The Failure That
Arrives Two Years Late

Early in my career I inherited a warranty case involving an electric
power steering module. The gearbox had begun to whine at low speed, and
teardown revealed metallic flakes distributed through the lubricant —
particles far too small to have been visible at any inspection station
on the line. The supplier had a clean assembly area, gloves, and good
intent. What they lacked was any systematic understanding of residual
contamination: the metal, fibre and mineral debris that rides inside
components from casting, machining, stamping and handling, and only
declares itself once the product is in service.

This is the defining problem of technical cleanliness. The
contamination that matters rarely fails the end-of-line test. A hard
metallic particle of a few hundred microns sits harmlessly in a
reservoir until vibration, thermal cycling or a pressure pulse moves it
into a narrow clearance, a valve seat, a ball track or a solder joint.
By then the trace back to the responsible process step is cold, the
production lot is dispersed across a continent, and the warranty claim
arrives with an unpleasant per-unit cost attached. Managing cleanliness
is therefore about preventing a failure you will never see on your own
test rig.

Technical cleanliness, as a discipline, treats the product itself as
the inspection lot. Rather than examining surfaces directly, we extract
the contamination and quantify it: gravimetrically by mass, and
particle-counted by size classes, with metallic content identified where
the application demands it. The methodology is codified in VDA Volume 19
(and its ISO 16232 counterpart), and it is the working language between
customer and supplier when cleanliness specifications are set. Doing it
properly changes how you build, not merely how you inspect.

How Extraction Actually
Works

The principle behind VDA 19 style extraction is simple: agitate the
part in a defined fluid under defined conditions, filter the fluid, and
analyse what the filter catches. The practice is anything but casual.
Parts are rinsed, sprayed or ultrasonically agitated — sometimes in
combination — using the method that best matches the geometry. Internal
channels and blind drillings need pressure rinsing or ultrasonic
support; large flat surfaces may be spray-rinsed with a defined pattern.
The fluid, its volume, temperature and pressure are all specified,
because the extraction efficiency of a method determines whether your
results mean anything at all.

Blank values matter more than most people expect. Every extraction
consumes filters, fluid, glassware and handling steps, each of which
contributes its own particle load. A blank run — the full procedure
without a part — must be executed and shown to be low relative to
expected part values, otherwise you are counting your own laboratory. I
have seen suppliers chase phantom particle sources for weeks because
their flush fluid was filtered to a specification coarser than their
limit values. The blank is not paperwork; it is the floor of your
measurement capability.

Validation of the extraction is what separates a number from a
result. If you rinse a part once, extract it a second time, and recover
a substantial share of the original load, your method is leaving
contamination behind. Extraction efficiency checks of this kind —
repeating the extraction on the same part until the recovery falls off —
tell you whether your declared value represents the part or merely the
fraction your method happens to reach. Geometry often dictates that full
extraction is impossible; that is acceptable, provided the method is
consistent and agreed with the customer, because you are managing a
relative indicator as much as an absolute one.

Finally, the environment. Analysis of the filter membrane happens
under a microscope in a clean bench, because ambient dust in a standard
quality lab will swamp anything the part contributed. Staining fluids
help distinguish metallic particles from fibres and non-metallic debris
on the membrane. The whole chain — extraction, filtration, membrane
handling, imaging — has to be controlled as a process, with competence
built deliberately. Half-implemented cleanliness labs produce confident
nonsense.

Reading
the Numbers: Limit Values That Mean Something

A cleanliness specification is usually built from three tiers of
limits, and each tier answers a different failure question. The first is
gravimetric: total mass of contamination per part, expressed in
milligrams, sometimes normalised to component mass or functional area.
Mass limits are cheap to verify and useful as a coarse process
indicator, but they are blind to particle size. A single large shaving
and a fine sludge of dust can weigh the same, with entirely different
consequences downstream.

The second tier is particle counting by size class — the real heart
of VDA 19 style reporting. Membrane analysis yields the number of
particles in coded size bands, typically from 50 micrometres upward,
extending down to 5 or 15 micrometres where functional risk demands it.
Size matters because geometry decides the failure mode: a particle must
fit into a clearance to wedge there, must bridge a land to block an
orifice, must be conductive and long enough to span contacts in
electronics. Size classes let you set limits where the physics of your
product lives.

