Water in Oil: Read the Report Before You Blame the Cooler

The 24 March report

Read the units and the test method first

On a Karl Fischer report, mg/kg is the same as parts per million by mass. So 540 mg/kg is 0.054 percent, and 1 percent is 10,000 ppm. A result in volume percent is on a different basis, because oil is lighter than water. Before comparing two reports, check that the units and the basis match.

Laboratories typically measure water in oil by Karl Fischer titration, in which a reagent reacts with the water in a weighed sample. The coulometric version used for oils is standardised as ASTM D6304. Some sulfur-containing additives can also react with the reagent and count as water, as NIST researchers noted in 2004. Metrohm, a titrator maker, advises heating such oils in an oven and sending just the vapour to the titration cell, which lowers the interference without always removing it.

Keep the laboratory and method the same for a trend, and if a result jumps, ask whether the oil was injected directly or heated in an oven. Here the laboratory confirmed the oven method used for every earlier report, which rules out a change of test.

About a teacup of water, spread through 400 litres of oil. No level gauge will register that, so the laboratory result may be the only sign.

Referenced here ISO 12937 preview; Mobil turbine oil guide; NIST oven study; Metrohm AB-209; Chevron water bulletin

Clear oil can still hold water

Water in oil can be dissolved, emulsified as fine droplets that make the oil hazy, or free and settled at the bottom. How much an oil can dissolve, its saturation point, depends on the base oil, additives, age and temperature. In Pall’s example of a typical hydraulic fluid, the same 200 ppm is about 80 percent of saturation at a 15 °C start-up and 40 percent at 50 °C, while the Karl Fischer result does not move.

So oil that is clear at running temperature can turn hazy on a laboratory bench. The US Bureau of Reclamation’s lubrication manual says to let a sample stand at room temperature for at least 24 hours before judging its appearance, and treats haze only as a possible sign of water. Haze from water means the oil held more than it could dissolve at that temperature. It does not say how much.

Clear oil is not proof of harmless oil either. In full-scale tests published in 1977, R. E. Cantley found tapered roller bearing fatigue life fell as water in an SAE 20 oil rose from 25 to 100 to 400 ppm. That is one oil and one bearing type, not a limit for this unit.

Referenced here Pall: water content as percent of saturation; Reclamation FIST 2-4; Cantley 1977, abstract

The 7 April resample

Check what the bottle represents

Free water and sediment settle. Reclamation, writing about bearing oil tubs in hydro plants, warns that a drain-line sample from the bottom of the tub will likely give “a much worse picture” of the oil’s condition. Mobil’s turbine oil guide lists bottom samples among the causes of a high water result.

Keep these the same from one sample to the next, and write any change on the label:

• The same valve each time, on flowing oil, with the stagnant oil in the line flushed out first
• The machine at its normal operating state, with the time since start-up recorded
• A clean, dry, sealed bottle, kept closed until it reaches the laboratory

Our guide to where you take an oil sample explains how to choose the point. Before a result changes the maintenance plan, confirm it with a second sample from the routine point.

This unit’s resample from the same valve on 7 April read 520 mg/kg. The rise is real.

Referenced here Reclamation FIST 2-4, section 5.1.1.1; Flender 7300en, section 7.3.1; Mobil turbine oil guide, appendix 1

Published limits

Judge the number against this machine

Published water limits differ several-fold between machines and oils:

Use the limit set for this machine by its maker, the oil supplier or your monitoring programme, and read the result against the machine’s own history. Mobil’s turbine oil guide gives similar advice about its own interpretation guidelines: treat them as a general guide, and take corrective action only with guidance from the equipment maker or the lubricant supplier.

The plant in this example follows the Rexroth figures for its hydraulic units. At 540 mg/kg the oil is above the 500 recommended for long life but under the 1,000 ceiling: worth acting on, not an emergency. The bottles turned faintly hazy on the bench, but no free water settled in them or showed in the sight glass. So once the resample confirmed the rise, the oil was dried in mid-April by a method the supplier accepted, and the search went on.

Referenced here Flender 7300en, table 7-2; Bosch Rexroth RE 90220; Mobil guide, appendices 1 and 2

January to June

Look at when the water arrived

Drying lowered the number. It did not explain it. In May, a second stop brought a second rise. By June, the history looked like this.

The cooler

Test the cooler theory with evidence from both sides

Some oil coolers are designed to keep the oil at a higher pressure than the cooling water, so that a leak sends oil out rather than letting water in. A 1994 CANTEACH training module for CANDU nuclear stations describes this arrangement and warns that water can still leak in during a shutdown, when the oil is no longer pressurised.

Reliability Engine’s explainer on heat exchanger leaks in data-centre cooling loops shows the same rule with a pressure slider: shift the balance across a breach and the leak reverses. Its fluids and limits differ from a hydraulic unit, but its advice carries over: look for evidence on the side that receives the leak.

Condensation fits the same timing: a reservoir that cools during a stop in humid weather draws moist air in. Bosch Rexroth lists condensation as a way water enters hydraulic fluid, and recommends an air dehumidifier at the reservoir vent in humid conditions.

Each remaining source leaves a different trace:

If the cooling water is treated, ask the laboratory whether sodium or boron is rising in the oil, and send a cooling-water sample for comparison. Oil additives can contain both elements, and sodium can also come from hard water, so compare with an analysis of the new oil first. With only about 0.14 L of water per stop, the elements may not rise enough to measure, so a flat result does not clear the cooler.

When the evidence points to the cooler, have authorised people test it by the maker’s procedure. Alfa Laval’s gasketed plate heat exchanger manual, for example, pressure-tests one side at a time with the other side open to ambient pressure.

Referenced here CANTEACH turbine lubricating oil module; Bosch Rexroth RE 90220; Mobil guide, appendix 1; Alfa Laval gasketed PHE manual

Write the decision down

A short review record keeps the reasoning visible to the next person and stops the same report being argued again next month.

Back to the March meeting. The oil did not need changing that day, and the cooler had not yet earned a pressure test. By June the timing had narrowed the suspects to the cooler and condensation. If the water side reads higher at the next stop and the neighbouring units stayed flat, test the cooler. A water case closes when the record shows where the water came from.