Cross-cut, X-cut, pull-off and knife tests answer different questions. Choosing the wrong one produces a number that looks authoritative and means nothing.
Adhesion is the property most often argued about and least often measured properly. When a coating fails in service, the discussion moves quickly to whether it adhered in the first place, and by then the evidence has usually been painted over. Testing at the right moment with the right method is the only way to have an answer.
Cross-cut is for thin films only
The cross-cut test, described in ISO 2409, cuts a lattice of six by six lines through the film to the substrate and assesses how much coating lifts when tape is applied and removed. It is quick, cheap and repeatable, and it is widely misapplied. The method is intended for films up to roughly 250 micrometres. On a high-build epoxy at 400 micrometres, or a glass-flake system at 600, the blade does not reliably reach the substrate and the lattice cannot detach a film that is thick enough to resist the tape. The result is a classification of zero that reflects the thickness rather than the adhesion.
Where thick films must be assessed with a cut test, the X-cut variant in ISO 16276-2 is the appropriate route, and it still gives a comparative rather than an absolute answer.
Pull-off gives a number, and the number is not the point
The pull-off test in ISO 4624 glues a dolly to the coating, cuts around it to the substrate and pulls it off, reporting the stress in megapascals. It is the only common method that produces a quantitative value, which is why specifications quote acceptance figures in MPa. It is also the method most often misread.
The failure mode matters more than the number. A result of eight megapascals with cohesive failure inside the coating says the coating is weaker than its bond to the steel, and the number is therefore a property of the coating. The same eight megapascals with adhesive failure at the substrate interface says something entirely different about surface preparation. A result where the glue fails, or where the dolly detaches at the topcoat interface, says nothing about the substrate bond at all. Any report that quotes a value without the failure mode is incomplete.
Practical details also decide the outcome. The dolly must be aligned, the adhesive must be fully cured, the cut must reach the substrate, and the test must be run after the coating has reached its full cure rather than at the end of the application shift. A pull-off test on a coating that is still hardening measures the hardening curve.
Knife tests are for diagnosis, not acceptance
The knife test in ASTM D6677 is a qualitative assessment: a blade is used to lift the coating at an edge or a scribe, and the resistance is rated. It is fast, requires no adhesive or curing time, and is the right tool for a surveyor trying to establish whether an existing system is sound enough to overcoat. It is the wrong tool for an acceptance criterion, because two competent inspectors can rate the same area differently.
What to specify, and when
- Cross-cut at the specified dry film thickness only, and record the thickness with the result
- Pull-off where a numerical criterion is required, reporting stress, failure mode and location of failure
- Test after full cure, at a recorded temperature and film thickness
- Repair every destructive test point to the full system, not with a touch-up brush
- Test the same witness areas that were prepared and coated as production, not specially prepared panels
The most common practical mistake is testing too early and testing too little. Two pull-off tests on a structure of five thousand square metres is not a sample, it is an anecdote. The second most common is treating a number as a verdict without reading which layer came apart.