The third tier is material characterisation: distinguishing hard
metallic particles (often further split into ferrous and non-ferrous,
using magnetic separation or reflective analysis) from soft or
non-metallic debris such as fibres, plastics and minerals. Hard metal is
the usual villain in bearings, valve seats and pump internals, where it
embeds, scores and initiates wear chains. Non-metallic particles, by
contrast, tend to matter in fluid systems as blockage agents and in
electronics as ionic contamination carriers. A limit value without
material discrimination throws these populations into one bucket.

Setting the limits themselves is an engineering judgement, not a copy
from the customer’s generic table. Start from functional clearances in
the design: the narrowest oil film, the smallest orifice, the tightest
ball-track geometry, the shortest creepage distance on a populated
board. Then decide what safety factor against those dimensions your
warranty position can support, and negotiate. I always advise suppliers
to know their own process capability against a limit before signing up
to it — a limit you cannot hold with margin is a delayed field claim
with your name on it.

Cleanliness as a Warranty
Driver

The uncomfortable economics of particulate contamination lie in the
detection delay. A missing screw is caught at end of line; a burr that
sheds fragments is caught by the customer, after thousands of operating
hours, in a warranty database. Field failures attributed to
contamination are notoriously under-analysed, because by the time the
part returns, the evidence has been ground into the lubricant or washed
into a sludge. The claim gets coded as “bearing failure” or “valve
sticking,” and the root cause — debris introduced at your or your
supplier’s bench — never surfaces as a cleanliness problem.

This is precisely why cleanliness limits belong in the warranty
conversation, not just in the quality agreement. When I negotiate
supplier contracts, residual cleanliness is treated the same as
dimensional capability: specified, measured on a sampling basis, and
tied to defined reactions when exceeded. The supplier who monitors
extraction results as a trend, not merely a pass/fail gate, gets early
warning of a degrading deburring process or a failing chip-evacuation
step — months before the field notices. That trend is, in effect, an
insurance policy priced in milligrams.

The second warranty mechanism is the mixture effect. Assembly is
where contamination from many contributors converges: purchased castings
with residual sand, machined housings with chips in cross-drillings,
stampings with slitter debris, fasteners with rolling-scale, and the
ambient particle load of your own hall. Each contributor may sit
comfortably inside its own limit, yet the sum in the finished product
crosses the functional threshold. Field failures from accumulated
cleanliness are then nobody’s clear responsibility, which makes the
commercial argument between supplier and OEM long, bitter and expensive
for both.

My practical advice: when a field failure shows wear debris, scoring
or blockage at teardown, run a full cleanliness-style analysis on the
returned unit and on retained samples from the same production period.
Retained samples are the single most valuable thing a plant can hold for
this purpose. Without them, you are reconstructing a crime scene from a
photograph of the wreckage.

The Failure Modes Worth
Knowing

Wedge-type abrasive wear is the classic. A hard particle of the same
order as the clearance enters a sliding or rolling interface, embeds in
one surface, and machines the counter-surface. In ball screws, EPS
gearsets, fuel injectors and compressor bearings, this manifests as
noise first, friction and heat second, seizure eventually. The particle
sizes that do this are typically the ones just above the operating film
thickness — which is why limit values tied to functional clearance,
rather than a generic table, actually protect you.

Orifice and passage blockage is the second family, dominant in
hydraulics, braking and thermal systems. Spool valves, damping orifices
and nozzle needles are designed with flow areas that a single fibre or a
cluster of fine particles can obstruct. Symptoms are intermittent and
notoriously temperature- and vibration-dependent — the failure that
appears in the field and refuses to reproduce on the bench. Fibres
deserve particular suspicion: they are long relative to their width,
they tangle, and they arrive from wiping cloths, packaging and clothing
despite everyone’s best intentions.

