What Oil Carryover Does Downstream

Why this matters

Nothing bad happens to a lubricated compressor when it passes oil downstream. The machine runs fine, makes its pressure and shows no fault. The bill arrives somewhere else entirely: in a desiccant bed that dies in a fraction of its rated life, in coalescing elements that reach their change point in a third of their interval, in a coating that fisheyes, in valves that get sticky, and in a condensate stream that has become a regulated disposal problem. Every one of those gets diagnosed locally, as its own fault, by a different person, and none of them points back at the compressor room.

The sump is pressurised, and the oil is hot

Before anything else on a lubricated rotary screw: the oil sump on that machine is a pressure vessel at system pressure while the unit is running and for as long as it stays pressurised after it stops. Do not crack a fill plug, a sight glass, a drain or a sample port until the machine is shut down, its air isolation is closed, the sump is vented through the machine's own blowdown, the sump gauge reads zero with the vent still open, and the isolation is locked and tagged under 29 CFR 1910.147, which is the standard covering stored mechanical and pressure energy. Compressor oil at operating temperature will cause a serious burn, so allow the cool-down time in the machine's manual before draining rather than working around it.

Where a panel has to come off to reach a temperature sensor or a starter, that is electrical work: de-energize at the disconnect, lock and tag under 29 CFR 1910.333(b)(2), prove dead live-dead-live per NFPA 70E-2021, 120.5, in the edition your employer's electrical safety program has adopted, with 29 CFR 1926.417 as the construction counterpart.

Read the lubricant's safety data sheet before handling it, which is your employer's obligation to make available under 29 CFR 1910.1200. Wear chemical-resistant gloves and sealed eye protection for any oil handling. Where a leaking fitting is producing a visible oil mist rather than a drip, the route of exposure is inhalation and it needs the leak stopped and the area ventilated, not a glove.

Carryover is two different substances, and only one of them filters out

Aerosol is oil suspended as fine droplets. A coalescing filter, correctly graded and correctly protected by a separator upstream, removes the large majority of it, and its performance is quoted as a remaining carryover concentration at a stated inlet concentration, flow and temperature.

Vapour is oil in the gas phase. It passes through every mechanical filter that has ever been built, because there is no droplet to intercept. The only thing that removes it is an adsorbent bed, usually activated carbon.

The split between the two is not fixed, and this is the relationship that explains most surprising carryover results: the vapour fraction rises with air temperature at the compressor discharge. A machine running at the top of its normal band, or above it, puts more oil vapour into the header than the same machine running cool, with the coalescer performing perfectly in both cases. That is why a hot compressor room shows up as a contamination problem and not only as a capacity problem, and it is why "we fitted a better filter and it did not help" is a common and entirely explicable outcome.

Where a purity class is specified, ISO 8573-1:2010 states oil as total oil, aerosol plus liquid plus vapour, in one class number, and that standard binds through a customer specification, a contract or an equipment listing in the edition named there, never on its own authority. Measuring it takes an adsorbent sampling tube drawn at a stated sample volume or a colorimetric indicator tube, not a rag held under a fitting.

The signal

A shop running a lubricated rotary screw, a heatless desiccant dryer and a coating line called about two things they did not connect. The desiccant bed had been replaced twice, at 14 months and then at 11 months, against a manufacturer figure of 3 to 5 years under stated conditions. And the coating line had begun rejecting parts for fisheyes, at a rate the shop described as roughly one part in twenty on bad days.

The dryer had never alarmed. The compressor had never tripped.

Eliminated one at a time

A bad batch of desiccant. This was the shop's own theory after the first failure, and buying the second bed from a different supplier was their test. It failed the same way in a shorter time, which kills the theory. The spent material also came out darkened and greasy rather than merely dusty, which is consistent with oil fouling and is not consistent with a bed that has simply reached the end of its cycling life.

Seasonal inlet moisture. The two bed failures spanned a humid summer and a dry winter, and the separator ahead of the dryer was discharging normally with the inlet condition within the range the dryer's manual expects. Water loading did not change between the two failures. Something that is constant across a variable cannot be the variable's effect.

A dryer control or valve fault. Cycle timing checked against the manual and matched. Tower pressures swung correctly on switchover. The purge muffler discharged on cycle. A dryer that is sequencing correctly and still losing bed capacity is being poisoned, not mis-operated.

