I have spent a good part of my career chasing bonded joints that failed in the field, and in nearly every case the post-mortem pointed to the same culprit: the surface, not the adhesive. Bonding is a special process in the fullest sense. You cannot look at a finished joint and tell whether the chemistry between adhesive and substrate is sound. The fillet looks perfect, the cure appears complete, and yet the joint peels apart with adhesive separating cleanly from the substrate in what we call adhesive failure rather than cohesive failure within the glue line.
That distinction matters more than almost anything else in bonded assembly. A cohesive failure, where the adhesive itself tears, tells you the surface preparation was adequate and the load exceeded the strength of the adhesive. An adhesive failure, where the bond releases from the substrate, tells you the surface was contaminated, poorly activated, or chemically unsuited to the adhesive system. When your destructive tests keep showing adhesive failure, your process is telling you something before your customers do.
The difficulty is that surface condition is invisible at the point of bonding. An operator cannot see a monolayer of mould release agent transferred from a supplier's forming tool, or silicone plasticiser migrated from a nearby gasket operation. Neither can an inspector. Quality must therefore be built into the process through disciplined surface preparation, verified surface energy, controlled contamination pathways, and a destructive testing programme that samples the invisible variable on your behalf.
Wetting and Surface Energy: The Physics You Must Get Right
Adhesion begins with wetting. The liquid adhesive must spread across the substrate, flow into its microstructure, and establish molecular contact before any chemical or mechanical bonding can occur. Wetting happens when the surface energy of the substrate exceeds the surface tension of the adhesive. Most structural adhesives have surface tensions in the range of roughly 30 to 40 mN/m, so the substrate needs to sit comfortably above that threshold, with margin, for the bond to form properly.
Many engineering materials arrive below that threshold. Polypropylene and polyethylene are notoriously low-energy surfaces; composites may carry release agents; even metals come with rolling oils, oxides and machining residues. Raising and activating the surface through plasma treatment, flame treatment, corona, abrasion followed by solvent wipe, or chemical etching in aerospace practice is not an optional refinement. It is the process step that makes bonding possible at all.
What should you actually measure? Dyne inks remain the shop-floor workhorse: a defined surface-tension fluid brushed onto the surface, with wetting or beading observed within a couple of seconds. Water-break tests on metals serve a similar purpose, since a continuous unbroken water film indicates a clean, high-energy surface while rivulets betray contamination. Contact-angle measurement with a goniometer gives a quantified, recordable value and is increasingly practical on production lines.
Whatever method you choose, measure after treatment and before bonding, and treat the measurement as a process control rather than a certificate to file away. A dyne check performed weekly on a convenience sample tells you nothing about the part going into the press today. The measurement belongs in the control plan, at the station, with a defined reaction plan when it fails.

Contamination Control: The Silent Killer of Bonds
Contamination is where most bonding processes quietly fall apart. Silicone is the classic offender: a single molecular layer is enough to destroy wetting entirely, and it migrates with ease from release agents, sealing compounds, cable insulation and even hand creams. Oils and greases from compressed-air blow-offs, particulates from trimming operations, and plasticiser migration from adjacent polymers all work against you. I have seen entire batches scrapped because a maintenance technician used a silicone-based spray on a conveyor two bays upstream of the bonding cell.
Zoning is your first line of defence. Physically separate the bonding area from operations that generate contaminants, control airflow so particulates drift away from rather than towards prepared surfaces, and restrict consumables in the cell to those verified as silicone-free. Gloves matter: nitrile rather than vinyl, changed frequently, because a glove that has touched a crate or a trolley rail carries whatever was on it. Time windows matter too, because a plasma-treated surface begins to re-equilibrate and attract airborne contamination the moment treatment stops. Define and enforce a maximum interval between preparation and bonding.
Verification of cleanliness deserves the same rigour as the bonding itself. Beyond dyne and water-break checks, consider analytic techniques for high-risk applications: FTIR spectroscopy of wipe samples, photoemission methods for detecting thin-film residues on metals, or gravimetric testing of cleaning baths. In aerospace bonding, process control of surface preparation is audited as heavily as the cure itself, and rightly so. A contamination failure found at final assembly is unfixable; one found at the preparation station costs you a re-clean.
Where Bonding Processes Fail
What failing plants do
- Dyne checks done occasionally, results filed not acted on
- Silicone sprays and release agents stored anywhere convenient
- No defined window between surface treatment and bonding
- Destructive test results read as pass or fail loads only
What controlled plants do
- Surface energy verified per control plan, with a stop reaction on failure
- Bonding cell zoned, airflow managed, consumables verified silicone-free
- Maximum preparation-to-bonding interval defined and enforced
- Failure mode trended coupon by coupon, sampling escalated on any change
The Special Process Discipline: What You Must Control and Record
Because you cannot inspect bond quality into the finished part, the process itself becomes the product. Every parameter that influences the joint must be defined, controlled and recorded: surface preparation method, treatment intensity, time between preparation and application, adhesive batch, mix ratio, application pattern, open time, fixturing pressure, cure temperature and duration. If a parameter is not recorded, assume an auditor will ask how you know it was correct, and assume the honest answer is that you do not.
