A mobile excavator hydraulic system that kept eating pumps. The pumps were new. The oil was changed twice a year. The oil was “clean” — it looked clear in the sight glass. The pumps failed at 1500 hours instead of the expected 8000. The lab report said the oil was ISO 4406 22/19/16. Nobody on the maintenance team knew what that meant. The pump manufacturer’s spec said 18/16/13 maximum. The oil was four classes dirtier than spec. The pumps were being fed with sandpaper.
What the ISO 4406 code actually says
The ISO 4406 code has three numbers, separated by slashes. They count particles per milliliter of oil in three size ranges: greater than 4 μm, greater than 6 μm, and greater than 14 μm. The numbers are not particle counts — they’re ranges on a logarithmic scale. A code of 22/19/16 means about 40,000 particles per milliliter larger than 4 μm, 10,000 larger than 6 μm, and 800 larger than 14 μm.
One milliliter is a drop. Forty thousand particles in a drop. The sight glass looked clear because the particles are smaller than the eye can see. A human eye resolves down to about 40 μm. The killer particles in a hydraulic system are in the 5-15 μm range — exactly the clearance between a pump piston and its bore. The oil that looks clean is full of the particles that destroy pumps.
The pump clearance math
A piston pump has a clearance between the piston and the cylinder bore of about 5-10 μm. Particles larger than the clearance get trapped at the inlet of the gap. They score the surface as the piston moves. The scoring enlarges the clearance, which increases internal leakage, which reduces volumetric efficiency, which raises the oil temperature, which thins the oil, which accelerates the wear. The pump dies from the inside out.
The 22/19/16 oil had 800 particles per milliliter in the 14 μm range. The pump spec wanted 18/16/13 — about 50 particles per milliliter above 14 μm. That’s a 16x reduction. The filter that was installed was a 10 μm nominal filter. It was catching particles above 10 μm, but the 5-10 μm particles — the ones that fit exactly into the pump clearance — passed straight through.
Why the filter wasn’t doing its job
The filter was a 10 μm nominal (β10=75) return-line filter. Three issues. First, nominal rating means 50% of particles at that size are captured. A β10=75 filter removes about 98.7% of particles larger than 10 μm — but only in the lab test, with a fresh filter and clean oil. In service, the filter efficiency drops as it loads. Second, the filter was on the return line, not the pressure line. The pump suction was drawing from a reservoir that had years of settled particles. Third, the filter was undersized — the flow through it was 2x its rated capacity, so the bypass valve opened most of the time and unfiltered oil flowed around the element.
The fix for this machine: a 6 μm absolute filter (β6=1000) on the return line, sized so it never bypasses, plus a 10 μm filter on the pressure line, plus a magnetic plug in the reservoir to catch ferrous wear particles. The oil was also flushed with a portable filtration cart (3 μm absolute) for 24 hours before the new filters were installed. The next oil analysis came back 17/15/12 — within spec. The pumps now last 6000+ hours.
The sampling discipline that makes the code mean something
An ISO 4406 code is only as good as the sample. Taking oil from the reservoir drain plug or the bottom of a bucket gives a sample that’s dirtier than the operating oil. The correct method: take the sample from a sampling valve on the pressure or return line, with the system running at operating temperature, after at least 30 minutes of operation. The sample bottle must be clean (lab-supplied). The sample must be taken before the filter, not after it — you want to know what’s going into the components, not what the filter is doing.
The plant that fixed the excavator samples monthly. They track the trend, not just the number. If the code climbs one class in a month, something is wrong — a leaking cylinder rod seal, a dusty breather, or a filter reaching the end of its life. The trend catches the problem before the pump does.
The target codes by component
| Component | Maximum ISO 4406 code |
|---|---|
| Piston pumps and motors | 18/16/13 |
| Vane pumps and motors | 19/17/14 |
| Gear pumps | 20/18/15 |
| Servo valves and proportional valves | 16/14/11 |
| Directional valves (standard) | 20/18/15 |
| Cylinders | 20/18/15 |
These are the numbers from the component manufacturers’ literature. If your oil analysis comes back dirtier than the component spec, you have two choices: filter the oil better, or replace components more often. The second choice is what most plants do without realizing it’s a choice.
The cost of clean oil
Clean oil costs money: a β6=1000 return filter costs about $80. A portable filtration cart rental is about $500 for a weekend. An oil analysis is $40 per sample. The alternative: a new piston pump is $3,500 plus the labor to change it. A pump failure at 1500 hours on a machine that runs 3000 hours a year means two pump replacements a year. The filtration program costs a fraction of one pump.
Clear oil is not clean oil. The ISO 4406 code is the only honest measure, and it has to be measured on a proper sample, not guessed from a sight glass. If your pumps fail before their rated life, measure the oil before you blame the pump. The excavator pumps weren’t defective — they were being fed particles that the filter was supposed to catch.
