Austenitic Girth Welds: UT Substituted for RT
An EPC contractor directs that girth welds in a 304L stainless line (323.9 mm OD × 12.7 mm wall) be examined by manual UT instead of RT; the line is congested and radiography would need a night shift and a 30 m exclusion zone. The Level II uses a 5 MHz, 70°, 6 mm single-crystal shear-wave probe and builds a DAC from a carbon steel block with a 1.5 mm side-drilled hole. He reports the welds 'acceptable; some geometric noise in the weld volume.' A second inspector, using a 2.25 MHz, 45°, dual-crystal focused probe, gets scattered signals from the weld body that he cannot size, and calls for a cut-out. The owner asks the Level III to arbitrate before the line is insulated.
1. Assess the first inspector's technique. Is his 'no indications' a valid result, and what is the 'geometric noise' he reports?
Read the worked answer
The result is not valid, and the 'geometric noise' is very likely grain scattering from the weld body being mislabelled.
What is happening acoustically in an austenitic weld:
- Columnar, dendritic solidification structure with elastic anisotropy. Grains are large and aligned with the heat flow, so the beam is scattered, skewed and split; energy that should return from a flaw is redirected.
- Velocity differences. Shear velocity in austenitic weld metal is roughly 3100 m/s against ~3230 m/s in carbon steel, so a carbon-steel DAC places every indication at the wrong depth and the wrong angle is used to reach it.
- Attenuation and mode conversion at the fusion line; the beam can cross into the weld and come back as a different mode or from a grain boundary.
- 5 MHz is too high for this microstructure; the wavelength is small enough that grain boundaries scatter strongly, and the resulting 'grass' is exactly what an operator calls geometric noise when he has no signal-to-noise criterion to test against.
- 70° with a 6 mm crystal in a 12.7 mm wall with internal misalignment produces legitimate geometry signals (root, counterbore, high-low) that cannot be separated from flaws by amplitude alone.
Conclusion: 'acceptable, with noise' is a statement that the technique did not achieve detection. A technique that cannot demonstrate signal-to-noise cannot report a weld as acceptable - it can only report that it does not know. The second inspector is closer to right, but his inability to size means his result is also not yet an acceptance decision.
2. What must the Level III do before any UT result on these welds can be used for acceptance, and under what conditions would you refuse the UT-for-RT substitution entirely?
Read the worked answer
Prerequisites, in order:
- Written procedure qualified by demonstration. A mock-up of the same material, diameter, wall thickness, weld prep and internal misalignment, containing known flaws (lack of sidewall fusion, toe crack, slag, root defect) at known depths. The demonstration must show detection and sizing, not just 'it beeps'.
- Technique matched to the metallurgy. Lower frequency (1–2.5 MHz), dual-crystal/transmit-receive or focused probes to control the beam in the weld, 45° preferred for near-vertical planar flaws, scanning from both sides and both surfaces, and a reference block in austenitic material or a validated correction.
- Personnel qualified to that procedure. A general UT Level II certificate does not confer competence on austenitic weld examination; the demonstration record must name the operators.
- Approval of the substitution. UT in lieu of RT is a contract and code decision, requiring the owner's and the AI's written agreement; the acceptance criteria must be defined for UT, not borrowed from the RT limits.
- Complementary coverage. PT/MT after hydrotest for surface-breaking toe cracks, which volume UT is poor at finding in this geometry.
Refuse the substitution if: the demonstration cannot show adequate signal-to-noise and sizing accuracy on the mock-up; the acceptance basis for UT is undefined; or the misalignment is so large that root geometry masks the region of interest. In those cases the answer is RT (with the exclusion zone and night shift priced in) or a mechanised/PAUT solution with a demonstrated capability - not a manual scan reported as clean.
3. The contractor argues that 12.7 mm is thin, radiation is a scheduling problem, and 'UT found nothing wrong.' How do you answer, and what is the practical outcome for the line?
Read the worked answer
Thin wall makes the problem harder, not easier, for manual UT here. At 12.7 mm the weld cap, root profile and internal misalignment occupy a large fraction of the sound path, and the austenitic microstructure degrades the beam in exactly the region of interest. 'UT found nothing' from an unqualified technique carries no more weight than 'RT found nothing' from a film with the wrong density - the method was not capable, so the null result is uninformative.
Practical outcome:
- Re-examine a defined sample of the welds with the demonstrated procedure, plus the welds the contractor most wants to keep, before insulation.
- Where the second inspector's scattered signals are located, use a sizing technique (TOFD or PAUT with a qualified procedure) to convert 'can't size' into a height and a length; then apply the agreed acceptance criteria.
- If a cut-out is confirmed, treat it as a process signal: review the welding procedure, the purge, the ferrite content and the welder's other joints, and widen the sample.
- Insulation goes on only after the examination basis is agreed in writing. Once insulated, the option to radiograph disappears and the argument is settled by default - which is the outcome to avoid.