That left contamination arriving at the bed, and the question became what was between the compressor and the bed.

What was found

Two things, in the same twenty minutes, once someone read part numbers instead of looking at housings.

The coalescing housing ahead of the dryer contained a general-purpose particulate element. The housing label said coalescer, the element inside it did not. At some earlier service the wrong element had gone into the right housing, and every oil droplet arriving since had gone straight through, because a particulate element removes solids and does not remove oil aerosol.

The activated carbon bed downstream had never been changed in the life of the installation. A carbon bed has a finite adsorption capacity, gives no differential pressure signature as it saturates, and once saturated it releases what it was holding rather than politely stopping. So the one stage that could have caught the vapour had stopped catching anything years before, silently, and the one stage that should have caught the aerosol had been unable to since the wrong part went in.

The compressor discharge temperature was measured at 207 F against the manual's normal operating band of 180 to 200 F, with high-temperature shutdown at 220 F. That is not a fault, it never tripped anything, and it was the amplifier: running above the band raised the vapour fraction on top of an aerosol path that was already wide open.

Why the bed died and why the finish did

The desiccant died because oil coats the adsorbent and blocks access to the internal surface that does the adsorbing. That capacity does not come back on regeneration the way water does, because purge air strips water and does not strip oil. So the bed loses capacity monotonically, and the pressure dew point at the outlet climbs as it does. Bed life measured 14 months and then 11 months, and the second number being shorter than the first is exactly what an increasing contamination load predicts: the same poisoning arriving faster.

The finish failed because oil vapour that condenses on a cool surface downstream leaves a film, and a coating applied over a film crawls away from it. Fisheyes are one of the few contamination symptoms visible to the naked eye, which is why the coating line noticed months before the compressor room did.

Two further consequences were present and had not been recognised. The coalescer's replacement element, once the correct part was fitted, reached its change point in about 4 months against its 12-month calendar interval, using the gate covered in the filter change-out article; on a system carrying this much oil an element earns its change on drop long before the calendar leg fires. And the condensate leaving the separator and the drains was carrying enough oil that its disposal is a pretreatment question for the local sewer authority, not a floor-drain decision. That obligation grows with carryover, and it does not appear on any gauge.

One thing was deliberately not concluded. The shop asked whether the oil had damaged the elastomer seals in the downstream cylinders. Seal compatibility depends on which lubricant class is in the machine and which elastomer compound is in the seal, and those interact specifically rather than generally, so that question went to the lubricant's data sheet and the seal manufacturer's compatibility data rather than being answered from the truck.

What confirmed it

Total oil measured downstream of the filter train with an adsorbent sampling tube at a stated sample volume, before any correction and again four weeks after fitting the correct coalescing element, replacing the carbon bed and restoring the compressor's cooling so discharge came back inside the manual's band. The post-correction figure came in at a small fraction of the pre-correction one, and the shop was given both numbers with their sample volumes and dates rather than a percentage improvement, because a percentage hides the fact that the pre-correction sample is the only evidence the correction was needed.

Bed life is the slower confirmation and it is the one that actually settles the question. The next desiccant charge has to survive well past 14 months before anybody can say the poisoning stopped, and that verdict is a year away. The report said so.

What would have changed the conclusion

If the coalescing element had been the correct part and the carbon bed within its interval, the finding moves to the compressor itself: a failed or bypassed air-oil separator element in the machine, which puts far more oil into the discharge than a filter train is designed to catch, and which is diagnosed at the machine rather than in the pipe. And if the desiccant had come out clean and dusty rather than darkened, oil was never the mechanism, the bed was simply worn out, and the whole trail above is the wrong trail.

References

• 29 CFR 1910.147 for isolating and verifying depressurisation of a compressor sump or a filter housing before it is opened, and 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, in the edition your employer's electrical safety program has adopted, plus 29 CFR 1926.417 as the construction counterpart, for electrical isolation
• 29 CFR 1910.1200 for access to the lubricant's safety data sheet before handling
• ISO 8573-1:2010 total oil classes, which bind through a customer specification, a contract or an equipment listing in the edition named there; compressor and lubricant manufacturer documentation for the normal discharge temperature band and for seal compatibility
• Local sewer authority pretreatment rules for oil-bearing condensate disposal
• See related: What Each Stage of Filtration Removes; Why a Filter That Is Not Changed Costs More Than One That Is