Why Contamination Is the Dominant Failure Cause in Hydraulics

Why this matters

The four faults a tech reaches for first on a misbehaving hydraulic machine are a worn pump, a valve that sticks, a seal that leaks and a setting that has drifted. All four are real, and on a machine that has been in service a while, all four are usually downstream of the same cause. Contamination is the dominant failure mechanism in fluid power by a wide margin, and the particles doing the damage are in the same size range as the working clearances they destroy, which puts every one of them below what a human eye can resolve. You will never see the problem on a dipstick, in a drain pan, or on a filter element you cut open.

That is what makes this a measurement discipline rather than a housekeeping one, and the measurement has a code, an arithmetic and a sampling method that all have to be right together.

The clearances set the particle size that matters

The moving parts inside a hydraulic pump or a proportional valve are separated by a film of oil, and the film is thin. Slipper, barrel and valve-plate films inside a piston pump, tooth-tip and side-plate clearances in a gear pump, and the spool-to-bore clearance in a servo or proportional valve all commonly sit in the single-digit micrometre range at operating conditions. The machine builder's own documentation owns the numbers for any specific unit, and they are worth asking for.

A human eye under good light resolves down to something like 40 micrometres. So the entire population of particles that matters is invisible, and the particles you can see are the minority. That is the single most counter-intuitive fact in this subject: the dirt you can see is not the dirt that is killing the machine. Large particles cause the dramatic, sudden failure, a spool jammed or a port blocked. Small ones cause the slow, expensive one.

Two feedback loops, and both of them accelerate

The particle loop. A particle roughly the size of a clearance gets dragged into it, abrades both surfaces, and comes out the other side along with new particles ground off the surfaces it just cut. Those new particles are, by the geometry of where they were made, also roughly the size of that clearance. The clearance is now slightly larger, so it admits more particles. Contamination is therefore self-multiplying, which is why a machine's condition degrades slowly for a long time and then falls over.

The heat and viscosity loop. A widened clearance leaks more internally, which lowers volumetric efficiency, which means the same work takes more pump flow and drops more pressure across the gaps. That loss becomes heat in the oil, which the card on what a hydraulic system is trading away works through. Hotter oil is thinner, thinner oil gives a thinner film, and a thinner film puts the surfaces closer to contact with the same particles between them.

The two loops feed each other, which is also why the component that finally fails is rarely the one that was contaminated first. A pump usually fails last, having been fed particles generated by everything upstream of it for a year.

Those two loops are what put all four of the usual suspects downstream of one cause. The worn pump is the particle loop finishing its work on the tightest clearance in the circuit. The sticking valve is silt packed into a spool clearance, which is why it sticks intermittently and frees itself when the machine is bumped. The leaking seal is a rod or shaft surface that has been scored by particles carried under the lip, so a new seal on an unpolished surface leaks again inside a season. And the drifted setting usually is not drifted at all: a relief or compensator spool with contamination in its pilot orifice regulates at a different pressure with the dial untouched, which is why the setting comes back after the valve is cleaned. Four complaints, one measurement.

Reading a cleanliness code

Fluid cleanliness is reported as an ISO 4406 code: three numbers, counting particles per millilitre larger than 4, 6 and 14 micrometres. Pin the edition and the adoption path. The 1999 revision changed both the size channels and the calibration reference for the particle counters that produce the numbers, so a code from an older report is not directly comparable to a modern one; and the standard binds you through the machine builder's specification or through your service contract, not through any regulation. Ask which edition the target was written in.

The property that makes the code usable in the field: each code number is a doubling of particle count. So the gap between two codes is a factor of two raised to the difference, and you can do the whole comparison in your head. Three code numbers is a factor of 8. Four is a factor of 16.

What a filter rating actually promises

A filter's beta ratio is the ratio of particles larger than a stated size upstream of the element to particles larger than that size downstream. It is meaningless without the size attached.

Efficiency is (beta minus 1) divided by beta:

Read the right-hand column, not the middle one. Going from beta 20 to beta 200 looks like a trivial move from 95 to 99.5 percent efficiency. It is a factor of ten in what actually reaches the pump, and for a given ingression rate and filter flow the steady-state particle count in the machine is proportional to the fraction that passes. A factor of ten is just over three code numbers, since two to the power 3.32 is ten.

Where the dirt comes from

Filtration is a rate control on a loop that has an input. Cutting the input is usually cheaper and more effective than adding filtration, and there are three inputs worth naming:

• The reservoir breather. Oil level rises and falls as cylinders extend and retract, and the tank breathes room air in and out to match. A bare vent, or a coarse breather, passes exactly the size range that does the damage. A fine breather element is the highest-yield single change on most machines and it costs a fraction of a filter housing.
• Cylinder rod seals and wipers. Every retraction draws the exposed rod back into the gland and whatever is on the rod goes with it. A hardened wiper is an open door.
• Service work. Every hose change, every filter change with the housing left open, every top-up from an open container. Which leads to the input most shops do not believe.

