Diagnosing a failed windscreen bond

When a windscreen bond fails, the conversation usually starts with the adhesive and ends with the preparation. A ten-minute diagnosis sorts out which one it is, and it is the difference between a claim and a rework.

Step 1: look at how the bead separated

Remove the glass and inspect the failure surface.

What you see	What it means
Adhesive torn, residue on both glass and frame	Cohesive failure — the adhesive itself failed; check batch and cure
Clean glass, adhesive stuck to frame	Interfacial failure at the glass — contamination or frit problem
Clean frame, adhesive stuck to glass	Interfacial failure at the frame — paint or primer problem
Foamy, crumbly bead	Moisture or solvent trapped; check storage and substrate
Bead torn at one corner only	Discontinuous bead or movement during cure

Cohesive failure is the only pattern that points at the product. Everything else points at the process.

Step 2: read the frit

The frit (the black ceramic band on the glass) is the bonding surface, and it is not inert. A frit that is under-cured, contaminated or damaged will release the adhesive cleanly.

Look for: a glossy or waxy feel (release agent), fingerprints in the bond area, or a frit edge that lifts with a fingernail.

Step 3: measure the bead

Cut a section and measure it. A bead specified at 8 × 10 mm that measures 5 × 6 mm has half the section and a fraction of the movement capacity. Tooling a bead to a feather edge at the corners is the most common geometry failure.

The batch record closes the loop

If the failure is cohesive, the batch number is the only way to compare against the retained sample. Without it, the conversation is opinion against opinion.

Run the peel test before you argue

The visual classification is faster, but a peel test settles the question when the failure surface is ambiguous.

sample     100 mm bead on an offcut of the actual substrate, same preparation
cure       24 h at 20–25 °C, or the full cure time for the bead section
peel       lift one end at roughly 90°, pull steadily, watch where it releases
record     cohesive (adhesive tears) vs interfacial (clean substrate)

What the result means:

• Cohesive over most of the length — the adhesive is weaker than the bond, which is normal and healthy. A failure in service with this pattern points to cure, bead section or load, not to preparation.
• Interfacial at the start, cohesive later — the first centimetres were contaminated or the activator had not flashed off.
• Interfacial over the whole length — the surface won. Preparation, frit condition or the wrong primer.

Photograph the peeled strip against a ruler. The proportion of cohesive to interfacial area is the number that matters, and it is the one thing a supplier will ask for.

What the evidence pack should contain

If the failure is going to a supplier, an insurer or a customer, four items decide whether the conversation is technical or emotional:

Item	Why it matters
Photograph of the failure surface before cleaning	Once it is cleaned, the evidence is gone
Bead cross-section measurement	Distinguishes a geometry failure from a product failure
Batch number from the cartridge	Links the failure to a production record
Installation temperature and time	Explains cure state at the moment of failure

A workshop that keeps these four on every glazing job card can answer a claim in an hour. One that does not will replace a windscreen and hope.

The failure that is neither product nor preparation

There is a third category, and it is the one most often missed: the bead was correct and the glass was wrong. A chipped edge, a crack initiated at a stone strike, or a previously repaired chip in the bond area will propagate under the bead and release it. The adhesive did its job; the substrate failed first.

Look at the glass side of the bond line before blaming either the product or the process. A crack running out from under the frit is visible if you look for it, and it changes the whole diagnosis.

What this means for buyers

1. Photograph the failure surface before cleaning anything up.
2. Use the table above to classify before contacting the supplier — it changes the conversation.
3. Record batch numbers on job cards so a future claim is checkable.
4. Treat a single corner failure as a process signal, not a product signal.