Personnel qualification is part of this discipline. Bonding operators and the engineers who set the process should be trained and certified against recognised practice; in aerospace, dedicated bonding personnel qualification schemes exist precisely because process discipline and operator technique carry so much of the quality burden. Training must cover the reasons behind the rules, not just the rules themselves. An operator who understands that touching a prepared surface with a bare finger can halve bond strength will behave differently from one who has merely been told not to.
Records must allow traceability both backwards and forwards. Which adhesive batch, applied by which qualified operator, to which surface preparation, cured in which oven run, on which serial numbers. When a field failure occurs, that traceability lets you bound the affected population instead of guessing. When a cure oven is later found to have drifted out of calibration, those records tell you exactly which joints to review. Without them, a special process is an act of faith.
Cure Verification: Proving the Invisible Conversion
Cure is the second invisible transformation. The adhesive goes into the oven or sits through its room-temperature cure window as a liquid or paste and emerges looking solid, but appearance tells you nothing about the degree of crosslinking. Under-cured adhesive is soft, weak and prone to creep under load; over-cured or overheated adhesive can be brittle or degraded. Neither condition is visible from outside the joint.
Direct verification methods exist and should be matched to the risk. Differential scanning calorimetry on a sample tells you whether residual exotherm remains, indicating incomplete cure. Hardness testing of the bondline or of witness coupons gives a quick proxy. Lap-shear specimens cured alongside production parts, then pulled to destruction, provide the most convincing evidence for structural applications.
Thermocouples placed on actual parts during oven cure, not merely measuring oven air temperature, reveal the gap between the oven display and the adhesive's true thermal history. That gap is routinely larger than people expect, particularly with heavy parts acting as heat sinks. Log the part temperature profile for every cure cycle and alarm on deviation.
The adhesive was rarely the problem. It rarely is. Everyone trusted the adhesive, and almost no one was watching the surface.
Adhesive shelf life and out-of-freezer work-life must be tracked with the same seriousness. An expired adhesive can cure to full hardness yet deliver a fraction of its specified strength, which is precisely the kind of failure that escapes visual checks and appears only in the field. I would rather scrap an hour of production over a suspect work-life record than sign off a batch on a verbal assurance.
Verification Chain for a Bonded Joint
- 01Measure surface energyDyne, water-break or contact angle after treatment, before bonding
- 02Control the time windowMaximum interval from preparation to adhesive application, enforced
- 03Record cure profileThermocouples on the part, not on the oven air; alarm on deviation
- 04Test witness couponsLap-shear and peel samples from current production conditions
- 05Trend failure modeCohesive failure is health; drift towards adhesive failure is an early alarm
Destructive Testing: Your Window into the Joint
Since the joint cannot be inspected non-destructively in any general way, destructive testing becomes your primary evidence. The principle is simple: sacrifice representative samples, at a defined frequency, from real production conditions, meaning the same surface preparation, the same operator, the same adhesive lot and the same cure, and break them. The mode of failure tells you the truth about the process, and the numbers tell you whether the truth is acceptable.
Lap-shear testing to established standards is the baseline for most structural joints, with peel testing added wherever the geometry or loading involves peel, which is far more punishing than shear. Wedge testing, where a wedge is driven into the bondline and crack propagation is monitored under hostile environmental conditions, is the aerospace method of choice for detecting weak boundary layers and marginal surface preparation. A poorly prepared surface shows accelerating crack growth in a way that static strength tests can miss entirely. Cross-sectioning of coupons reveals voids, unbonded areas and bondline thickness anomalies.
What you must insist on is the failure-mode record, not just the load figure. A coupon that exceeds minimum strength but fails adhesively at the interface is a warning, not a pass. Cohesive failure with adequate strength is the healthy state. Trend the failure modes over time, because a drift from cohesive towards adhesive failure is an early indicator of degrading surface preparation, a slipping plasma nozzle, or a new contaminant entering the cell. Escalate the sampling when anything in the process changes, whether a new substrate supplier, a new adhesive batch or maintenance on the treatment equipment. Destructive testing only protects you if the samples represent current conditions.
Making the Invisible Variable Visible
Treat surface condition as a defined product characteristic, not a background assumption. Put surface energy limits on the drawing or in the process specification, verify them with a stated method at a stated frequency, and give the operator authority to stop when a surface fails the check. Coupled with disciplined contamination zoning, fully parameterised and recorded bonding processes, thermal-profiled cures, and a destructive testing programme interpreted by people who understand failure modes, that converts an uninspectable process into a controlled one.
The failures I have investigated across automotive body shops and aerospace assembly halls share a pattern: everyone trusted the adhesive, and almost no one was watching the surface. The adhesive was rarely the problem. It rarely is. Respect the chemistry, control the surface, prove the cure, and break coupons honestly. That is how you earn confidence in joints you will never truly see.
