The sump was charged three weeks ago at 8.0%. The refractometer still reads 8.0. pH is 9.2. And the machine already stinks, cast iron parts come off the tool with rust spots in the pockets, and tool life is down a third. Nothing in the concentrate changed. Three numbers did, and none of them appear on the concentrate datasheet.
Metalworking fluids are unusual among industrial chemicals in that their service life is decided mostly by what happens to them rather than what is in them. Additive depletion is real, but it is slow and predictable. Contamination is fast and invisible. This article is about the three physical quantities that do the killing, and the arithmetic that makes each of them a purchasing and maintenance decision rather than a smell you complain about.
Three numbers that are never on the datasheet
Every coolant TDS lists concentration, pH of the emulsion, foam, corrosion performance and often a refractometer factor. None of them lists the three quantities below, because all three are properties of the system the fluid lives in, not of the drum it arrived in.
| Quantity | How it is measured | Well-managed target | What happens above it |
|---|---|---|---|
| Tramp oil | Skim test, separator, or lab analysis of the free oil layer against sump volume | Under 2% of sump volume | Oxygen transfer falls, anaerobic bacteria bloom, wetting and corrosion protection drop, oil mist and residues rise |
| Suspended fines | Gravimetric (mg/L) or ISO 4406 particle count | Under 100 mg/L | Abrasive wear at the cutting edge, scoring of bushings and pump seals, plugged high-pressure orifices |
| Filtration rating | Beta ratio at a stated size, ISO 16889 multipass | Absolute rating where the machine demands it | Particles above the rated size circulate freely while the label still says the right number |
Tramp oil: a threshold, and a piece of arithmetic
Tramp oil is any oil that enters the sump without being asked: hydraulic oil from a weeping cylinder, way lube from a metering pump, spindle and gearbox oil, rust preventive carried in on parts from an earlier operation, and whatever gets washed off the floor during cleaning. It exists in three states, and the states behave differently. Free oil floats on the surface. Dispersed oil is mechanically spread through the fluid by the pump and will rise to the top if the system is left quiet long enough. Emulsified oil has been stabilised into the emulsion by the fluid's own surfactants and will not separate at all.
Mineral oil has a lower specific gravity than water, so free oil forms a continuous film across the sump surface. That film is the problem: it seals the fluid off from air. Oxygen transfer stops, aerobic organisms die back, and anaerobic strains multiply, releasing hydrogen sulphide. The smell is the symptom; the mechanism is asphyxiation. At the same time the film reduces wetting, so chips and part surfaces lose their corrosion protection, and the detergents in some way lubricants generate foam that the fluid's antifoam was never sized for.
The thresholds in the literature are unusually consistent for a field this messy:
- Below 2% of sump volume, few observable or quantifiable problems.
- Above 2%, bacterial activity begins to rise measurably.
- Above 4%, tool life, surface finish, oil mist and residues all become problems at once.
- Source matters more than percentage. As little as 0.5% of a wax-based rust preventive can cause what 4% of way lubricant does, because the wax forms a tacky film that holds chips and blocks wetting entirely.
Now the arithmetic that makes the 2% line a schedule rather than an aspiration. A drip of oil at one drop per second, at the conventional 20 drops per millilitre, is 0.05 mL per second. Over a day that is 4.3 litres. Over a week, 30 litres. In a 1000-litre sump that is 3.0% - through the 2% threshold and halfway to the 4% threshold in seven days, from a leak nobody would call a leak. A 2000-litre central system gives you two weeks instead of one.
Which is why the control measure is not a bigger skimmer but a continuously running one. A belt or disc skimmer that runs on a timer for six hours leaves an unbroken film on the surface for the other eighteen, and the anaerobic population does not observe the timetable. Dispersed oil needs a coalescing separator; emulsified oil needs a centrifuge or it does not come out at all.
There is a diagnostic tell worth teaching every operator: a hazy or fuzzy boundary line on the refractometer is usually tramp oil, and the reading itself is inflated, because oil has its own refractive index and the instrument cannot tell it from concentrate.
Fines: what an ISO 4406 code actually means inside a sump
The human eye resolves about 40 micrometres. Human hair is 50–70 µm. Everything that matters in a coolant is below that line, which is why "the fluid looks clean" is not an observation, it is a guess.