New oil is not clean oil. Fluid delivered in a drum or a pail is routinely dirtier than a machine's target, because it has been transferred, stored and shipped since it was made. Adding it straight to a reservoir through an open fill port is one of the largest single ingression events a machine experiences. It goes in through a transfer cart with a filter on it, or a dedicated filtered fill point. Where a supplier will give you a certificate with a code on it, compare that code against the machine's target rather than assuming it passes.

Worked example: the sample that came back 21/19/16

A machine that has eaten two pumps in three years. The builder's specification calls for a fluid cleanliness of 17/15/12 for this pump and valve package.

The first sample was thrown away. It was drawn with a dipper from the reservoir through the fill port and came back 23/21/18. A static reservoir stratifies, the bottom collects the heavy particles and the top does not represent what the pump is drinking, so that number describes the tank, not the fluid in circulation. The rule stated above is the one that governs: a valid sample comes from a live line under flow, at operating temperature, after the machine has run at least half an hour, from a dedicated sampling valve, with several times the sample line's own volume flushed through and discarded first. During that warm-up run the machine is cycling normally with its guards in place and nobody in the motion path.

The valid sample: 21/19/16.

The gap, computed on the code's own basis. Every channel is 4 code numbers above target. Two to the fourth power is 16, so the machine is carrying roughly sixteen times the particle count it is specified for, at every size that was counted. Not sixteen percent, sixteen times.

What the existing filter can contribute. The return filter fitted is beta 20 at 10 micrometres, so one particle in twenty above that size passes. Moving to a beta 200 at 5 micrometres element, where the housing and the pressure drop allow it, takes the pass-through from 1 in 20 to 1 in 200, a factor of ten, which is just over three code numbers.

The correction does not close the gap, and saying so is the finding. Four code numbers is a factor of 16. The filter change delivers a factor of 10. The remaining factor of 1.6, which is about 0.7 of a code number, has to come off the ingression side, because a filter can only remove what is already in the oil and it competes with whatever is being added. On this machine the breather cap was found finger-tight over a missing element, and the fill port had been used with an open pail for at least two years.

Why the pumps failed and the valves did not. The valves here are simple directional spools with generous clearances. The pump is a piston unit whose slipper and valve-plate films are the tightest clearances in the circuit, so it is the component filtering the fluid with its own surfaces. The general rule is worth carrying: the tightest clearance in a circuit fails first and is the least likely to be the cause. A pressure-compensated pump raises the stakes further, its compensator spool running tighter still, which the card on what a compensated pump changes covers.

The failure mode of doing this without measuring. The common version is a shop that changes elements on a calendar and never samples: the elements go in on time, the machine keeps failing pumps, and nobody learns the ingression path was a breather. The second version takes one sample, sees a bad number, and adds filtration without finding the input, which buys whatever the new element buys and leaves the loop running.

Taking a sample that means something

• Use a dedicated sampling valve or probe. Do not crack a fitting or loosen a plug on a pressurized line to get fluid; escaping oil injects through skin at pressures far below any working pressure in these circuits, and that hazard is covered in the card on what a hydraulic system is trading away.
• Sample hot and running. Cold, still oil has let its particles settle, the same error the discarded first sample made in another form.
• Flush and discard several times the sample line's own volume before you fill the bottle, and use a clean certified bottle, because the bottle contributes particles too.
• Wear nitrile gloves and eye protection and wash after handling used fluid. Where the machine runs a fire-resistant phosphate-ester fluid rather than a petroleum oil, the skin and eye handling is different and that fluid's own safety data sheet governs, not the habits you built on mineral oil.
• Sample the same point, the same way, every time. A trend across four samples from one port tells you more than one absolute number, because whatever the method contributes it contributes consistently, so it falls out of the comparison.
• Never run a machine with an element removed or a filter bypassed to speed up a flush. A tripped bypass indicator means the element is already passing unfiltered oil, and that is a reason to stop, not to keep going.

References

• ISO 4406 for fluid cleanliness coding, in the edition your machine builder's specification or your service contract calls out, which is what binds you rather than any regulation, and noting that the 1999 revision changed the size channels and the counter calibration reference
• Machine and component manufacturer documentation for the required cleanliness code, the working clearances and the filter element specification for each housing
• Safety data sheet for the specific hydraulic fluid, which governs skin and eye handling and disposal of used fluid
• See related: What a Hydraulic System Is Actually Trading Away; What a Pressure Compensated Pump Changes About the Circuit; What a Directional Control Valve Is Doing in Each Position