Interpretation, Characterisation and False Calls

Learning outcomes

By the end of this chapter, the learner should be able to apply a structured interpretation process; distinguish relevant, non-relevant and false indications; use corroborating views and scans; manage uncertainty and analyst bias; and define escalation thresholds for ambiguous data.

13.1 The interpretation funnel

Interpretation should narrow possibilities in a controlled sequence:

1. confirm data validity and calibration status;
2. establish the indication coordinate and sound path;
3. compare with known component geometry;
4. examine response dynamics and orientation;
5. test alternative wave paths and mode conversions;
6. compare with credible material or damage mechanisms;
7. obtain corroborating scans or another method;
8. assign classification and confidence within authorised terminology.

Skipping the first step risks analysing an acquisition error. Skipping alternative explanations creates confirmation bias.

13.2 False and non-relevant indications

A false indication is produced by the examination system rather than a physical condition of interest, such as electrical noise, loose connections, multiplexing artefacts, ghost echoes or encoder errors. A non-relevant indication is produced by a real feature that is not a reportable discontinuity under the inspection objective, such as known geometry.

The distinction matters. A false indication calls for system correction. A non-relevant response calls for justified classification. Both should be traceable when they could affect review.

13.3 Geometry and mode conversion

Corners, radii, counterbores, holes, threads and backing can produce direct reflections, creeping waves, mode-converted legs and reverberation. Calculate alternative paths rather than forcing every signal into the intended mode. Changes in probe position and skew can test whether the response follows the predicted feature.

13.4 Data views and corroboration

An A-scan preserves time-amplitude detail. Encoded B-, C- and D-scan views add position and projection. Sectorial or reconstructed views can show relationships more clearly, but each depends on setup geometry and velocity. Corroboration should return to raw A-scans and independent scan directions where possible.

13.5 Bias controls

Useful controls include blind review, independent second analysis, standardised indication worksheets, predefined classification rules and separation of analysis from acceptance disposition. Analysts should record evidence that contradicts their preferred explanation. Confidence terms should be defined rather than improvised.

13.6 Ambiguous-case rule

If plausible competing explanations have different safety or acceptance consequences, the Level 3 should not select the convenient one without additional evidence. The response may be reported as unresolved, followed by rescanning, surface examination, radiography, profile measurement, metallurgical review or engineering assessment as appropriate.

Failure modes and decision control

Weak practice commonly includes analysing before confirming data quality; fitting every signal to the nominal beam path; calling known geometry false rather than non-relevant; using colour maps as amplitude truth; ignoring contradictory evidence; and resolving ambiguity by production preference.