A laminated packet is an acoustic structure before it contains any defect. Steel, coatings, adhesive and contact boundaries can generate repeated reflections. An echo that looks unusual may represent a bond discontinuity, a normal layer boundary, a coupling change or several overlapping responses.
The first task is to establish the response of the intended construction. An inspection calibrated on solid steel cannot automatically distinguish a void inside a multilayer packet. Access, surface condition, packet thickness and the arrangement of bonded interfaces belong in the method qualification.
Create known comparisons
Prepare an intact reference and representative specimens containing independently known discontinuities. Vary defect position and relevant geometry deliberately. A surface disbond that produces a clear echo does not establish sensitivity to the same area buried behind many interfaces.
Ground truth needs a method independent of the ultrasonic classification. Sectioning a retained trial specimen or another suitable physical examination can establish defect extent, although destructive preparation must not create the very discontinuity being measured. Record that uncertainty explicitly.
| Reference comparison | Response sought | Capability being tested |
|---|---|---|
| Intact representative packet | Normal repeated echoes and variability | Background classification |
| Known near-surface discontinuity | Difference from intact reference | Detection with limited interface crossings |
| Known deeper discontinuity | Attenuation and echo overlap | Depth-dependent detection |
| Different surface coupling on an intact sample | Changed signal without a defect | Resistance to false calls |
The specimen series should include challenging sound paths, rather than only conspicuous defects that are easy to classify.
Preserve the raw signal
Retain time-domain traces and inspection coordinates as well as processed images. Color maps can conceal whether a classification resulted from a gate, threshold or filtering choice. Document the signal-processing settings used to distinguish a candidate indication from the intact response.
Repeat scans after removing and re-establishing coupling. Compare this variability with the difference between good and defective references. If recoupling changes the signal more than the defect does, the method has not yet established a reliable decision boundary for that defect class.
Trace the acoustic path through the actual stack. Changes in thickness or incidence can move expected echoes into a gate intended for a different interface. The inspection plan should state which regions and depths it can evaluate and which remain inaccessible or ambiguous.
Report a detected class, not a universal pass
A qualified result should identify the validated defect types, geometries, depths and accessible surfaces. It should also state how indications will be resolved. An ambiguous region may require another measurement or a representative destructive check rather than a forced binary classification.
Keep a coverage map for the known defect specimens and the tested multilayer construction. It should identify the locations and defect sizes the ultrasound method resolves, along with the regions where a bond discontinuity remains indistinguishable from the normal stack response.
Technical references
- voestalpine. Backlack: Lamination bonding.
- voestalpine. Backlack: Testing methods and production of test specimens.