
7. Decide the three possible outcomes in advance, and who raises the work order.

None of this calls for special equipment on board. It calls for consistency: the same point, the same interval, the same label, the same laboratory. That consistency is what turns a stack of isolated reports into a curve you can make decisions from.