ISO 4406 reports particle counts per millilitre in three cumulative channels - at or above 4, 6 and 14 µm - as a three-number code on a logarithmic scale, where each single-step increase doubles the count. A code of 18/15/12 means 1,300–2,500 particles per mL above 4 µm, 160–320 above 6 µm, and 20–40 above 14 µm.
Put that into a 1000-litre sump. One thousand litres is one million millilitres. At 2,000 particles per mL above 4 µm, the sump contains roughly two billion particles. The fluid is still translucent. The operator still calls it clean. This is the single most useful number in the article, because it converts an abstract code into a quantity nobody can dismiss.
The mass view is complementary and easier to write into a log: a well-managed sample typically carries under 100 mg/L of suspended fines, which is 100 g in that same 1000-litre sump - and that is only the fraction still suspended, not the sludge on the bottom.
Size matters more than mass. Machine clearances in lubricated contacts are on the order of 1–10 µm, and the particles that do the most damage are the ones comparable to the clearance. The third channel of the code, above 14 µm, is the one that scores guide bushings and erodes mechanical seal faces; the fourth problem, specific to high-pressure through-spindle coolant, is that fine particles accumulate in drill-head orifices and quietly reduce flow.
One practical trap: magnetic separators only remove ferromagnetic material. Cast iron and carbon steel respond. Aluminium, copper alloys and austenitic stainless steel do not. A shop cutting aluminium with a magnetic separator and no other filtration is removing almost nothing, and no amount of maintenance will fix a device that is physically unable to capture the contaminant.
Fines also have a second-order effect that connects directly to the next section. Sludge settles and displaces fluid, so the effective sump volume shrinks. Less volume means less thermal mass and a faster temperature rise; above roughly 30 °C microbial growth accelerates sharply and evaporation climbs, which drives the concentration creep described below.
Filter ratings: the beta ratio is the only number that means anything
Two elements can both be sold as "10 micron" and differ by two orders of magnitude in what they let through. The reason is that a micron number without a test method is a description of a label, not of a filter.
| Rating | Definition | Efficiency at the stated size | Where it belongs |
|---|---|---|---|
| Nominal | Loose, supplier-defined capture at the stated size; no universal test | Typically 50–95% | Prefiltration, bulk solids, cheap protection ahead of a fine element |
| Absolute | Verified by ISO 16889 multipass, expressed as a beta ratio at that size | 98.7% or better (beta 75 and up) | Final filtration where the downstream component is expensive |
The beta ratio is simply the upstream particle count divided by the downstream count at a stated size, and efficiency follows as (beta - 1) / beta:
| Beta ratio at 10 µm | Particles passing, per 100 entering | Efficiency |
|---|---|---|
| beta 2 | 50 | 50.0% |
| beta 10 | 10 | 90.0% |
| beta 75 | 1.3 | 98.7% |
| beta 200 | 0.5 | 99.5% |
| beta 1000 | 0.1 | 99.9% |
Read the first and fourth rows together. Both elements are legitimately "10 micron". One passes half of everything at that size, the other passes one particle in two hundred. If you specify filtration by micron number alone, you have specified nothing you can hold a supplier to. Ask for the beta value, and ask at which size - a beta of 200 at 25 µm says nothing at all about 10 µm.
Matching the rating to the machine is where the money is. A general return-line element in the 10–25 µm range is normal for flood coolant. High-pressure through-spindle systems, 70 bar and above, are far less tolerant: a 50 µm particle that a flood machine never notices will erode a spindle seal, so the final stage wants an absolute rating around 10–20 µm.
Two operational points that are routinely missed. First, elements are changed on differential pressure - commonly around 2.5 bar - not on a calendar, because a filter that is loaded is a filter that is bypassing. Second, check whether the housing has a bypass valve: if it opens at 3 bar and the element is changed at 4, the system runs unfiltered exactly when the dirt load is highest.
And the question everyone asks: does fine filtration strip the additives? Generally no. Corrosion inhibitors and lubricity additives are dissolved or emulsified at molecular scale, orders of magnitude below a 10 µm pore, so a filter captures solids, not chemistry. The exception is a coalescing separator, which can carry some emulsified oil with the separated phase - so after any change to the cleaning setup, re-measure concentration before concluding the fluid has changed.