In electronics, the modes shift from mechanical to electrical.
Conductive metallic particles bridging adjacent pads or connector
contacts cause intermittent shorts that come and go with vibration — the
hardest category of field complaint to diagnose. Non-conductive
contamination contributes differently: hygroscopic residue plus dust
supports leakage currents and corrosion under bias, especially in humid
climates. Assembly cleanliness for control units therefore couples
particulate control with ionic cleanliness methods, and the two
specifications should be developed together, not bolted on
separately.

Seal and diaphragma damage rounds out the set. Elastomer seals are
cut and abraded by hard particles trapped at the lip, and a single nick
in a hydraulic seal becomes a slow leak that a customer tolerates until
it becomes a fast one. Membrane-style diaphragms in pressure sensors and
pumps fail similarly. In all these modes, the message is the same: the
particle does its work over time, in the field, under loads your test
cycle never quite reproduces.

Designing a
Cleanliness Chain in Production

Cleanliness is won or lost at the process steps that generate debris,
not at the inspection that counts it. Machining leaves chips and burrs;
casting leaves sand and shell residue; stamping leaves slitter burrs and
slugs; welding leaves spatter. Each generator needs its own removal step
— deburring, flushing, ultrasonic cleaning, chip evacuation from
drillings — validated to a defined end condition. If deburring of a
cross-drilled oilway is your critical step, then that step needs process
parameters, monitoring and a cleanliness-based validation, treated with
the same seriousness as a heat-treat furnace.

Handling and logistics are the quiet saboteurs. A part cleaned to
specification and then packed in a corrugated carton that sheds fibres,
or transported with open cavities facing upward collecting hall dust,
has simply been recontaminated by the time it reaches assembly.
Packaging specification — material, closure, orientation, single-use
rules — belongs in the cleanliness plan, as do rules for interim storage
and re-opening. I have found recontamination between cleaning and
assembly to be among the most common findings, and among the cheapest to
fix once someone owns it.

Within the assembly hall itself, the standard toolkit applies:
segregated flows for cleanliness-critical parts, defined routing that
avoids cutting, drilling or filing in their vicinity, controlled wiping
materials, and air quality appropriate to the product class. Full
cleanrooms are rarely necessary for mechanical assemblies; a disciplined
zone with controlled airflow, correct garment policy and
cleanroom-compatible consumables usually suffices. The aim is
proportionality — spend where functional risk is, not uniformly
everywhere, which is the fastest way to exhaust any budget and any
patience.

Measurement then closes the loop. Establish a sampling rhythm for
extraction analysis on critical part numbers, plot the trends, and
define reaction limits below the contractual limit so that drift
triggers action before rejection. When a lot fails, the analysis of
particle type and size often names the culprit process directly — cast
iron debris points to a machining step, fibre clusters to packaging,
stainless flakes to a fastener lot. That diagnostic power is the
everyday payoff of a competent cleanliness laboratory.

Auditing Suppliers
and Defending the Field

When I audit a supplier against cleanliness requirements, I ask a
short list of sharp questions. Where does the specification come from —
functional analysis or a copy-pasted customer table? What is the
extraction method, its validated efficiency, and its blank values? Are
limit values trended, or only gated? Who analyses the membranes, in what
environment, with what training? Can they show a case where cleanliness
data caught a process problem early? A supplier who can answer those six
things coherently is managing contamination; one who produces only a
certificate is measuring it, at best.

On the customer side, my message is equally blunt: cleanliness is a
shared system. Specify sensibly, audit the method rather than the
paperwork, and give suppliers access to field teardown findings so that
limit values can be recalibrated against reality. Nothing improves a
cleanliness programme faster than an engineer seeing, under the
microscope, the exact particle population recovered from a failed unit
they shipped.

The field is the final laboratory. Every returned unit with
contamination-driven failure is free tuition in where your limits are
too loose, your extraction too gentle, or your packaging too optimistic.
Build the discipline — extraction to VDA 19 or ISO 16232, limits
anchored in functional geometry, trends watched, suppliers audited on
method — and the particles you cannot see stop becoming the warranty
claims you cannot explain.


Peter Stasko is a Quality Director with 25+ years of experience
leading quality management systems across the automotive and aerospace
industries.