Why a filter cannot replace a skimmer
This is the part that catches people who have just bought a good filter. A soluble-oil emulsion has oil droplets in the range of roughly 0.2–10 µm. Droplet size is not a constant; it is set largely by how well the base oil is solvated by the emulsifier system. In a published comparison of model metalworking fluids, naphthenic base oil formulations produced mean droplet sizes of 1 µm and below, while the paraffinic oils, at their optimum HLB, never got below 10 µm - more than twenty times larger.
A 10 µm absolute element passes a 1 µm oil droplet completely. The filter is doing its job; the droplet is simply not a particle. Filtration separates by size, skimming separates by density, and no amount of filter quality substitutes for the density difference. Conversely, a skimmer cannot touch fines, because fines do not float.
The corollary is for formulators rather than users: base-oil solvency, which the aniline point measures, sets droplet size, and droplet size sets how much of your emulsion a downstream filter will treat as particulate. A fluid built on a well-solvated naphthenic base is both more stable and less likely to be damaged by aggressive filtration.
The top-up arithmetic, or why the refractometer creeps upward
Here is a piece of bookkeeping that explains a complaint heard in every plant: "nobody added concentrate, but the reading went up." Oil leaves a sump in exactly one way - drag-out on chips and parts. Water leaves in two ways: drag-out, and evaporation. That asymmetry is the whole story.
If a sump runs at concentration C, loses D litres a day to drag-out and E litres a day to evaporation, then to restore the level you must add D + E litres, but you only need to replace C x D litres of oil. The correct top-up concentration is therefore:
- Top-up concentration = C x D / (D + E)
Work an example. A 1000-litre sump at 8%, losing 10 litres a day to drag-out and 20 litres a day to evaporation. You add 30 litres a day, and you owe the sump 0.8 litres of oil. That is a 2.7% top-up, not 8%. Top up at 8% instead and you add 2.4 litres of oil against a 0.8-litre loss: a surplus of 1.6 litres a day, 11 litres a week, and the charge climbs from 8.0% to about 9.1% while the free-oil layer thickens and the fluid starts to leave a tacky residue on the machine.
This is also where the familiar rule of thumb comes from. "Top up at half the running concentration" is exactly right when drag-out and evaporation are equal - put D = E into the formula and you get C/2. It is wrong when evaporation dominates, which is the normal case in summer, on small sumps, and around high-speed spindles; there the correct top-up is well below half. The rule is not a law, it is one point on a curve, and the curve is the formula.
The same arithmetic explains the rising reading without any addition: the water evaporated, the oil stayed. And it explains why a refractometer reading should never be used as the concentration of record when tramp oil is present, because the instrument reads the oil phase too. Titration, or the supplier's stated factor applied to a clean sample, gives concentration; the refractometer gives a trend. The two disagreeing is itself the most useful signal you have.
A weekly log with four lines and two thresholds
None of this requires a laboratory. It requires four numbers, written down weekly, at the same time of day and from the same sample point. A written trend turns a smell complaint into a case for capital, which is the only form of argument that gets a filtration upgrade approved.
| Line | How | Target | What a bad number tells you |
|---|---|---|---|
| Concentration | Titration, plus refractometer for trend | Set point ±10% | Refractometer rising while titration is flat means water loss or tramp oil, not a rich mix |
| pH | Meter, daily on critical machines | 9.0–9.5 fresh | Below 8.5 the corrosion package is gone; below 8.0 the sump is a culture |
| Tramp oil | Skim test or separator reading | Under 2% | Back at 2% a week after skimming means find the leak, not buy a bigger skimmer |
| Suspended fines | Gravimetric mg/L or ISO 4406 count | Under 100 mg/L | The 14 µm channel climbing means the element is spent or the housing is bypassing |
| Temperature | Sump probe | Below 30 °C | Above it, microbial growth accelerates and evaporation drives concentration creep |
Two thresholds decide whether the fluid is worth saving. If pH will not hold 8.5 after correction, or tramp oil returns to 2% within a week of continuous skimming, the charge is finished - and refilling is almost always cheaper than chasing a dead sump with biocide, which also risks releasing endotoxins when a system is already out of control. A third signal is a persistent divergence of more than 20% between titration and refractometer after cleaning, which usually means the emulsion has taken up enough foreign oil that it is no longer the product you bought.
What the concentrate formulator can actually do
Formulators cannot skim a customer's sump, but the fluid can be built to survive a badly maintained one.
- Choose the base oil for solvency, not price. A lower aniline point means better solvation, smaller droplets and a more stable emulsion at the same emulsifier dose - the 1 µm versus 10 µm difference above comes almost entirely from base-oil selection.
- Build hardness tolerance into the emulsifier package. Anionic systems are the sensitive ones; a fluid that splits at 400 ppm hardness will be blamed on the customer's water and will not be reordered.
- Keep foam margin, because the customer's sump is dirtier than your lab beaker. Detergent-bearing way oils and cleaning residues are foam generators you cannot formulate against, so leave headroom.
- Pick lubricity additives that survive the cleaning loop. butyl oleate (CAS 142-77-8), butyl stearate (123-95-5), oleamide (301-02-0) and stearamide (124-26-5) are the ester and amide workhorses of the boundary-lubricity package; they are also the components most affected by a coalescer that carries emulsified oil out of the system.
- Use polyol esters where temperature and life matter. Pentaerythritol (115-77-5) and dipentaerythritol (126-58-9) based esters resist hydrolysis and oxidation far better than trimethylolpropane or simple diesters, which is why they appear in long-life and high-temperature systems.
- Treat preservatives as control, not repair. BIT (CAS 2634-33-5) and sodium pyrithione (3811-73-2) suppress planktonic bacteria effectively, but a biocide cannot reach an anaerobic biofilm living under an oil blanket. The blanket is the problem; the dose is not. Biocide in a heavily fouled sump mostly generates endotoxins and resistant strains.
- Choose the leaking oil deliberately. The hydraulic and way oils that end up in the sump set the ceiling on fluid life. Low-additive paraffinic white oils such as 5#, 10#, 22# and 32# (CAS 8042-47-5) contribute almost no detergents, no ZDDP and no wax film, and are the least harmful thing that can leak into a coolant. Anti-wear hydraulic oils such as HD HM T10, T22 and T100 do their own job well, but when they leak they bring zinc dialkyldithiophosphate and detergent chemistry into a system that was never formulated for either.
Procurement checklist
- Specify filter elements by beta value and size (for example beta 200 at 10 µm), never by micron number alone.
- Write a fines target in mg/L or an ISO 4406 code into the maintenance contract, not "clean coolant".
- Ask for the skimmer throughput in litres per hour and confirm continuous duty, not timer duty.
- Confirm the refractometer factor for the specific product; a factor borrowed from a previous supplier's fluid will systematically misreport concentration.
- Get the hardness and pH window in writing before the first charge, because that is the first thing that will be blamed if the emulsion splits.
Fluids do not die of old age. They die of an oil film that cut off their air, a sludge that ate their sump volume, and a filter label that promised a number nobody verified. Fix those three and most additive packages will comfortably reach the service life the formulator designed for.
Products referenced in this guide
We supply the base oils and additives that sit on both sides of this problem: the low-additive paraffinic white oils that do the least damage when they leak (5#, 10#, 22#, 32#, CAS 8042-47-5), the ester and amide lubricity components (butyl oleate 142-77-8, butyl stearate 123-95-5, oleamide 301-02-0, stearamide 124-26-5), the polyol cores for hydrolysis-resistant synthetic esters (pentaerythritol 115-77-5, dipentaerythritol 126-58-9), and the metalworking preservative pair (YS-MK, BIT, 2634-33-5; YS-HEY, sodium pyrithione, 3811-73-2). Full specifications, COA and sample quantities are available on each product page; tell us the sump volume, the metal being cut and the water hardness, and we will match the grade rather than the price.
Related reading: how to read concentration correctly on a refractometer (refractometer factors and dilution), preservative selection for industrial fluids (industrial preservatives), viscosity grade selection for the hydraulic oils that leak into your sump (hydraulic oil ISO VG), boundary lubricity additives (fatty amides and esters), and why synthetic esters last longer (polyol ester base stocks).